research paper on effects of deforestation

• Search Menu
• Sign in through your institution
• Advance Articles
• Author Guidelines
• Submission Site
• Open Access
• Self-Archiving Policy
• Why Submit?
• About Journal of Forestry
• About Society of American Foresters
• Editorial Board
• Advertising & Corporate Services
• Journals Career Network
• Journals on Oxford Academic
• Books on Oxford Academic

Article Contents

Addressing the complexities and impacts of deforestation.

• < Previous

Addressing the Complexities and Impacts of Deforestation

• Article contents
• Figures & tables
• Supplementary Data

John Stanturf, Addressing the Complexities and Impacts of Deforestation, Journal of Forestry , Volume 115, Issue 4, July 2017, Pages 319–320, https://doi.org/10.5849/jof.2016-069

• Permissions Icon Permissions

Deforestation, the removal of forest canopy and conversion to another land use, has engendered significant international efforts to stop or reverse loss of tree cover. Seen today, deforestation would appear to be an issue of tropical forests even though the cumulative loss of forestland over the last 5,000 years is estimated globally at 1.8 billion ha. In Global Deforestation , Runyan and D'Odorico provide a thorough analysis of the literature, and to say that their book is broad in scope is perhaps an understatement. It is multidisciplinary, addressing perspectives from the biophysical as well as social sciences and examines processes at multiple spatial and temporal scales. The book addresses the complexity of deforestation, including considering the benefits of deforestation (mostly private and short-term) as well as the costs (social and long-term). A further strength of the book is that effects examined go beyond the deforested area, for example, including the potential regional climatic effects of deforestation and afforestation. This comprehensive approach, which should appeal to researchers and students, may seem a bit too “academic” for managers. The summaries of main points provided by the authors within and at the end of chapters, however, should appeal to all readers.

Global Deforestation is a relatively short book, comprising just six chapters but well referenced, and some figures are reproduced in color at the end. It begins with an introduction to the drivers of deforestation, including a historical perspective that describes the trajectory of clearing and in some cases recovery by region. Potential future drivers (principally meeting food security and dietary preferences of an expanding human population) are introduced in Chapter 1 (“Introduction: Patterns and Drivers”) and explored in-depth in Chapter 6 (“Synthesis and Future Impacts of Deforestation”). The intervening chapters address impacts of deforestation on hydrology and climate, biogeochemistry, ecological irreversibility, and, of course, economic impacts and drivers. Each chapter on impacts has a similar structure, first describing underlying processes (or baseline conditions) and then the effects of deforestation on important aspects of a process. For example, Chapter 2, “Hydrologic and Climatic Impacts,” has sections describing precipitation and forest canopies and infiltration and runoff generation before launching into the effects of deforestation on the hydrologic responses of flood dynamics and water yields. Also considered in this chapter are the effects of forests on groundwater and deforestation on wetlands, evaporation and transpiration and the effects of forests on precipitation and microclimate, and the effects of deforestation on large-scale climate. The chapter effectively presents a minireview of the importance of forests in climate change discussions that venture beyond carbon sequestration.

The chapter on irreversibility and ecosystem impacts ties the previous chapters to current thinking on disturbance and ecosystem resilience and resistance. Specific feedbacks that modify resource availability are presented in terms of operating principles, geographic extent, available observational evidence, modeling techniques to address the complexity, and management implications. The situations considered are potentially where deforestation leads to conditions that cannot support forest vegetation, such as following landslides, wildfire, exposure to freezing events, or increased soil salinity. These are cases where an abrupt transition to a seemingly irreversible state, with loss of economic and ecological functions, has significant implications for restoration efforts. Unfortunately, the authors do not draw out these implications in greater detail; more on this later.

Chapter 5, “Economic Impacts and Drivers of Deforestation,” is where many authors would begin their story; to the credit of Runyan and D'Odorico, however, they lay the biophysical groundwork first before venturing into the very complex social context of land-use decisions. They use the framework that is well established in the literature of immediate (proximate) versus underlying (ultimate) causes and note that agricultural expansion is the primary cause of deforestation, although other factors such as road building that improves access to markets can determine the historical development of reduced forest cover in a region. Two paradoxes of importance to conservation and sustainable development emerge regarding improved agricultural productivity and secure tenure. First, agricultural intensification has been promoted as a way to reduce pressure on native forests (i.e., avoided deforestation). In actuality, yield increases and labor-saving technology may lead to increased deforestation because agriculture is more profitable. Second, insecure land tenure, typical of many areas in the tropics where deforestation is occurring, could lead a farmer to decline to invest in improvements or long-term investments such as forestry; thus, secure tenure could lead to greater emphasis on future benefits from value accumulated in timber. Conversely, secure property rights could induce the farmer to invest in agriculture and use the land as collateral for loans to cover the costs of improvement. In both instances, other sociocultural conditions play a role in which scenario plays out.

On the whole, Global Deforestation is a valuable addition to the literature on an important international issue of great interest to foresters. Although it may be slow going for some readers in parts due to the comprehensive treatment of technical issues, the brief summaries within some chapters and at the end of each provide needed rest from the details and helpful reminders of the larger picture. Considering the benefits to some people and costs to others makes the case that deforestation occurs for economic and social reasons and is not due to moral failure or ignorance. Reversing deforestation, therefore, requires addressing the immediate and underlying drivers, which is shown to be far from a simple task and one that requires more effort at multiple levels of society and not just in communities near tropical forests. I would be remiss as a reviewer if I couldn't identify two shortcomings. Even though the authors define deforestation clearly in the Introduction as involving long-term reduction of forest cover to below 10%, they still treat logging in several places as a general deforestation factor. This will be a burr under the saddle of most foresters. To their credit, the authors place relatively little emphasis on this, however. Finally, I would have appreciated the authors addressing current international efforts to address and reverse deforestation. Even though REDD+ is briefly mentioned, other efforts such as the Bonn Challenge to restore 150 million ha of forested landscapes by 2020 and 350 million ha by 2030 or the Convention on Biological Diversity Aichi Targets are ignored. Giving some insight into how to ensure these efforts are successful would have been a valuable conclusion to an interesting book.

Email alerts

Citing articles via.

• Recommend to Your Librarian
• Advertising and Corporate Services

Affiliations

• Online ISSN 1938-3746
• Print ISSN 0022-1201
• Copyright © 2024 Society of American Foresters
• About Oxford Academic
• Publish journals with us
• University press partners
• What we publish
• New features  
• Open access
• Institutional account management
• Rights and permissions
• Get help with access
• Accessibility
• Advertising
• Media enquiries
• Oxford University Press
• Oxford Languages
• University of Oxford

Oxford University Press is a department of the University of Oxford. It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide

• Copyright © 2024 Oxford University Press
• Cookie settings
• Cookie policy
• Privacy policy
• Legal notice

This Feature Is Available To Subscribers Only

Sign In or Create an Account

This PDF is available to Subscribers Only

For full access to this pdf, sign in to an existing account, or purchase an annual subscription.

The Effects of Deforestation: its Role on Climate Change and Impact on Ecosystems

Original Research 13 March 2023 The impact of meteorological changes on the quality of life regarding thermal comfort in the Amazon region Adrian Felipe dos Santos ,  4 more  and  Fernando Augusto Ribeiro Costa 1,622 views 1 citations

Original Research 20 October 2022 Towards sustainable adaptation: A tool for estimating adaptation costs to climate change for smallholder farmers Dumisani Shoko Kori  and  Edmore Kori 1,323 views 2 citations

Loading... Original Research 12 October 2022 RED, PEF, and EPD: Conflicting rules for determining the carbon footprint of biofuels give unclear signals to fuel producers and customers Miguel Brandão ,  6 more  and  Tomas Rydberg 6,872 views 4 citations

Loading... Original Research 08 February 2022 Understanding the Greenhouse Gas Impact of Deforestation Fires in Indonesia and Brazil in 2019 and 2020 Aparajita Datta  and  Ramanan Krishnamoorti 11,172 views 11 citations

November 13, 2012

Deforestation and Its Extreme Effect on Global Warming

From logging, agricultural production and other economic activities, deforestation adds more atmospheric CO2 than the sum total of cars and trucks on the world's roads

Land cleared of forest by timber industry.

Nazar Abbas Getty Images

Dear EarthTalk : Is it true that cutting and burning trees adds more global warming pollution to the atmosphere than all the cars and trucks in the world combined? — Mitchell Vale, Houston

By most accounts, deforestation in tropical rainforests adds more carbon dioxide to the atmosphere than the sum total of cars and trucks on the world’s roads. According to the World Carfree Network (WCN), cars and trucks account for about 14 percent of global carbon emissions, while most analysts attribute upwards of 15 percent to deforestation.

The reason that logging is so bad for the climate is that when trees are felled they release the carbon they are storing into the atmosphere, where it mingles with greenhouse gases from other sources and contributes to global warming accordingly. The upshot is that we should be doing as much to prevent deforestation as we are to increase fuel efficiency and reduce automobile usage.

On supporting science journalism

If you're enjoying this article, consider supporting our award-winning journalism by subscribing . By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.

According to the Environmental Defense Fund (EDF), a leading green group, 32 million acres of tropical rainforest were cut down each year between 2000 and 2009—and the pace of deforestation is only increasing. “Unless we change the present system that rewards forest destruction, forest clearing will put another 200 billion tons of carbon into the atmosphere in coming decades…,” says EDF.

“Any realistic plan to reduce global warming pollution sufficiently—and in time—to avoid dangerous consequences must rely in part on preserving tropical forests,” reports EDF. But it’s hard to convince the poor residents of the Amazon basin and other tropical regions of the world to stop cutting down trees when the forests are still worth more dead than alive. “Conservation costs money, while profits from timber, charcoal, pasture and cropland drive people to cut down forests,” adds EDF. Exacerbating global warming isn’t the only negative impact of tropical deforestation. It also wipes out biodiversity: More than half of the world’s plant and animal species live in tropical rainforests.

One way some tropical countries are reducing deforestation is through participation in the United Nations’ Reducing Emissions from Deforestation and Forest Degradation (REDD) program. REDD essentially works to establish incentives for the people who care for the forest to manage it sustainably while still being able to benefit economically. Examples include using less land (and therefore cutting fewer trees) for activities such as coffee growing and meat and milk production. Participating nations can then accrue and sell carbon pollution credits when they can prove they have lowered deforestation below a baseline. The REDD program has channeled over $117 million in direct financial aid and educational support into national deforestation reduction efforts in 44 developing countries across Africa, Asia and Latin America since its 2008 inception.

Brazil is among the countries embracing REDD among other efforts to reduce carbon emissions. Thanks to the program, Brazil has slowed deforestation within its borders by 40 percent since 2008 and is on track to achieve an 80 percent reduction by 2020. Environmentalists are optimistic that the initial success of REDD in Brazil bodes well for reducing deforestation in other parts of the tropics as well.

CONTACTS : WCN, www.worldcarfree.net; EDF, www.edf.org; REDD, www.un-redd.org .

EarthTalk® is written and edited by Roddy Scheer and Doug Moss and is a registered trademark of E - The Environmental Magazine (www.emagazine.com). Send questions to: [email protected] . Subscribe : www.emagazine.com/subscribe . Free Trial Issue : www.emagazine.com/trial .

Suggestions or feedback?

MIT News | Massachusetts Institute of Technology

• Machine learning
• Sustainability
• Black holes
• Classes and programs

Departments

• Aeronautics and Astronautics
• Brain and Cognitive Sciences
• Architecture
• Political Science
• Mechanical Engineering

Centers, Labs, & Programs

• Abdul Latif Jameel Poverty Action Lab (J-PAL)
• Picower Institute for Learning and Memory
• Lincoln Laboratory
• School of Architecture + Planning
• School of Engineering
• School of Humanities, Arts, and Social Sciences
• Sloan School of Management
• School of Science
• MIT Schwarzman College of Computing

How to tackle the global deforestation crisis

Press contact :, media download.

*Terms of Use:

Images for download on the MIT News office website are made available to non-commercial entities, press and the general public under a Creative Commons Attribution Non-Commercial No Derivatives license . You may not alter the images provided, other than to crop them to size. A credit line must be used when reproducing images; if one is not provided below, credit the images to "MIT."

Previous image Next image

Imagine if France, Germany, and Spain were completely blanketed in forests — and then all those trees were quickly chopped down. That’s nearly the amount of deforestation that occurred globally between 2001 and 2020, with profound consequences.

Deforestation is a major contributor to climate change, producing between 6 and 17 percent of global greenhouse gas emissions, according to a 2009 study. Meanwhile, because trees also absorb carbon dioxide, removing it from the atmosphere, they help keep the Earth cooler. And climate change aside, forests protect biodiversity.

“Climate change and biodiversity make this a global problem, not a local problem,” says MIT economist Ben Olken. “Deciding to cut down trees or not has huge implications for the world.”

But deforestation is often financially profitable, so it continues at a rapid rate. Researchers can now measure this trend closely: In the last quarter-century, satellite-based technology has led to a paradigm change in charting deforestation. New deforestation datasets, based on the Landsat satellites, for instance, track forest change since 2000 with resolution at 30 meters, while many other products now offer frequent imaging at close resolution.

“Part of this revolution in measurement is accuracy, and the other part is coverage,” says Clare Balboni, an assistant professor of economics at the London School of Economics (LSE). “On-site observation is very expensive and logistically challenging, and you’re talking about case studies. These satellite-based data sets just open up opportunities to see deforestation at scale, systematically, across the globe.”

Balboni and Olken have now helped write a new paper providing a road map for thinking about this crisis. The open-access article, “ The Economics of Tropical Deforestation ,” appears this month in the Annual Review of Economics . The co-authors are Balboni, a former MIT faculty member; Aaron Berman, a PhD candidate in MIT’s Department of Economics; Robin Burgess, an LSE professor; and Olken, MIT’s Jane Berkowitz Carlton and Dennis William Carlton Professor of Microeconomics. Balboni and Olken have also conducted primary research in this area, along with Burgess.

So, how can the world tackle deforestation? It starts with understanding the problem.

Replacing forests with farms

Several decades ago, some thinkers, including the famous MIT economist Paul Samuelson in the 1970s, built models to study forests as a renewable resource; Samuelson calculated the “maximum sustained yield” at which a forest could be cleared while being regrown. These frameworks were designed to think about tree farms or the U.S. national forest system, where a fraction of trees would be cut each year, and then new trees would be grown over time to take their place.

But deforestation today, particularly in tropical areas, often looks very different, and forest regeneration is not common.

Indeed, as Balboni and Olken emphasize, deforestation is now rampant partly because the profits from chopping down trees come not just from timber, but from replacing forests with agriculture. In Brazil, deforestation has increased along with agricultural prices; in Indonesia, clearing trees accelerated as the global price of palm oil went up, leading companies to replace forests with palm tree orchards.

All this tree-clearing creates a familiar situation: The globally shared costs of climate change from deforestation are “externalities,” as economists say, imposed on everyone else by the people removing forest land. It is akin to a company that pollutes into a river, affecting the water quality of residents.

“Economics has changed the way it thinks about this over the last 50 years, and two things are central,” Olken says. “The relevance of global externalities is very important, and the conceptualization of alternate land uses is very important.” This also means traditional forest-management guidance about regrowth is not enough. With the economic dynamics in mind, which policies might work, and why?

The search for solutions

As Balboni and Olken note, economists often recommend “Pigouvian” taxes (named after the British economist Arthur Pigou) in these cases, levied against people imposing externalities on others. And yet, it can be hard to identify who is doing the deforesting.

Instead of taxing people for clearing forests, governments can pay people to keep forests intact. The UN uses Payments for Environmental Services (PES) as part of its REDD+ (Reducing Emissions from Deforestation and forest Degradation) program. However, it is similarly tough to identify the optimal landowners to subsidize, and these payments may not match the quick cash-in of deforestation. A 2017 study in Uganda showed PES reduced deforestation somewhat; a 2022 study in Indonesia found no reduction; another 2022 study, in Brazil, showed again that some forest protection resulted.

“There’s mixed evidence from many of these [studies],” Balboni says. These policies, she notes, must reach people who would otherwise clear forests, and a key question is, “How can we assess their success compared to what would have happened anyway?”

Some places have tried cash transfer programs for larger populations. In Indonesia, a 2020 study found such subsidies reduced deforestation near villages by 30 percent. But in Mexico, a similar program meant more people could afford milk and meat, again creating demand for more agriculture and thus leading to more forest-clearing.

At this point, it might seem that laws simply banning deforestation in key areas would work best — indeed, about 16 percent of the world’s land overall is protected in some way. Yet the dynamics of protection are tricky. Even with protected areas in place, there is still “leakage” of deforestation into other regions. 

Still more approaches exist, including “nonstate agreements,” such as the Amazon Soy Moratorium in Brazil, in which grain traders pledged not to buy soy from deforested lands, and reduced deforestation without “leakage.”

Also, intriguingly, a 2008 policy change in the Brazilian Amazon made agricultural credit harder to obtain by requiring recipients to comply with environmental and land registration rules. The result? Deforestation dropped by up to 60 percent over nearly a decade. 

Politics and pulp

Overall, Balboni and Olken observe, beyond “externalities,” two major challenges exist. One, it is often unclear who holds property rights in forests. In these circumstances, deforestation seems to increase. Two, deforestation is subject to political battles.

For instance, as economist Bard Harstad of Stanford University has observed, environmental lobbying is asymmetric. Balboni and Olken write: “The conservationist lobby must pay the government in perpetuity … while the deforestation-oriented lobby need pay only once to deforest in the present.” And political instability leads to more deforestation because “the current administration places lower value on future conservation payments.”

Even so, national political measures can work. In the Amazon from 2001 to 2005, Brazilian deforestation rates were three to four times higher than on similar land across the border, but that imbalance vanished once the country passed conservation measures in 2006. However, deforestation ramped up again after a 2014 change in government. Looking at particular monitoring approaches, a study of Brazil’s satellite-based Real-Time System for Detection of Deforestation (DETER), launched in 2004, suggests that a 50 percent annual increase in its use in municipalities created a 25 percent reduction in deforestation from 2006 to 2016.

How precisely politics matters may depend on the context. In a 2021 paper, Balboni and Olken (with three colleagues) found that deforestation actually decreased around elections in Indonesia. Conversely, in Brazil, one study found that deforestation rates were 8 to 10 percent higher where mayors were running for re-election between 2002 and 2012, suggesting incumbents had deforestation industry support.

“The research there is aiming to understand what the political economy drivers are,” Olken says, “with the idea that if you understand those things, reform in those countries is more likely.”

Looking ahead, Balboni and Olken also suggest that new research estimating the value of intact forest land intact could influence public debates. And while many scholars have studied deforestation in Brazil and Indonesia, fewer have examined the Democratic Republic of Congo, another deforestation leader, and sub-Saharan Africa.

Deforestation is an ongoing crisis. But thanks to satellites and many recent studies, experts know vastly more about the problem than they did a decade or two ago, and with an economics toolkit, can evaluate the incentives and dynamics at play.

“To the extent that there’s ambuiguity across different contexts with different findings, part of the point of our review piece is to draw out common themes — the important considerations in determining which policy levers can [work] in different circumstances,” Balboni says. “That’s a fast-evolving area. We don’t have all the answers, but part of the process is bringing together growing evidence about [everything] that affects how successful those choices can be.”

Share this news article on:

Related links.

• Department of Economics

Related Topics

• Climate change
• Agriculture
• International relations
• School of Humanities Arts and Social Sciences

Related Articles

A clean alternative to one of the world’s most common ingredients

Keeping humanity central to solving climate change

3 Questions: Benjamin Olken on the economic impact of climate change

The economic cost of increased temperatures

Previous item Next item

More MIT News

Study: Transparency is often lacking in datasets used to train large language models

Read full story →

How MIT’s online resources provide a “highly motivating, even transformative experience”

Students learn theater design through the power of play

A framework for solving parabolic partial differential equations

Designing better delivery for medical therapies

Making a measurable economic impact

• More news on MIT News homepage →

Massachusetts Institute of Technology 77 Massachusetts Avenue, Cambridge, MA, USA

• Map (opens in new window)
• Events (opens in new window)
• People (opens in new window)
• Careers (opens in new window)
• Accessibility
• Social Media Hub
• MIT on Facebook
• MIT on YouTube
• MIT on Instagram

• DOI: 10.1515/IGBP-2015-0003
• Corpus ID: 131650904

Deforestation And Its Impacts On Climate Change An Overview Of Pakistan

• Arshad Ali , Somana Riaz , S. Iqbal
• Published 2014
• Environmental Science, Geography

Figures from this paper

13 Citations

Features of climatic zones, climate change and natural hazards, impacts of climate change on agricultural sector of pakistan: status, consequences, and adoption options, long term hydrological and environmental monitoring of the stryi river using remote sensing data and gis technologies, carbon stock estimation on oil palm plantations and oil palm-based agroforestry in gunung mas regency, salinity and the health of the poor in coastal bangladesh, after the paris agreement: how india can use climate financing to implement a sustainable clean cookstove program, household production and time allocation behaviour to climate change of nepal, assessing print media coverage on environmental issues in pakistan: a study of major newspapers in pakistan, s atellite r emote s ensing f orest m anagement in s elangor r iver b asin a rea from 1990 t o 2020, advocacy journalism and climate justice in a global southern country, 21 references, managing the indus river basin in light of climate change, spatial patterns and variation of suspended sediment yield in the upper indus river basin, northern pakistan, five centuries of upper indus river flow from tree rings, modeling snowmelt-runoff under climate scenarios in the hunza river basin, karakoram range, northern pakistan, catastrophic landslides and their effects on the upper indus streams, karakoram himalaya, northern pakistan, quaternary glacial history of the karakoram mountains and northwest himalayas: a review, flooding in the indus river basin — a spatiotemporal analysis of precipitation records, geomorphological analysis of the drainage system on the growing makran accretionary wedge, projected climate regime shift under future global warming from multi-model, multi-scenario cmip5 simulations, sea surface temperature anomalies at the mediterranean coast of turkey (period of 1968-2010), related papers.

Showing 1 through 3 of 0 Related Papers

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Published: 01 September 2021

How deregulation, drought and increasing fire impact Amazonian biodiversity

• Xiao Feng   ORCID: orcid.org/0000-0003-4638-3927 1   na1 ,
• Cory Merow 2   na1 ,
• Zhihua Liu   ORCID: orcid.org/0000-0002-0086-5659 3   na1 ,
• Daniel S. Park   ORCID: orcid.org/0000-0003-2783-530X 4 , 5   na1 ,
• Patrick R. Roehrdanz   ORCID: orcid.org/0000-0003-4047-5011 6   na1 ,
• Brian Maitner   ORCID: orcid.org/0000-0002-2118-9880 2   na1 ,
• Erica A. Newman 7 , 8   na1 ,
• Brad L. Boyle 7 , 9 ,
• Aaron Lien 8 , 10 ,
• Joseph R. Burger 7 , 8 , 11 ,
• Mathias M. Pires 12 ,
• Paulo M. Brando   ORCID: orcid.org/0000-0001-8952-7025 13 , 14 , 15 ,
• Mark B. Bush   ORCID: orcid.org/0000-0001-6894-8613 16 ,
• Crystal N. H. McMichael   ORCID: orcid.org/0000-0002-1064-1499 17 ,
• Danilo M. Neves 18 ,
• Efthymios I. Nikolopoulos 19 ,
• Scott R. Saleska 7 ,
• Lee Hannah 6 ,
• David D. Breshears   ORCID: orcid.org/0000-0001-6601-0058 10 ,
• Tom P. Evans   ORCID: orcid.org/0000-0003-4591-1011 20 ,
• José R. Soto 10 ,
• Kacey C. Ernst 21 &
• Brian J. Enquist 7 , 22   na1  

Nature volume  597 ,  pages 516–521 ( 2021 ) Cite this article

17k Accesses

66 Citations

658 Altmetric

Metrics details

• Biodiversity
• Biogeography

Biodiversity contributes to the ecological and climatic stability of the Amazon Basin 1 , 2 , but is increasingly threatened by deforestation and fire 3 , 4 . Here we quantify these impacts over the past two decades using remote-sensing estimates of fire and deforestation and comprehensive range estimates of 11,514 plant species and 3,079 vertebrate species in the Amazon. Deforestation has led to large amounts of habitat loss, and fires further exacerbate this already substantial impact on Amazonian biodiversity. Since 2001, 103,079–189,755 km 2 of Amazon rainforest has been impacted by fires, potentially impacting the ranges of 77.3–85.2% of species that are listed as threatened in this region 5 . The impacts of fire on the ranges of species in Amazonia could be as high as 64%, and greater impacts are typically associated with species that have restricted ranges. We find close associations between forest policy, fire-impacted forest area and their potential impacts on biodiversity. In Brazil, forest policies that were initiated in the mid-2000s corresponded to reduced rates of burning. However, relaxed enforcement of these policies in 2019 has seemingly begun to reverse this trend: approximately 4,253–10,343 km 2 of forest has been impacted by fire, leading to some of the most severe potential impacts on biodiversity since 2009. These results highlight the critical role of policy enforcement in the preservation of biodiversity in the Amazon.

This is a preview of subscription content, access via your institution

Access options

Access Nature and 54 other Nature Portfolio journals

Get Nature+, our best-value online-access subscription

24,99 € / 30 days

cancel any time

Subscribe to this journal

Receive 51 print issues and online access

185,98 € per year

only 3,65 € per issue

Buy this article

• Purchase on SpringerLink
• Instant access to full article PDF

Prices may be subject to local taxes which are calculated during checkout

Similar content being viewed by others

Shifting agriculture is the dominant driver of forest disturbance in threatened forest species’ ranges

Implications of the 2019–2020 megafires for the biogeography and conservation of Australian vegetation

Impact of 2019–2020 mega-fires on Australian fauna habitat

Data availability.

The plant occurrences from the BIEN database are accessible using the RBIEN package ( https://github.com/bmaitner/RBIEN ). The climatic data are accessible from http://worldclim.org and the soil data are available from http://soilgrids.org . MODIS active fire and burned area products are available at http://modis-fire.umd.edu . The MODIS Vegetation Continuous Fields data are publicly available from https://lpdaac.usgs.gov/products/mod44bv006/ . The annual forest loss layers are available from http://earthenginepartners.appspot.com/science-2013-global-forest . The plant range maps are accessible at https://github.com/shandongfx/paper_Amazon_biodiversity_2021 . The vertebrate range maps are available from https://www.iucnredlist.org/resources/spatial-data-download . The SPEI data are available from SPEI Global Drought Monitor ( https://spei.csic.es/map ).

Code availability

The code to process the remote-sensing data is available at https://github.com/shandongfx/paper_Amazon_biodiversity_2021 .

Yachi, S. & Loreau, M. Biodiversity and ecosystem productivity in a fluctuating environment: the insurance hypothesis. Proc. Natl Acad. Sci. USA 96 , 1463–1468 (1999).

Article   ADS   CAS   PubMed   PubMed Central   Google Scholar  

Oliver, T. H. et al. Biodiversity and resilience of ecosystem functions. Trends Ecol. Evol. 30 , 673–684 (2015).

Article   PubMed   Google Scholar  

Barlow, J., Berenguer, E., Carmenta, R. & França, F. Clarifying Amazonia’s burning crisis. Glob. Change Biol. 9 , 1 (2019).

Google Scholar  

Brando, P. M. et al. The gathering firestorm in southern Amazonia. Sci. Adv. 6 , eaay1632 (2020).

IUCN. IUCN Red List of Threatened Species version 6.2 . https://www.iucnredlist.org/ (2019).

Flores, M. et al. WWF’s Living Amazon Initiative (Grambs Corporación Gráfica, 2010).

Hubbell, S. P. et al. How many tree species are there in the Amazon and how many of them will go extinct? Proc. Natl Acad. Sci. USA 105 Suppl. 1 , 11498–11504 (2008).

Nepstad, D. C., Stickler, C. M., Filho, B. S.- & Merry, F. Interactions among Amazon land use, forests and climate: prospects for a near-term forest tipping point. Philos. Trans. R. Soc. Lond. B 363 , 1737–1746 (2008).

Article   Google Scholar  

Rankin-de-Mérona, J. M. et al. Preliminary results of a large-scale tree inventory of upland rain forest in the Central Amazon. Acta Amazon. 22 , 493–534 (1992).

Sakschewski, B. et al. Resilience of Amazon forests emerges from plant trait diversity. Nat. Clim. Change 6 , 1032–1036 (2016).

Article   ADS   Google Scholar  

Poorter, L. et al. Biomass resilience of Neotropical secondary forests. Nature 530 , 211–214 (2016).

Article   ADS   CAS   PubMed   Google Scholar  

Beisner, B. E., Haydon, D. T. & Cuddington, K. Alternative stable states in ecology. Front. Ecol. Environ. 1 , 376–382 (2003).

Lovejoy, T. E. & Nobre, C. Amazon tipping point. Sci. Adv. 4 , eaat2340 (2018).

Article   ADS   PubMed   PubMed Central   Google Scholar  

Veldman, J. W. Clarifying the confusion: old-growth savannahs and tropical ecosystem degradation. Philos. Trans. R. Soc. Lond. B 371 , (2016).

Arruda, D., Candido, H. G. & Fonseca, R. Amazon fires threaten Brazil’s agribusiness. Science 365 , 1387 (2019).

Article   ADS   PubMed   CAS   Google Scholar  

Ter Steege, H. et al. Estimating the global conservation status of more than 15,000 Amazonian tree species. Sci. Adv. 1 , e1500936 (2015).

Gomes, V. H. F., Vieira, I. C. G., Salomão, R. P. & ter Steege, H. Amazonian tree species threatened by deforestation and climate change. Nat. Clim. Change 9 , 547–553 (2019).

Brando, P. et al. Amazon wildfires: scenes from a foreseeable disaster. Flora 268 , 151609 (2020).

Balch, J. K. et al. The susceptibility of southeastern Amazon forests to fire: insights from a large-scale burn experiment. Bioscience 65 , 893–905 (2015).

Barlow, J. et al. The critical importance of considering fire in REDD+ programs. Biol. Conserv. 154 , 1–8 (2012).

Cochrane, M. A. & Schulze, M. D. Fire as a recurrent event in tropical forests of the eastern Amazon: effects on forest structure, biomass, and species composition. Biotropica 31 , 2–16 (1999).

Brando, P. M. et al. Prolonged tropical forest degradation due to compounding disturbances: Implications for CO 2 and H 2 O fluxes. Glob. Change Biol. 25 , 2855–2868 (2019).

Barlow, J. & Peres, C. A. Fire-mediated dieback and compositional cascade in an Amazonian forest. Philos. Trans. R. Soc. Lond. B 363 , 1787–1794 (2008).

Cochrane, M. Tropical Fire Ecology: Climate Change, Land Use and Ecosystem Dynamics (Springer, 2010).

Uhl, C. & Kauffman, J. B. Deforestation, fire susceptibility, and potential tree responses to fire in the eastern Amazon. Ecology 71 , 437–449 (1990).

Cochrane, M. A. Fire science for rainforests. Nature 421 , 913–919 (2003).

Cochrane, M. A. & Laurance, W. F. Synergisms among fire, land use, and climate change in the Amazon. Ambio 37 , 522–527 (2008).

Nepstad, D. C. et al. Large-scale impoverishment of Amazonian forests by logging and fire. Nature 398 , 505–508 (1999).

Article   ADS   CAS   Google Scholar  

Aragão, L. E. O. C. et al. 21st Century drought-related fires counteract the decline of Amazon deforestation carbon emissions. Nat. Commun. 9 , 536 (2018).

Article   ADS   PubMed   PubMed Central   CAS   Google Scholar  

Nepstad, D. et al. Slowing Amazon deforestation through public policy and interventions in beef and soy supply chains. Science 344 , 1118–1123 (2014).

Hope, M. The Brazilian development agenda driving Amazon devastation. Lancet Planet. Health 3 , e409–e411 (2019).

Brown, J. H. On the relationship between abundance and distribution of species. Am. Nat. 124 , 255–279 (1984).

Barnagaud, J.-Y. et al. Ecological traits influence the phylogenetic structure of bird species co-occurrences worldwide. Ecol. Lett. 17 , 811–820 (2014).

Šímová, I. et al. Spatial patterns and climate relationships of major plant traits in the New World differ between woody and herbaceous species. J. Biogeogr. 45 , 895–916 (2018).

Enquist, B. J. et al. The commonness of rarity: Global and future distribution of rarity across land plants. Sci. Adv. 5 , eaaz0414 (2019).

May, P. H., Gebara, M. F., de Barcellos, L. M., Rizek, M. B. & Millikan, B. The Context of REDD+ in Brazil: Drivers, Agents, and Institutions, 3rd edition, https://doi.org/10.17528/cifor/006338 (Center for International Forestry Research, 2016).

Neves, D. M., Dexter, K. G., Baker, T. R., Coelho de Souza, F. & Oliveira-Filho, A. T. Evolutionary diversity in tropical tree communities peaks at intermediate precipitation. Sci. Rep. 10 , 1188 (2020).

Cadotte, M. W., Cardinale, B. J. & Oakley, T. H. Evolutionary history and the effect of biodiversity on plant productivity. Proc. Natl Acad. Sci. USA 105 , 17012–17017 (2008).

Hopkins, M. J. G. Modelling the known and unknown plant biodiversity of the Amazon Basin. J. Biogeogr. 34 , 1400–1411 (2007).

Wilson, E. O. in Biodiversity (eds Wilson E. O. & Peter F. M.) Ch. 1 (National Academies Press, 1988).

Brooks, T. M. et al. Habitat loss and extinction in the hotspots of biodiversity. Conserv. Biol. 16 , 909–923 (2002).

Gibbs, H. K. et al. Brazil’s soy moratorium. Science 347 , 377–378 (2015).

Alix-Garcia, J. & Gibbs, H. K. Forest conservation effects of Brazil’s zero deforestation cattle agreements undermined by leakage. Glob. Environ. Change 47 , 201–217 (2017).

Escobar, H. There’s no doubt that Brazil’s fires are linked to deforestation, scientists say. Science https://doi.org/10.1126/science.aaz2689 (2019).

Amazon fires: Brazil sends army to help tackle blazes. BBC News https://www.bbc.co.uk/news/world-latin-america-49452789 (24 August 2019).

Marengo, J. A., Tomasella, J., Soares, W. R., Alves, L. M. & Nobre, C. A. Extreme climatic events in the Amazon basin. Theor. Appl. Climatol. 107 , 73–85 (2012).

Malhi, Y. et al. Exploring the likelihood and mechanism of a climate-change-induced dieback of the Amazon rainforest. Proc. Natl Acad. Sci. USA 106 , 20610–20615 (2009).

Swann, A. L. S. et al. Continental-scale consequences of tree die-offs in North America: identifying where forest loss matters most. Environ. Res. Lett. 13 , 055014 (2018).

McCoy, T. Amazon fires dropped unexpectedly in September, after summer spike. Washington Post https://www.washingtonpost.com/world/the_americas/amazon-fires-dropped-unexpectedly-in-september-after-spiking-over-the-summer/2019/10/02/4ddc0026-e516-11e9-b403-f738899982d2_story.html (2 October 2019).

Moutinho, P., Guerra, R. & Azevedo-Ramos, C. Achieving zero deforestation in the Brazilian Amazon: what is missing? Elementa 4 , 000125 (2016).

Olson, D. M. et al. Terrestrial ecoregions of the world: a new map of life on Earth: A new global map of terrestrial ecoregions provides an innovative tool for conserving biodiversity. Bioscience 51 , 933–938 (2001).

Hansen, M. C. et al. High-resolution global maps of 21st-century forest cover change. Science 342 , 850–853 (2013).

Giglio, L., Schroeder, W. & Justice, C. O. The collection 6 MODIS active fire detection algorithm and fire products. Remote Sens. Environ. 178 , 31–41 (2016).

Giglio, L. MODIS Collection 6 Active Fire Product User’s Guide Revision A (Univ. Maryland, 2015).

Barlow, J., Lagan, B. O. & Peres, C. A. Morphological correlates of fire-induced tree mortality in a central Amazonian forest. J. Trop. Ecol. 19 , 291–299 (2003).

Brando, P. M. et al. Fire-induced tree mortality in a neotropical forest: the roles of bark traits, tree size, wood density and fire behavior. Glob. Change Biol. 18 , 630–641 (2012).

Gibbs, H. K. et al. Tropical forests were the primary sources of new agricultural land in the 1980s and 1990s. Proc. Natl Acad. Sci. USA 107 , 16732–16737 (2010).

Barlow, J. & Peres, C. in Emerging Threats to Tropical Forests (eds. Laurance, W. F. & Peres, C. A.) 225–240 (Univ. Chicago Press, 2006).

Barlow, J. et al. Wildfires in bamboo-dominated Amazonian forest: impacts on above-ground biomass and biodiversity. PLoS ONE 7 , e33373 (2012).

Gerwing, J. J. Degradation of forests through logging and fire in the eastern Brazilian Amazon. For. Ecol. Manage. 157 , 131–141 (2002).

Brando, P. M. et al. Abrupt increases in Amazonian tree mortality due to drought-fire interactions. Proc. Natl Acad. Sci. USA 111 , 6347–6352 (2014).

Barlow, J. & Peres, C. A. Avifaunal responses to single and recurrent wildfires in Amazonian forests. Ecol. Appl. 14 , 1358–1373 (2004).

Paolucci, L. N., Schoereder, J. H., Brando, P. M. & Andersen, A. N. Fire-induced forest transition to derived savannas: cascading effects on ant communities. Biol. Conserv. 214 , 295–302 (2017).

Roy, D. P. & Kumar, S. S. Multi-year MODIS active fire type classification over the Brazilian Tropical Moist Forest Biome. Int. J. Digital Earth 10 , 54–84 (2017).

Giglio, L., Schroeder, W., Hall, J. V. & Justice, C. O. MODIS Collection 6 Active Fire Product User’s Guide Revision B (Univ. Maryland, 2018).

Barriopedro, D., Fischer, E. M., Luterbacher, J., Trigo, R. M. & García-Herrera, R. The hot summer of 2010: redrawing the temperature record map of Europe. Science 332 , 220–224 (2011).

Chen, Y. et al. Forecasting fire season severity in South America using sea surface temperature anomalies. Science 334 , 787–791 (2011).

Giglio, L. et al. Assessing variability and long-term trends in burned area by merging multiple satellite fire products. Biogeosciences 7 , 1171–1186 (2010).

Justice, C. O. et al. The MODIS fire products. Remote Sens. Environ. 83 , 244–262 (2002).

Giglio, L., Boschetti, L., Roy, D. P., Humber, M. L. & Justice, C. O. The Collection 6 MODIS burned area mapping algorithm and product. Remote Sens. Environ. 217 , 72–85 (2018).

Nóbrega, C. C., Brando, P. M., Silvério, D. V., Maracahipes, L. & de Marco, P. Effects of experimental fires on the phylogenetic and functional diversity of woody species in a neotropical forest. For. Ecol. Manage. 450 , 117497 (2019).

Alencar, A., Nepstad, D. & Diaz, M. C. V. Forest understory fire in the Brazilian Amazon in ENSO and Non-ENSO years: area burned and committed carbon emissions. Earth Interact. 10 , 1–17 (2006).

Siegert, F., Ruecker, G., Hinrichs, A. & Hoffmann, A. A. Increased damage from fires in logged forests during droughts caused by El Niño. Nature 414 , 437–440 (2001).

Cochrane, M. A. & Laurance, W. F. Fire as a large-scale edge effect in Amazonian forests. J. Trop. Ecol. 18 , 311–325 (2002).

Ray, D., Nepstad, D. & Moutinho, P. Micrometeorological and canopy controls of fire susceptibility in a forested Amazon landscape. Ecol. Appl. 15 , 1664–1678 (2005).

Silvério, D. V. et al. Fire, fragmentation, and windstorms: a recipe for tropical forest degradation. J. Ecol. 107 , 656–667 (2019).

Guisan, A. & Zimmermann, N. E. Predictive habitat distribution models in ecology. Ecol. Modell. 135 , 147–186 (2000).

Fegraus, E. Tropical Ecology Assessment and Monitoring Network (TEAM Network). Biodivers. Ecol. 4 , 287–287 (2012).

Peet, R. K., Lee, M. T., Jennings, M. D. & Faber-Langendoen, D. VegBank: a permanent, open-access archive for vegetation plot data. Biodivers. Ecol. 4 , 233–241 (2012).

DeWalt, S. J., Bourdy, G., Chavez de Michel, L. R. & Quenevo, C. Ethnobotany of the Tacana: quantitative inventories of two permanent plots of Northwestern Bolivia. Econ. Bot. 53 , 237–260 (1999).

USDA Forest Service. Forest Inventory and Analysis National Program, http://www.fia.fs.fed.us/ (2013).

Wiser, S. K., Bellingham, P. J. & Burrows, L. E. Managing biodiversity information: development of New Zealand’s National Vegetation Survey databank. N. Z. J. Ecol. 25 , 1–17 (2001).

Anderson-Teixeira, K. J. et al. CTFS-ForestGEO: a worldwide network monitoring forests in an era of global change. Glob. Change Biol. 21 , 528–549 (2015).

Enquist, B. & Boyle, B. SALVIAS – the SALVIAS vegetation inventory database. Biodivers. Ecol. 4 , 288 (2012).

GBIF.org. GBIF Occurrence Download https://doi.org/10.15468/dl.yubndf (2018).

Dauby, G. et al. RAINBIO: a mega-database of tropical African vascular plants distributions. PhytoKeys 74 , 1–18 (2016).

Arellano, G. et al. A standard protocol for woody plant inventories and soil characterisation using temporary 0.1-ha plots in tropical forests. J. Trop. For. Sci. 28 , 508–516 (2016).

O’Connell, B. M. et al. The Forest Inventory and Analysis Database: Database Description and User Guide for Phase 2 (version 6.1), https://doi.org/10.2737/fs-fiadb-p2-6.1 (USDA Forest Service, 2016).

Oliveira-Filho, A. T. NeoTropTree, Flora arbórea da Região Neotropical: Um Banco de Dados Envolvendo Biogeografia, Diversidade e Conservação, http://www.neotroptree.info (Univ. Federal de Minas Gerais, 2017).

Peet, R. K., Lee, M. T., Jennings, M. D. & Faber-Langendoen, D. VegBank: The Vegetation Plot Archive of the Ecological Society of America , http://vegbank.org (accessed 2013).

Boyle, B. et al. The taxonomic name resolution service: an online tool for automated standardization of plant names. BMC Bioinf. 14 , 16 (2013).

Goldsmith, G. R. et al. Plant-O-Matic: a dynamic and mobile guide to all plants of the Americas. Methods Ecol. Evol. 7 , 960–965 (2016).

McFadden, I. R. et al. Temperature shapes opposing latitudinal gradients of plant taxonomic and phylogenetic β diversity. Ecol. Lett. 22 , 1126–1135 (2019).

Enquist, B. J., Condit, R., Peet, R. K., Schildhauer, M. & Thiers, B. M. Cyberinfrastructure for an integrated botanical information network to investigate the ecological impacts of global climate change on plant biodiversity. Preprint at https://doi.org/10.7287/peerj.preprints.2615v2 (2016).

Maitner, B. S. et al. The BIEN R package: A tool to access the Botanical Information and Ecology Network (BIEN) database. Methods Ecol. Evol. 9 , 373–379 (2017).

Phillips, S. J. & Dudik, M. Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. Ecography 31 , 161–175 (2008).

Merow, C. & Silander, J. A. A comparison of Maxlike and Maxent for modelling species distributions. Methods Ecol. Evol. 5 , 215–225 (2014).

Aiello-Lammens, M. E., Boria, R. A., Radosavljevic, A., Vilela, B. & Anderson, R. P. spThin: an R package for spatial thinning of species occurrence records for use in ecological niche models. Ecography 38 , 541–545 (2015).

Grubbs, F. E. Sample criteria for testing outlying observations. Ann. Math. Statist. 21 , 27–58 (1950).

Article   MathSciNet   MATH   Google Scholar  

Komsta, L. outliers: Tests for outliers. R package v.0.14 (2011).

Fick, S. E. & Hijmans, R. J. WorldClim 2: new 1-km spatial resolution climate surfaces for global land areas. Int. J. Climatol. 37 , 4302–4315 (2017).

Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G. & Jarvis, A. Very high resolution interpolated climate surfaces for global land areas. Int. J. Climatol. 25 , 1965–1978 (2005).

Mueller-Dombois, D. & Ellenberg, H. Aims and Methods of Vegetation Ecology (Wiley, 1974).

Friedman, J., Hastie, T. & Tibshirani, R. glmnet: Lasso and elastic-net regularized generalized linear models. R package v.4.0-2 (2020).

Phillips, S. J., Anderson, R. P. & Schapire, R. E. Maximum entropy modeling of species geographic distributions. Ecol. Modell. 190 , 231–259 (2006).

Drake, J. M. Range bagging: a new method for ecological niche modelling from presence-only data. J. R. Soc. Interface 12 , 20150086 (2015).

Article   PubMed   PubMed Central   Google Scholar  

Cardoso, D. et al. Amazon plant diversity revealed by a taxonomically verified species list. Proc. Natl Acad. Sci. USA 114 , 10695–10700 (2017).

Article   CAS   PubMed   PubMed Central   Google Scholar  

Warton, D. I. & Shepherd, L. C. Poisson point process models solve the “pseudo-absence problem” for presence-only data in ecology. Ann. Appl. Stat. 4 , 1383–1402 (2010).

MathSciNet   MATH   Google Scholar  

Renner, I. W. et al. Point process models for presence-only analysis. Methods Ecol. Evol. 6 , 366–379 (2015).

Dinerstein, E. et al. An ecoregion-based approach to protecting half the terrestrial realm. Bioscience 67 , 534–545 (2017).

Roberts, D. R. et al. Cross-validation strategies for data with temporal, spatial, hierarchical, or phylogenetic structure. Ecography 40 , 913–929 (2016).

Phillips, S. J. Transferability, sample selection bias and background data in presence-only modelling: a response to Peterson et al . (2007). Ecography 31 , 272–278 (2008).

Merow, C., Smith, M. J. & Silander, J. A. Jr. A practical guide to MaxEnt for modeling species’ distributions: what it does, and why inputs and settings matter. Ecography 36 , 1058–1069 (2013).

Qiao, H. et al. An evaluation of transferability of ecological niche models. Ecography 42 , 521–534 (2019).

Peterson, A. T., Papeş, M. & Soberón, J. Rethinking receiver operating characteristic analysis applications in ecological niche modeling. Ecol. Modell. 213 , 63–72 (2008).

Allouche, O., Tsoar, A. & Kadmon, R. Assessing the accuracy of species distribution models: prevalence, kappa and the true skill statistic (TSS). J. Appl. Ecol. 43 , 1223–1232 (2006).

Jung, M. et al. Areas of global importance for terrestrial biodiversity, carbon, and water. Preprint at https://doi.org/10.1101/2020.04.16.021444 (2020).

Carlson, C. J. et al. Climate change will drive novel cross-species viral transmission. Preprint at https://doi.org/10.1101/2020.01.24.918755 (2020).

BirdLife International. IUCN Red List for Birds http://www.birdlife.org (2019).

Brooks, T. M. et al. Measuring terrestrial area of habitat (AOH) and its utility for the IUCN Red List. Trends Ecol. Evol. 34 , 977–986 (2019).

de Area Leão Pereira, E. J., de Santana Ribeiro, L. C., da Silva Freitas, L. F. & de Barros Pereira, H. B. Brazilian policy and agribusiness damage the Amazon rainforest. Land Use Policy 92 , 104491 (2020).

Garcia, R. T. After Brazil’s summer of fire, the militarization of the Amazon remains. Foreign Policy https://foreignpolicy.com/2019/11/19/militarization-amazon-legacy-brazil-forest-fire-bolsonaro/ (19 November 2019).

Vicente-Serrano, S. M., Beguería, S. & López-Moreno, J. I. A multiscalar drought index sensitive to global warming: The Standardized Precipitation Evapotranspiration Index. J. Clim. 23 , 1696–1718 (2010).

Feldpausch, T. R. et al. Amazon forest response to repeated droughts. Global Biogeochem. Cycles 30 , 964–982 (2016).

Marin, P.-G., Julio, C. J., Arturo, R.-T. D. & Jose, V.-N. D. Drought and spatiotemporal variability of forest fires across Mexico. Chin. Geogr. Sci. 28 , 25–37 (2018).

Adams, H. D. et al. Temperature sensitivity of drought-induced tree mortality portends increased regional die-off under global-change-type drought. Proc. Natl Acad. Sci. USA 106 , 7063–7066 (2009).

Download references

Acknowledgements

We acknowledge the herbaria that contributed data to this work: HA, FCO, MFU, UNEX, VDB, ASDM, BPI, BRI, CLF, L, LPB, AD, TAES, FEN, FHO, A, ANSM, BCMEX, RB, TRH, AAH, ACOR, AJOU, UI, AK, ALCB, AKPM, EA, AAU, ALU, AMES, AMNH, AMO, ANA, GH, ARAN, ARM, AS, CICY, ASU, BAI, AUT, B, BA, BAA, BAB, BACP, BAF, BAL, COCA, BARC, BBS, BC, BCN, BCRU, BEREA, BG, BH, BIO, BISH, SEV, BLA, BM, MJG, BOL, CVRD, BOLV, BONN, BOUM, BR, BREM, BRLU, BSB, BUT, C, CAMU, CAN, CANB, CAS, CAY, CBG, CBM, CEN, CEPEC, CESJ, CHR, ENCB, CHRB, CIIDIR, CIMI, CLEMS, COA, COAH, COFC, CP, COL, COLO, CONC, CORD, CPAP, CPUN, CR, CRAI, FURB, CU, CRP, CS, CSU, CTES, CTESN, CUZ, DAO, HB, DAV, DLF, DNA, DS, DUKE, DUSS, E, HUA, EAC, ECU, EIF, EIU, GI, GLM, GMNHJ, K, GOET, GUA, EKY, EMMA, HUAZ, ERA, ESA, F, FAA, FAU, UVIC, FI, GZU, H, FLAS, FLOR, HCIB, FR, FTG, FUEL, G, GB, GDA, HPL, GENT, GEO, HUAA, HUJ, CGE, HAL, HAM, IAC, HAMAB, HAS, HAST, IB, HASU, HBG, IBUG, HBR, IEB, HGI, HIP, IBGE, ICEL, ICN, ILL, SF, NWOSU, HO, HRCB, HRP, HSS, HU, HUAL, HUEFS, HUEM, HUSA, HUT, IAA, HYO, IAN, ILLS, IPRN, FCQ, ABH, BAFC, BBB, INPA, IPA, BO, NAS, INB, INEGI, INM, MW, EAN, IZTA, ISKW, ISC, GAT, IBSC, UCSB, ISU, IZAC, JBAG, JE, SD, JUA, JYV, KIEL, ECON, TOYA, MPN, USF, TALL, RELC, CATA, AQP, KMN, KMNH, KOR, KPM, KSTC, LAGU, UESC, GRA, IBK, KTU, KU, PSU, KYO, LA, LOMA, SUU, UNITEC, NAC, IEA, LAE, LAF, GMDRC, LCR, LD, LE, LEB, LI, LIL, LINN, AV, HUCP, MBML, FAUC, CNH, MACF, CATIE, LTB, LISI, LISU, MEXU, LL, LOJA, LP, LPAG, MGC, LPD, LPS, IRVC, MICH, JOTR, LSU, LBG, WOLL, LTR, MNHN, CDBI, LYJB, LISC, MOL, DBG, AWH, NH, HSC, LMS, MELU, NZFRI, M, MA, UU, UBT, CSUSB, MAF, MAK, MB, KUN, MARY, MASS, MBK, MBM, UCSC, UCS, JBGP, OBI, BESA, LSUM, FULD, MCNS, ICESI, MEL, MEN, TUB, MERL, CGMS, FSU, MG, HIB, TRT, BABY, ETH, YAMA, SCFS, SACT, ER, JCT, JROH, SBBG, SAV, PDD, MIN, SJSU, MISS, PAMP, MNHM, SDSU, BOTU, MPU, MSB, MSC, CANU, SFV, RSA, CNS, JEPS, BKF, MSUN, CIB, VIT, MU, MUB, MVFA, SLPM, MVFQ, PGM, MVJB, MVM, MY, PASA, N, HGM, TAM, BOON, MHA, MARS, COI, CMM, NA, NCSC, ND, NU, NE, NHM, NHMC, NHT, UFMA, NLH, UFRJ, UFRN, UFS, ULS, UNL, US, NMNL, USP, NMR, NMSU, XAL, NSW, ZMT, BRIT, MO, NCU, NY, TEX, U, UNCC, NUM, O, OCLA, CHSC, LINC, CHAS, ODU, OKL, OKLA, CDA, OS, OSA, OSC, OSH, OULU, OXF, P, PACA, PAR, UPS, PE, PEL, SGO, PEUFR, PH, PKDC, SI, PMA, POM, PORT, PR, PRC, TRA, PRE, PY, QMEX, QCA, TROM, QCNE, QRS, UH, R, REG, RFA, RIOC, RM, RNG, RYU, S, SALA, SANT, SAPS, SASK, SBT, SEL, SING, SIU, SJRP, SMDB, SNM, SOM, SP, SRFA, SPF, STL, STU, SUVA, SVG, SZU, TAI, TAIF, TAMU, TAN, TEF, TENN, TEPB, TI, TKPM, TNS, TO, TU, TULS, UADY, UAM, UAS, UB, UC, UCR, UEC, UFG, UFMT, UFP, UGDA, UJAT, ULM, UME, UMO, UNA, UNM, UNR, UNSL, UPCB, UPNA, USAS, USJ, USM, USNC, USZ, UT, UTC, UTEP, UV, VAL, VEN, VMSL, VT, W, WAG, WII, WELT, WIS, WMNH, WS, WTU, WU, Z, ZSS, ZT, CUVC, AAS, AFS, BHCB, CHAM, FM, PERTH and SAN. X.F., D.S.P., E.A.N., A.L. and J.R.B. were supported by the University of Arizona Bridging Biodiversity and Conservation Science program. Z.L. was supported by NSFC (41922006) and K. C. Wong Education Foundation. The BIEN working group was supported by the National Center for Ecological Analysis and Synthesis, a centre funded by NSF EF-0553768 at the University of California, Santa Barbara, and the State of California. Additional support for the BIEN working group was provided by iPlant/Cyverse via NSF DBI-0735191. B.J.E., B.M. and C.M. were supported by NSF ABI-1565118. B.J.E. and C.M. were supported by NSF ABI-1565118 and NSF HDR-1934790. B.J.E., L.H. and P.R.R. were supported by the Global Environment Facility SPARC project grant (GEF-5810). D.D.B. was supported in part by NSF DEB-1824796 and NSF DEB-1550686. S.R.S. was supported by NSF DEB-1754803. X.F. and A.L. were partly supported by NSF DEB-1824796. B.J.E. and D.M.N. were supported by NSF DEB-1556651. M.M.P. is supported by the São Paulo Research Foundation (FAPESP), grant 2019/25478-7. D.M.N. was supported by Instituto Serrapilheira/Brazil (Serra-1912-32082). E.I.N. was supported by NSF HDR-1934712. We thank L. López-Hoffman and L. Baldwin for constructive comments.

Author information

These authors contributed equally: Xiao Feng, Cory Merow, Zhihua Liu, Daniel S. Park, Patrick R. Roehrdanz, Brian Maitner, Erica A. Newman, Brian J. Enquist

Authors and Affiliations

Department of Geography, Florida State University, Tallahassee, FL, USA

Eversource Energy Center and Department of Ecology and Evolutionary Biology, University of Connecticut, Storrs, CT, USA

Cory Merow & Brian Maitner

CAS Key Laboratory of Forest Ecology and Management, Institute of Applied Ecology, Chinese Academy of Sciences, Shenyang, China

Department of Biological Sciences, Purdue University, West Lafayette, IN, USA

Daniel S. Park

Purdue Center for Plant Biology, Purdue University, West Lafayette, IN, USA

The Moore Center for Science, Conservation International, Arlington, VA, USA

Patrick R. Roehrdanz & Lee Hannah

Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, AZ, USA

Erica A. Newman, Brad L. Boyle, Joseph R. Burger, Scott R. Saleska & Brian J. Enquist

Arizona Institutes for Resilience, University of Arizona, Tucson, AZ, USA

Erica A. Newman, Aaron Lien & Joseph R. Burger

Hardner & Gullison Associates, Amherst, NH, USA

Brad L. Boyle

School of Natural Resources and the Environment, University of Arizona, Tucson, AZ, USA

Aaron Lien, David D. Breshears & José R. Soto

Department of Biology, University of Kentucky, Lexington, KY, USA

Joseph R. Burger

Departamento de Biologia Animal, Universidade Estadual de Campinas, Campinas, Brazil

Mathias M. Pires

Department of Earth System Science, University of California, Irvine, Irvine, CA, USA

Paulo M. Brando

Woodwell Climate Research Center, Falmouth, MA, USA

Instituto de Pesquisa Ambiental da Amazônia (IPAM), Brasilia, Brazil

Insitute for Global Ecology, Florida Institute of Technology, Melbourne, FL, USA

Mark B. Bush

Department of Ecosystem and Landscape Dynamics, Institute for Biodiversity and Ecosystem Dynamics, University of Amsterdam, Amsterdam, The Netherlands

Crystal N. H. McMichael

Institute of Biological Sciences, Federal University of Minas Gerais, Belo Horizonte, Brazil

Danilo M. Neves

Department of Mechanical and Civil Engineering, Florida Institute of Technology, Melbourne, FL, USA

Efthymios I. Nikolopoulos

School of Geography, Development and Environment, University of Arizona, Tucson, AZ, USA

Tom P. Evans

Department of Epidemiology and Biostatistics, College of Public Health, University of Arizona, Tucson, AZ, USA

Kacey C. Ernst

The Santa Fe Institute, Santa Fe, NM, USA

Brian J. Enquist

You can also search for this author in PubMed   Google Scholar

Contributions

X.F. conceived the idea, which was refined by discussion with D.S.P., C.M., B.M., P.R.R., E.A.N., B.L.B., A.L., J.R.B., D.D.B., J.R.S., K.C.E. and B.J.E.; X.F. and Z.L. processed the remote-sensing data; C.M., X.F., B.M., B.L.B., D.S.P. and B.J.E. conducted the analyses of plant data; P.R.R., C.M., B.M., X.F. and D.S.P. conducted the analyses of vertebrate data; X.F., C.M., S.R.S. and E.A.N. processed the drought data; D.S.P., X.F., C.M., P.R.R. and B.M. designed the illustrations with help from B.J.E., D.D.B., K.C.E. and E.A.N.; E.A.N., X.F., and D.S.P. conducted the statistical analyses with help from B.J.E.; X.F., B.J.E., B.M., A.L., J.R.B., D.S.P., C.M., E.A.N., Z.L. and P.R.R. wrote the original draft; all authors contributed to interpreting the results and the editing of manuscript drafts. B.J.E., C.M., K.C.E. and D.D.B. led to the acquisition of the financial support for the project. X.F., C.M., B.M., D.S.P., P.R.R., Z.L., E.A.N. and B.J.E. contributed equally to data, analyses and writing.

Corresponding author

Correspondence to Xiao Feng .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Additional information

Peer review information Nature thanks the anonymous reviewers for their contribution to the peer review of this work.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data figures and tables

Extended data fig. 1 fire-impacted forest and forest loss in the amazon basin..

a – h , Visualization of fire-impacted forest ( a , b ), forest loss without fire ( c , d ), fire-impacted forest with forest loss ( e , f ), and fire-impacted forest without forest loss ( g , h ) in the Amazon Basin based on MODIS burned area (left panels) and active fire (right panels). Data in a – d are resampled from the 500m (MODIS burned area) or 1 km (MODIS active fire) to 10 km resolution using mean function and thresholded at 0.01 to illustrate the temporal dynamics. Black represents non-forested areas masked out from this study. The cumulative fire-impacted forest is classified into two categories: fire-impacted forest with forest loss ( e , f ) and fire-impacted forest without forest loss ( g , h ). Data in e – h are resampled to 10 km using mean function to illustrate the cumulative percentages of impacts.

Extended Data Fig. 2 Scatter plot of species’ range impacted by fire.

Scatter plot of species’ range size in Amazon forest (x-axis) and percentage of total range impacted by fire (red) and forest loss without fire (black) up to 2019 for plants (left panel) and vertebrates (right panel).

Extended Data Fig. 3 Density plot of species’ cumulative range impacted by fire.

Density plot of species’ cumulative range impacted by fire. The different colours represent years 2001-2019. The x-axis is log10 transformed.

Extended Data Fig. 4 Summary of forest impacts in the Amazon Basin.

Areas of forest impact in the Amazon Basin estimated from MODIS burned area (top) and MODIS active fire (bottom).

Extended Data Fig. 5 Cumulative impacts on biodiversity in the Amazon Basin.

Cumulative effects of forest loss without fire on biodiversity in the Amazon rainforest. In the left panels, the black and grey shading represent the cumulative forest loss without fire based on MODIS burned area and MODIS active fire, respectively. Coloured areas represent the lower and upper bounds of cumulative numbers of a , plant and c , vertebrate species’ ranges impacted. Right panels depict the relationships between the cumulative forest loss without fire (based on MODIS burned area) and cumulative number of b , plant and d , vertebrate species. Coloured lines represent predicted values of an ordinary least squares linear regression and grey bands define the two-sided 95% confidence interval (two-sided, p values = 0.00). The silhouette of the tree is from http://phylopic.org/ ; silhouette of the monkey is courtesy of Mathias M. Pires.

Extended Data Fig. 6 Fire-impacted forest in Brazil.

Newly fire-impacted forest in Brazil (based on MODIS active fire). a shows the area of fire-impacted forest not explained by drought conditions. Different colours represent years from different policy regimes: pre-regulations in light red (mean value in dark red), regulation in grey (mean value in black dashed line), and 2019 in blue. The y-axis represents the difference between actual area and area predicted by drought conditions calibrated by data from regulation years ( Methods ). A positive value on the y-axis represents more area than expected, using the regulation years as a baseline. b shows a scatter plot of newly fire-impacted forest in Brazil and drought conditions (SPEI); The lines represent the ordinary least squares linear regression between fire-impacted forest and drought conditions for pre-regulation (red) and regulation (black) respectively.

Extended Data Fig. 7 Fire-impacted forest in different countries.

The contribution (0–1) of different countries to the newly fire-impacted forest each year based on MODIS active fire (top) and MODIS burned area (bottom).

Extended Data Figure 8 Impacts of fire on forest and biodiversity in Brazil.

a , Newly fire-impacted forest, b , new range impact on plants and c , new range impacts on vertebrate species in Brazil each year (based on MODIS active fire) that are not predicted by drought conditions. The colours represent three policy regimes: pre-regulation in red, regulation in grey and 2019 in blue. The y-axis represents the difference between actual value (area or range impacted by fire) and the values predicted by drought conditions calibrated by data from regulation years ( Methods ). A positive value on the y-axis represents more area or range impacted by fire than the expectation using the regulation years as a baseline. The dotted lines represent a smooth curve fitted to the values based on the loess method.

Supplementary information

‘supplementary discussion’ and ‘this file contains the supplementary discussion’, reporting summary, rights and permissions.

Reprints and permissions

About this article

Cite this article.

Feng, X., Merow, C., Liu, Z. et al. How deregulation, drought and increasing fire impact Amazonian biodiversity. Nature 597 , 516–521 (2021). https://doi.org/10.1038/s41586-021-03876-7

Download citation

Received : 22 November 2019

Accepted : 04 August 2021

Published : 01 September 2021

Issue Date : 23 September 2021

DOI : https://doi.org/10.1038/s41586-021-03876-7

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

This article is cited by

Variation of floristic diversity, community composition, endemism, and conservation status of tree species in tropical rainforests of sri lanka across a wide altitudinal gradient.

• Nimalka Sanjeewani
• Dilum Samarasinghe
• Janendra De Costa

Scientific Reports (2024)

Comparative analysis of seed biomass from Amazonian fruits for activated carbon production

• Edina Ruth Mendes Leal Mafra
• Thiago de Paula Protásio
• Lina Bufalino

Biomass Conversion and Biorefinery (2024)

Assessing changes in global fire regimes

• Sayedeh Sara Sayedi
• Benjamin W. Abbott
• Anne-Laure Daniau

Fire Ecology (2024)

The role of high-biodiversity regions in preserving Nature’s Contributions to People

• Marta Cimatti
• Rebecca Chaplin-Kramer
• Moreno Di Marco

Nature Sustainability (2023)

Forest conservation in Indigenous territories and protected areas in the Brazilian Amazon

• Yuanwei Qin
• Xiangming Xiao
• Philip Martin Fearnside

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

The Negative Effect of Temperature Variability on Household Wealth in Low- and Middle-Income Countries

• Open access
• Published: 28 August 2024

Cite this article

You have full access to this open access article

• Ida Brzezinska 1 &
• Paul Jasper 1  

Climate change distributes its negative impacts on societies unequally. Low- and middle-income countries (LMICs) are disproportionately affected by climate change, which threatens to undo global progress in poverty reduction. Temperature variability has been shown to reduce macro-economic growth and to negatively affect household wealth. Climate models predict that LMICs are located in areas that will experience the largest increases in temperature variability. Understanding the effects of an increasingly variable climate on social and economic outcomes is thus of key importance. Yet current literature on the link between temperature variability and poverty has limited geographical coverage, with evidence from many LMICs still missing. In addition, the data used in studies have tended to be at a low level of spatial disaggregation. This paper combines remote sensing temperature data from the Reanalysis Fifth Generation (ERA5) global climate dataset for 2018 and micro-estimates of relative household wealth at a 2.4 km spatial resolution predicted by machine learning (ML) algorithms from (Chi et al., 2022 ) in a Spatial First Differences (SFD) research design. We find a remarkably robust negative effect of increases in day-to-day temperature variability on household wealth across 89 LMICs. Our heterogeneity analysis shows that the magnitude of negative effects of day-to-day temperature variability on household wealth is largest in Sub-Saharan Africa, in low-income countries, and in tropical climates. Altogether our findings highlight the need to build climate resilience across LMICs, for instance through climate-responsive social protection programmes – particularly in the most affected geographies and climates .

Similar content being viewed by others

Climate shocks and wealth inequality in the UK: evidence from monthly data

The Effects of Climate Change on Poverty and Income Distribution: A Case Study for Rural Mexico

How many people will live in poverty because of climate change? A macro-level projection analysis to 2070

Avoid common mistakes on your manuscript.

Introduction

Climate change continues to increase the frequency and intensity of extreme weather events, exerting widespread negative impacts on societies around the world – particularly on people living in poverty (Jafino et al. 2020 ). Recent climate models show that low- and middle-income countries (LMICs) are located in ‘hot-spot’ areas, where increases in temperature variability due to climate change are projected to be the greatest (Bathiany et al. 2018 ). According to projections of relative changes in temperature variability until the end of the twenty-first century in Bathiany et al. ( 2018 ), climate variability is expected to decrease in most high-income countries – while LMICs will experience the largest increases in variability. This is one of the exemplifiers of the issue of climate injustice, with countries that contributed the least to climate change likely to be the ones most affected by its negative effects.

Given the projected changes in climate variability, a crucial question relates to the impact of increased temperature variability on economic and social outcomes. Recent evidence shows that variations in temperature can negatively affect macro-economic growth. Kotz et al. ( 2021 ) analyse subnational economic data for 1,537 regions worldwide over a 40 year period, finding that, on average, an extra degree of day-to-day temperature variability reduces the regional economic growth rate by 5 percentage points. Linsenmeier ( 2021 ) uses satellite data on nightlights as a proxy for economic activity and similarly finds that day-to-day variability has negative effects on economic growth at most temperature levels. Consistent with the unequal impacts of climate change mentioned above, Dell et al. ( 2008 ) show that temperature fluctuations reduce economic growth in low-income countries but have little effect in rich countries.

Alongside those looking at macro-economic outcomes, many studies have analysed the impact of temperature variability on household poverty. Agriculture is a common focus area. Temperature variability has been associated with lower agricultural productivity, decreased yields and thus reduced farm income in Togo (Yevessé, 2021 ), Nigeria (Ishaya et al. 2014 ), and El Salvador (Ibanez et al. 2021 ). Also, temperature variability and temperature shocks have been directly linked to reductions of food and non-food consumption in Vietnam (Narloch & Bangalore 2016 ), Mexico (Skoufias et al. 2011 ), and Malawi (Asfaw & Maggio 2018 ). Other studies find that temperature variability can negatively affect educational outcomes, including the likelihood of a child completing any education in Ethiopia (Randell & Gray 2016 ). Finally, adverse effects of temperature variability on poverty can materialise through lower labour productivity (Ibanez et al. 2021 ; Markham & Markham 2005 ; Cachon et al. 2012 ).

Although there is a sizeable literature on the link between temperature variability and poverty, geographical coverage remains limited, with evidence from most LMICs still lacking. In addition, studies to date have typically considered poverty outcomes at a coarse level of spatial disaggregation, obscuring the spatial distribution of poverty at smaller areas. We leverage a novel high-resolution dataset on household wealth and poverty from Chi et al. ( 2022 ). This consists of micro-estimates of relative household asset wealth for 89 LMICs at an unprecedented 2.4 km spatial resolution predicted from machine-learning (ML) algorithms on the basis of Demographic and Health Surveys (DHS), connectivity data from Facebook, demographic maps, and geo-spatial datasets. We combine this with remote sensing temperature data from 2018 at a 0.5 degrees spatial resolution taken from the Reanalysis Fifth Generation (ERA5) global climate time-series data. We apply a Spatial First Differences (SFD) approach that addresses unobserved heterogeneity in cross-sectional data that is dense in space (Druckenmiller & Hsiang 2018 ), and a variety of robustness checks, to identify the causal impact of day-to-day temperature variability on household wealth in LMICs.

Through the use of high-resolution ‘big data’ on climate and household wealth, as well as novel analytical approaches, we aim to address the following research question:

Is there a causal relationship between day-to-day temperature variability and wealth in LMICs and does it vary across regions?

Existing Literature

The literature that links temperature variability to poverty and wealth in LMICs is broad. Although many studies have looked at the effects of absolute changes in temperature and temperature extremes on wealth and poverty, we focus on the literature that specifically investigates temperature variability . We first very briefly touch on the literature that investigates factors that explain variation and changes in temperature variability, then review theoretical approaches that seek to explain the link between temperature variability and poverty or wealth in LMICs, before turning to a summary of existing empirical evidence related to this issue.

Factors Affecting Changes in Temperature Variability

Like many climatic variables, temperature variability is affected by local, regional, and global factors whose impacts materialise at different time-scales. For instance, studies that have identified global factors affecting changes in temperature variability have highlighted the effects of climate change and how these materialise differently across regions (Bathiany et al. 2018 ). More locally, studies have identified urbanisation (Tam et al. 2015 ), deforestation (Ge et al. 2022 ), and changes in soil moisture and sea-ice loss (Bathiany et al. 2018 ) as factors associated with a change in temperature variability. Hence, while changes in temperature variability can generally be explained by global drivers, there is some evidence that at a local level they might also be affected by human activity, such as for example urban build up.

Theory: Temperature Variability Affecting Poverty and Wealth in LMICs

Three main theoretical mechanisms via which temperature variability might affect poverty have been discussed in the literature. First, temperature variability has been linked to the concept of poverty traps (Banerjee & Duflo 2011 ). Given that households experiencing extreme poverty are more likely to live in rural areas and to be engaged in agriculture than other households (The World Bank 2022 ), temperature variability increases uncertainty and volatility of farm income for many poor households. This, in turn, makes it more difficult for households to break out of poverty traps, and increases the likelihood of families being pushed into them. Second, and in a similar vein, theoretical modelling has linked climate change, temperature increases, and increased temperature variability to a reduction of farm productivity and rents, hence reducing a key source of income for most of the world’s poor (Mendelsohn et al. 1994 ). Third, increases in temperature variability and the likelihood of temperature shocks have been theoretically linked to increases in firm production costs, in turn reducing firm productivity and, via this channel, wealth (Melitz 2003 ; Foltz et al. 2018 ).

Existing Empirical Evidence on the Effects of Temperature Variability

There is a growing body of evidence that temperature variability negatively affects aggregate levels of economic growth. For instance, Kotz et al. ( 2021 ), find that an extra degree of day-to-day variability results in a 5-percentage point reduction in growth rates in the regions included in their study. Other studies find similar negative relationships between temperature variability and growth, although effects vary by location, with LMICs, often in hotter climates, being more severely affected (Linsenmeier 2021 ; Dell et al. 2008 ). Other studies have found evidence for negative effects of temperature variability on health-related outcomes, such as mortality rates and birth outcomes (Andalón et al. 2014 ; Hovdahl 2020 ).

At the household level, studies have directly linked increases in temperature variability and temperature shocks with reductions of food and non-food consumption in Vietnam (Narloch & Bangalore 2016 ), Mexico (Skoufias et al. 2011 ), and Malawi (Asfaw & Maggio 2018 ). Others have focused explicitly on rural households and linked temperature to decreases in agricultural income in Ghana (Nkegbe & Kuunibe 2014 ), and rural consumption growth in Tanzania (Letta et al. 2018 ). Haile et al. ( 2018 ) find that negative effects of heat waves on rural households vary by climatic condition, when comparing results from Ghana, Uganda, and Tanzania. This highlights the need for broad analyses that cover a variety of different contextual conditions while also allowing for localised disaggregation of effect estimates.

Looking at other welfare-related outcomes, studies that use household-level data have found evidence that temperature fluctuations are associated with lower education completion rates (Randell & Gray 2016 ), lower levels of income from crops (Yevessé, 2021 ), and lower cereal yields (Ishaya et al. 2014 ). In addition, temperature variability has been found to lead to lower labour productivity in non-agricultural sectors, both in LMICs and in rich economies, particularly in precarious employment, such as at industrial plants or construction sites, where exposure to health risks is higher (Ibanez et al. 2021 ; Markham & Markham 2005 ; Cachon et al. 2012 ).

Gaps and our Contribution

There is hence a significant body of evidence for negative effects of temperature variability and shocks on aggregate economic indicators and household-level outcomes, including consumption, expenditure, and income. The size of these effects seems to vary, depending on the agroecological and climatological characteristics of the area a particular study covers.

As far as we know, no large-scale study has been conducted so far that aggregates data across a global set of LMICs, while at the same time focussing on high resolution, i.e. small area measurements of household wealth. Our study aims to close this gap, providing both a globe-circumventing analysis of this relationship between temperature variability and household wealth in LMICs, and a high-resolution perspective on this. This allows for disaggregating effect estimates by a variety of different contextual variables.

Data and Methods

Climatic data, data sources.

The main data source for climatic variables is the Reanalysis Fifth Generation (ERA5) global climate time-series data for the years 1979 to 2019 taken from the European Centre for Medium-Range Weather Forecasts (ECMWF). ERA5 combines satellite data with historical observations from land stations, aircrafts and radars into global estimates using advanced modelling and data assimilation systems (Hersbach et al. 2020 ). For our main estimation, we use only the climatic data for 2018 to match the time period of our household wealth data (see below). Variables are recorded on a regular grid of 0.5 by 0.5 degrees spatial resolution with global geographical coverage. For our purposes, a crucial advantage of this dataset is the highly disaggregated hourly temporal resolution. This allows us to construct our measure of day-to-day temperature variability from the temperature variable, defined as the temperature of air at two metres above the surface of land, sea, or inland waters, in Kelvins (K). We also process the rainfall flux variable, defined as the rate of rain that falls to the Earth’s surface. For additional analyses that incorporates historical and long-term measures of temperature variability, we make use of pre-2018 observations of this time-series.

To remove the influence of the seasonal temperature cycle, we use an additional climatic dataset to construct a measure of seasonal temperature difference, which serves as a control variable in our main analytical specification. This data comes from the Climatic Research Unit (CRU) Version 4 gridded time-series climatic dataset on a 0.5 latitude by 0.5 longitude regular grid over all land domains of the world except Antarctica. The spatial resolution of this climatic time-series matches that of our main data source – ERA5. Observations are derived by the interpolation of monthly climate anomalies from extensive networks of weather station measurements (Harris et al. 2020 ). The available time range spans 1901–2018. We concentrate on monthly values of temperature, defined as mean two meters above-surface, in the historical reference period of 1979 to 2018. All values are defined in degrees Celsius. Additionally, we rely on a global 0.5 degree resolution dataset with the Koppen-Geiger climate classification in order to identify climatic zones into which grid-cell observations of climatic variables fall (Beck et al. 2018 ).

Constructing Temperature Variability Indicators

Following Kotz et al. ( 2021 ), we construct our main measure of day-to-day temperature variability from the ERA5 dataset as the deviation of daily temperatures from their monthly means, averaged for 2018. For each grid cell, we implement the following steps: 1) aggregate hourly measurements to a mean daily temperature, 2) calculate monthly standard deviations, and 3) average the values of monthly standard deviations for the whole year. This approach is a better measure of day-to-day variability than absolute changes between days, which can be sensitive to changes in observational procedures and not fully reflect seasonal variation (Moberg et al. 2000 ). The result is presented in Fig.  1 . Day-to-day temperature variability displays a strong latitudinal dependence, being the smallest around the equator and increasing at higher latitudes.

Day-to-day temperature variability (in Kelvins)

For some analyses, we also include long-term measures of day-to-day temperature variability, which are averages of the day-to-day temperature variability measure described above for two historical periods: 1) 1989–2001; and 2) 1989–2017. For our main dependent variable, we rely on estimates of household wealth produced by machine learning (ML) algorithms trained on the latest available “ground truth” DHS data and various Big Data sources. While the DHS data used is mostly from 2018, in some cases this data is not available for that year and an earlier DHS is used. The earliest DHS used to train the algorithm is from 2002–03. Footnote 1 For analyses using historical temperature variability data, and to check whether results are sensitive to a change in how we look at historical temperature variability, we therefore choose two historical periods to construct long-term measures of temperature variability: First, 1989–2001, which excludes any time-period for which DHS data is included in the construction of our main dependent variable. Second, 1989–2017, which covers a longer time span, but which overlaps to a limited extent with DHS data collection years.

Day-to-day variations in temperature could also be driven by the variation introduced by the seasonal temperature cycle. We include a measure of seasonal temperature difference as a control variable in our main specification to filter out the confounding effect of seasonal variation. This is calculated using the CRU Version 4 dataset at the grid-cell level as the difference between the maximum and minimum average monthly temperature in a year. In order to capture the historical trend and minimise the influence of anomalies in a particular year, we calculate this difference for every year during the period 1979 to 2018 and average the values to yield one unique observation per grid cell. We convert temperature units from Celsius to Kelvin for consistency with our main dataset, ERA5. The result is plotted in Fig.  2 . Similar to day-to-day temperature variability, seasonal variation is generally larger at higher latitudes. Note that this dataset does not cover land masses in Antarctica.

Seasonal temperature difference (in Kelvins)

Apart from temperature variability, previous research has found that rainfall has a significant impact on poverty in LMICs, particularly through droughts. For example, Brunckhorst ( 2020 ) finds that the 2015 drought in Ethiopia reduced annual consumption of affected households by 12 percent, while Hirvonen et al. ( 2020 ) show that chronic undernutrition rates increased in drought-exposed areas with limited road networks. Drought was also found to have negatively affected household consumption in Somalia (Pape & Wollburg 2019 ) and Malawi (Baquie & Fuje 2020 ) and reduced agricultural productivity in Zambia (Chonabayashi et al. 2020 ). Given the importance of rainfall for poverty outcomes, and its potential relationship with other climatic variables such as temperature variability, we include daily rainfall as an additional control in our key regressions below. This variable is constructed on the basis of the hourly rainfall flux observations from the ERA5 dataset for every grid cell. These values, originally expressed in kg m-2 s-1, are converted into daily rainfall in millimetres for ease of interpretation.

Household Wealth

Data source.

As a measure of relative wealth, we rely on micro-estimates of relative wealth for LMICs at a 2.4 km spatial resolution provided by Chi et al. ( 2022 ). These estimates are produced by applying machine learning algorithms to the ‘ground truth’ of household survey data coming from the USAID Demographic and Health Surveys (DHS), data from satellites, mobile phone networks, and topographic maps, as well as deidentified connectivity data from Facebook. The DHS, which is available for 56 LMICs, is geo-referenced, linked with alternative data sources, and fed into the algorithm, which then predicts relative wealth for all 135 LMICs at a 2.4 km grid-cell spatial resolution. The resulting output is cross-sectional in nature and relies on the latest available DHS in any given country, with the latest information coming from 2018. Footnote 2 Importantly, 5 out of the 10 most predicative features in the model used to produce relative wealth predictions are related to Facebook connectivity data, rather than any geo-spatial covariates that could directly capture our measure of day-to-day temperature variability. Footnote 3

A crucial question relates to the accuracy of these micro-estimates of relative wealth. Chi et al. ( 2022 ) conduct validation exercises comparing their household wealth estimates to actual values from four independent data sources that are not used to train the ML algorithm: 1) census data collected from 15 LMICs since 2001; 2) nationally representative samples collected by the Government of Togo from 6,172 households in 2018/2019; 3) nationally representative samples collected by the Government of Nigeria from 22,104 households in 2019; and 4) data from 5,703 Kenyan households in 2018 collected by GiveDirectly, a non-profit humanitarian aid organisation. For the census data, the prediction model explains between 72–86% of the variation in household wealth. Footnote 4 For Togo and Nigeria, the model explains 76–84% and 50–71% Footnote 5 of variation, respectively. Finally, in Kenya, the coefficient of correlation between the wealth predicted by the ML model and actual values in the survey is between 0.41 and 0.78. In addition to validity tests with external data sources, the data set includes an estimated average model error for each 2.4 km grid cell, which we refer to in more detail when accounting for prediction error.

Variables Used

The measure of wealth produced by micro-estimates in Chi et al. ( 2022 ) is the Relative Wealth Index (RWI), derived from ‘ground truth’ DHS data. This is an asset-based measure of wealth constructed by implementing a Principal Component Analysis (PCA) of questions related to asset holdings of households surveyed in the DHS (such as electricity, TV, motorcycle, roof material etc.) (Rutstein & Johnson 2004 ). The resulting RWI is in itself a standardised value, with a mean of zero and a standard deviation of one. Crucially, the RWI is defined relative to other locations within the country. It is therefore not comparable across countries, unless transformed into an absolute wealth index. Since a conversion into absolute values requires relying on per capita GDP and Gini coefficient estimates, which are unavailable or unreliable at the right temporal or geographical resolution in many LMICs, we use the RWI in LMICs as our main measure of household wealth. Because we are particularly interested in how temperature variability affects poverty, we would ideally want to directly observe households’ poverty status, which is typically measured using consumption or income aggregates. However, such data is unavailable at a sufficiently high spatial resolution needed for this study. While we acknowledge that using a consumption or income-based measure would be more suitable to analyses of poverty outcomes, asset-based measures can nonetheless provide important insights into the relationship between temperature variability and household wealth, all while being available at an unprecedented spatial granularity and covering areas across LMICs with relatively higher levels of poverty, compared to high-income countries.

Figure  3 plots the micro-estimates of the RWI derived from machine learning models in Chi et al. ( 2022 ) for all LMICs. Each data point represents a grid cell with a 2.4 km spatial resolution and the associated predicted value of relative wealth. Values should be interpreted as relative to other locations within the same country, where a value of 0 represents the mean. Given the relative nature of our wealth measure, we apply our econometric framework, Spatial First Differences, within countries, i.e. we are only comparing and differencing with regards to observational points that lie within the same country.

Relative Wealth Index at 2.4 km spatial resolution

Aggregating Data to the Right Spatial Resolution

The first step to processing our temperature and household wealth data for analysis was aggregation to the same spatial resolution. While the ERA5 temperature dataset offers observations at a regular grid of 0.5 longitude by 0.5 latitude (around 50 km at the equator), the micro-estimates of wealth from Chi et al. ( 2022 ) are produced at an exceptionally highly disaggregated spatial level of an irregular grid with a 2.4 km spatial resolution. To obtain comparable units of observation, we aggregated the wealth data to a regular grid of 0.5 × 0.5 degrees. Given our earlier discussion on the RWI being a relative measure of wealth, and therefore its non-comparability across borders, we perform this aggregation within countries. For each 0.5 × 0.5 degrees grid cell, we calculate the mean values of the RWI and ML model error across all data points at a 2.4 km spatial resolution that fall within that grid cell. The process is visualised in Fig.  4 for the example of Ethiopia. The left side shows the initial structure of our household wealth data, with an irregular grid of data points at a 2.4 km spatial disaggregation, each representing an estimate of the RWI. The right-hand side is the outcome of the aggregation, with averaged values overlaid on a regular 0.5 degrees grid.

Aggregating household wealth data to a 0.5 degrees spatial resolution

Once both datasets had the same spatial resolution of 0.5 degrees, our next step was to set up a geographical panel structure. This is analogous to methods used with longitudinal panel data in relation to time, except that the geographical panel is defined in space, exploiting the spatial structure of the data. The two identifying dimensions of our panel are: 1) the country code (for which we use i as an index); and 2) ordered grid cells starting from the most north-western observation in each country, moving east, then turning south and reversing the order to go east–west in a ‘snake-like’ manner (for which we use g as an index). A sketch of this process is presented in Fig.  5 .

Building a geospatial panel dataset

A crucial question relates to a case where a single temperature grid cell crosses borders, falling within two different countries – and which country it should be assigned to. This is illustrated in Fig.  6 , where yellow dots represent data points assigned to Ethiopia, red dots to Eritrea, and the orange dots for those that fall at the border between the two countries. In this case, we would count the grid cell twice : once as a data point forming part of the panel for Ethiopia, and once as part of the panel for Eritrea. We aggregate the higher resolution wealth data in such a way, however, that each country’s data is only counted once. This means that we have duplicate temperature observations in our data for these overlapping grid cells, but no duplicate wealth observations.

Grid cells crossing borders

Population Density Estimates

To control for population density in a sub-set of our regressions, we use UN-adjusted population density estimates at a 30 arc spatial resolution (approximately 1 km at the equator) for 2018 produced by WorldPop ( 2018 ). These high-resolution estimates are available for all 89 countries in our sample, listed in Table 10 in the Annex. The original units are number of people per square kilometre based on country totals adjusted to match the corresponding official United Nations population estimates that have been prepared by the Population Division of the Department of Economic and Social Affairs of the United Nations Secretariat. In order to match the spatial resolution of the temperature data, we aggregate the population estimates to a 0.5-degree resolution using the “sum” resampling algorithm and divide the population estimates by 1,000. The resulting unit is number of people (1,000s) per 0.5-degree grid cell.

The Final Dataset

The structure of our dataset is shown in Table  1 . At the original spatial resolution of 2.4 km, we have 2,428,569 micro-estimates of wealth. After spatial aggregation and calculation of first differences, we obtain a total of 18,594 grid cells across 89 LMICs spread among six regions. Although in the original (Chi et al. 2022 ) paper, the micro-estimates of wealth are calculated for all 135 LMICs, only 89 of those were made publicly available for download at the time we accessed the data. Footnote 6 The regions that are represented in our data are: Sub-Saharan Africa, Latin America and the Caribbean, South Asia, Middle East and North Africa, East Asia and the Pacific, and Europe and Central Asia. See Table 10 in the Annex for a full list of countries in our data and corresponding regions.

Our Analytical Approach

Spatial first differences.

To identify the causal effect of temperature variability on household wealth in LMICs, we adopt a SFD research design, which can be used to reduce the effect of unobserved heterogeneity in settings with cross-sectional data that is dense in space (Druckenmiller & Hsiang 2018 ). This methodology is conceptually analogous to first differences in time, except the difference is calculated between two spatially adjacent units, i.e. the time index is replaced by a location in space. Since only the neighbouring observations in space are compared, the conditional mean assumption necessary for identification only needs to hold locally , as shown in the equation below .

where observations are indexed by \(i\) (the country they belong to), \(g\) (their location within the country), \(Y\) is the outcome variable (relative household wealth), \(T\) is the treatment variable (temperature variability in our case), and X represents a set of control variables (rainfall and seasonal temperature variation in our main specification, other controls in additional robustness checks and sensitivity analyses). This assumption states that conditional on receiving the same treatment and a set of controls, two spatially adjacent observations should have the same expected outcome. This is a much weaker identifying assumption than the conditional mean assumption in cross-sectional regressions in levels, which requires that all observations with the same treatment conditional on observable covariates need to have the same expected outcome. The SFD design thus reduces the bias due to omitted variables by exploiting the variation between two units adjacent in space that are ‘comparable’ and differencing out the error term that might contain unobserved heterogeneity.

Spatial First Differences using Temperature Variability and Micro-Data on Household Wealth

Main approach.

Our main specification is a linear model of the relationship between temperature variability and asset-based household wealth, with observations indexed by their location in space \(g\) within each country \(i\) .

Here, \(\Delta\) denotes first order differences with respect to \(g\) , \(rwi\) represents the RWI, \(temp\) is the degree of temperature variability, \(rain\) additionally controls for rainfall, \(season\) filters out the influence of seasonal temperature variation, and \(\varepsilon\) is the error term. The unit of observation is the 0.5 degrees latitude by 0.5 degrees longitude grid cell. Standard errors are clustered at the country level to account for spatial autocorrelation across units. The coefficient of interest is \({\beta }_{1}\) , which identifies the causal impact of a one-unit change in the degree of temperature variability on the RWI. Given the nature of this RWI, this coefficient should be interpreted as the average unit change in the wealth index at grid-cell level relative to other grid cells within a country caused by a one-unit change in the degree of temperature variability.

The SFD design relies on calculating first differences between spatially adjacent units, indexed by \(g\) . To check that our results are not sensitive to a specific implementation of the indexing approach, i.e. the snake approach presented in Fig.  5 and hence the direction in which first differences are defined, we implement the SFD estimation with a second indexing approach in addition to the one presented in Fig.  5 . If the identifying assumption of SFD is met, the coefficients obtained by different ways of indexing the data should be similar. For instance, running the ‘snake’ in a different direction (e.g. going south-north first and then east–west) should not lead to different results, as we would still be differencing between neighbours. In this paper, we present results from two approaches:

Method 1: Snake : within each country, we defined the order in which observations should be numbered by the ‘snake’ approach as presented in Fig.  5 . We started from the north-western most observation, moving east while keeping latitude constant, then turning south and reversing the order to go east–west in a ‘snake’. At each step, the first differences are calculated. Note that though observations are dense in space, in some cases the next available grid cell might be far away, potentially violating the local conditional mean assumption. In order to ensure comparability between every two adjacent observations, an empty row was inserted whenever the next observation fell outside of a 100 km radius and no difference was calculated.

Method 2: 100 km radius : for each observation, we also identified all observations that lie within a 100 km radius of the initial point and fall within the same country and calculated the average of the outcome variable (RWI), temperature variability, and control variables (rainfall and seasonal temperature variation). We then subtracted the values for the initial observation from this 100km average and regressed the first differences. In this case, the ‘neighbouring’ cell is an artificial unit that represents averages across neighbours in all directions, hence being a good robustness check on our first method.

The indexing approach for both methods is shown with the example of Northern Ethiopia in Fig.  7 : the ‘snake’ in the left-hand side panel and the radius approach in the right-hand side panel.

Two methods for indexing data

Heterogeneity Analysis by Region, Income Classification, and Koppen-Geiger Climate Classification

We also conduct heterogeneity analyses to understand if estimated effects vary by geographical region, by running our main specification, but additionally including an interaction term between the spatially differenced day-to-day temperature variability at grid cell level and a categorical variable indicating the region each country and grid cell is located in. The full model used for heterogeneity analysis is shown in Eq. ( 3 ) below. We follow the World Bank Country Classification Footnote 7 for the fiscal year 2024 to define which region each of the countries in our sample belongs to: Sub-Saharan Africa, Latin America and the Caribbean, South Asia, Middle East and North Africa, East Asia and the Pacific, and Europe and Central Asia. A full list of countries and regions in our dataset is shown in Table 10 in the Annex.

Our coefficients of interest are \({{\varvec{\upbeta}}}_{5}\) , which is the vector of individual coefficients for each of the interaction terms of regions and differenced temperature variability included in the regression (represented by the \(\text{i}.region\) term, with the exception of the base region). Each of these coefficients in \({{\varvec{\upbeta}}}_{5}\) represent how the impact of temperature variability on the RWI changes when comparing the respective region with the base region. If our estimated coefficients ( \({\widehat{{\varvec{\upbeta}}}}_{5})\) are significantly different from zero, this indicates that effects vary significantly across regions – compared to the base region. We set the base region to Europe & Central Asia. We are also interested in the sum of \({\upbeta }_{1}\) with each of the individual components of \({{\varvec{\upbeta}}}_{5}\) (e.g. \({\upbeta }_{1}+{\beta }_{5,Latin America}\) ), which represents the full effect of temperature variability on the RWI in each region. If estimates of combinations of these coefficients are different from zero in any given region, this indicates that there is a significant effect of temperature variability in this region. Footnote 8

We replicate this approach for an additional analysis using a different type of country grouping: the World Bank’s 2018 income classification system. Footnote 9 This system assigns countries to low-income, lower-middle-income, and upper-middle-income groups based on national per-capita income. For countries in our sample, the 2018 classification can be found in Table 10 in the Annex. We use the same specification as in Eq. ( 3 ), just changing the groupings of countries from regions to income groups. Again, we are interested in whether effects vary across groupings and whether effects are significantly different from zero within each group.

Similarly, we conduct heterogeneity analysis by climate zones, using the Koppen-Geiger climate classification. We estimate Eq. ( 4 ) below, where \(KG\) indicates a binary variable which assigns each grid cell to the Koppen-Geiger climate classification. We rely on a global 0.5 degree resolution dataset, which classifies each grid cell according to the Koppen-Geiger climate classification (Beck et al. 2018 ). We use the newest release of the dataset, which covers the present-day classification (1991–2020). The five main groups according to the Koppen-Geiger climate classification are: A (tropical), B (arid), C (temperate), D (continental), and E (polar).

The coefficients in the vector \({{\varvec{\upbeta}}}_{5}\) indicate how the impact of day-to-day temperature variability on household wealth differs for each of the Koppen-Geiger climate classifications, when compared to the base category. We set the base category to be Group A (tropical climate). The sum of coefficients \({\upbeta }_{1}\) and each individual component of \({{\varvec{\upbeta}}}_{5}\) – as above – shows the full effect of temperature variability on the RWI for each of the Koppen-Geiger climate categories.

Main Robustness Checks

Spatial double difference.

We estimate a few key variations to our main specification defined above for robustness check purposes. First, we follow Druckenmiller and Hsiang ( 2018 ) and take higher order differences of observations in adjacent units within countries and estimate the SFD on these. This means that we rely on the same linear model as our main specification but take double spatial differences of all variables. This so-called Spatial Double Difference (SDD) model is given by the equation below:

Here, \({\Delta }^{2}\) represent the spatial ‘double-difference’ transformations, i.e. for example: \({{\Delta }^{2} rwi}_{ig}= {\Delta rwi}_{ig}-\Delta rw{i}_{i\left(g-1\right)}.\) Druckenmiller and Hsiang ( 2018 ) argue that if the estimated coefficients of the simple SFD procedure and SDD are similar, it is likely that local conditional independence holds.

Controlling for Modelling Uncertainty in Underlying Wealth Data

In a second robustness check, we re-estimate all our main specifications using an amended household wealth dataset. Given that our household wealth data is based on modelling output, i.e. predictions produced by Chi et al. ( 2022 ), one could argue that our estimates are influenced by prediction uncertainty or prediction modelling error. In order to assess the plausibility of this, we use the modelling error data at 2.4 km grid-level shared by Chi et al. ( 2022 ) and re-estimate our specification on two new datasets: 1) excluding all countries with average model error in the top 50% of the distribution, and 2) excluding all 2.4 km grid-cells that lie in the top 50% of the modelling error distribution. These datasets are hence comprised of ‘high-quality’ predictions of our key outcome variable only.

Reverse Causality

It is possible that the relationship between temperature variability and relative household wealth could be driven by reverse causality. For example, if wealthier households build up their living environment with increased infrastructure, which in turn affects temperature variability, then higher levels of household wealth in certain grid cells might be driving differences in temperature variability across cells and not vice-versa. Inspired by Linsenmeier's ( 2021 ) approach to tackle a similar problem, we deal with this by re-running our main specifications with two different measures of long-term historical temperature variability: one covering the years 1989–2001 and one covering the years 1989–2017. Both measures precede the year for which the relative wealth index is estimated (2018), while the first one precedes any time-period in which underlying ground-truth DHS data was collected. We argue that there is no possibility of reverse causality in the form of observed wealth distributions affecting temperature variability in these specifications.

Controlling for Population Density

We run additional specifications in which we include population density as a control variable to investigate and address two key issues: first, our main outcome variable is measured using predictions from a machine learning algorithm. Some of the features (though as we explain in the above section the less important ones) used to produce these predictions – such as road density, population, and if urban or built up – could be related to temperature variability. Our analysis might hence be influenced by these underlying relationships between predictors and temperature variability. We argue that including population density in our specifications allows us to assess how sensitive our results are to controlling for some of these features. Second, it is possible that our SFD approach is not able to appropriately account for how temperature variability might be related both to household wealth and the location people live in. The relationship we identify might simply reflect how wealthier households live in more densely populated areas – which have different temperature variability characteristics than less densely populated areas. Including population density in our specifications means that we are comparing grid-cells with similar density levels when trying to identify the effect of temperature variability on wealth, hence controlling for this issue.

Headline Results

Ordinary least squares.

To illustrate the potential Omitted Variable Bias (OVB) problem that could be introduced by unobserved heterogeneity in cross-sectional regressions in levels, we first run a simple Ordinary Least Squares (OLS) regression of temperature variability on the RWI, without taking first differences. Table 2 shows the results for four models: (1) no control variables, (2) including a control for rainfall, (3) additionally controlling for seasonal temperature variability, and (4) including country fixed effects. Standard errors are clustered at the country level to allow for autocorrelation. In all cases, the coefficient of day-to-day temperature variability on the RWI is positive and statistically significant at the 1% level, indicating a positive impact of temperature variability on the asset-based index of wealth. More specifically, a one-unit increase in the degree of day-to-day temperature variability is found to be related to an increase in the RWI of 0.04–0.10 units.

This result is not only counterintuitive, as it suggests that higher degrees of climate variability possibly increase wealth, but also most likely does not have a meaningful interpretation due to the unobserved heterogeneity that we expect in the error term, biasing the coefficient. The identifying assumption for the cross-sectional regression in levels states that all units of observation that receive the same treatment (temperature variability) should have the same expected outcome conditional on control variables. This is unlikely to hold, as we are regressing data points from 89 LMICs. Each country (and – in the case of column (4) – each region within a country) has different characteristics that might affect household wealth through channels other than temperature variability. The omission of these factors in the regression will cause OVB. On the other hand, in an SFD setting, the conditional mean assumption only needs to hold locally, between points close in physical space. We look at results from our SFD estimations next.

Spatial First Differences Method 1: Snake

In Table  3 , we present our main SFD estimation results, based on specification Eq. ( 2 ) and on Method 1 (‘snake’ SFD) indexing method presented in Fig.  5 . We present three models, which only differ by covariates included in the regressions. The results vary starkly from OLS results in Table  2 . First, the coefficient on the day-to-day temperature variability variable has now turned negative, and significantly so. Across all three models, it is highly significant.

These results imply that – once accounting for unobserved heterogeneity by performing spatial first differences – an increase in temperature variability is significantly and negatively related with average relative wealth among households in each geographical grid-cell in our sample. The negative effect of a one-unit increase in temperature variability on relative wealth lies between -0.16 and -0.19 on average across all countries in our sample. We interpret these results to imply that temperature variability negatively affects household wealth and – as a consequence – likely increases poverty in LMICs.

Spatial First Differences Method 2: 100 km radius

To demonstrate that these results are not a feature of our ‘snake’ indexing approach, Table  4 presents SFD results implemented using Method 2 (100 km radius) approach. This means that we run specification Eq. ( 2 ) by taking first differences of outcomes between every unit of observation and the 100 km radius average. As before, the first model in column (1) considers the relationship between temperature variability and household wealth without any control variables. Column (2) adds a control for rainfall, and column (3) additionally adjusts for the seasonal temperature cycle. These results confirm the findings derived from implementing SFD using our ‘snake’ approach: in all cases, there is a highly statistically significant (at 1%) negative impact of day-to-day temperature variation on the RWI. In fact, our main coefficient estimates of interest are remarkably similar to the ‘snake’ approach, varying between -0.16 and -0.19.

Heterogeneity Analysis by Region and Income Classification

We present results from our heterogeneity analysis by region in Fig.  8 . We plot estimated combined coefficients on temperature variability and the regional interaction terms in specification Eq. ( 3 ) – i.e. \(\widehat{{\beta }_{1}+{{\varvec{\beta}}}_{5}}\) – together with the 95% confidence interval, indicated by the width of the whiskers in the graph below. A full regression table that shows coefficients, including the specific coefficients estimated for each interaction term and control variables included in this specification, is presented in Table 13 in the Annex.

Effect of day-to-day temperature variability by region

The results show that in Sub-Saharan Africa, Middle East & North Africa (MENA), East Asia & the Pacific, and South Asia the effects of day-to-day temperature variability on household wealth are negative and significantly different from zero. As shown by the value of the coefficients, the magnitude of the negative effect is the largest in Sub-Saharan Africa, followed by MENA, East Asia & the Pacific, and South Asia. Effects are not statistically significantly different from zero in Europe & Central Asia.

In Fig.  9 , we present results from a similar heterogeneity analysis, this time grouping countries according to the World Bank’s income classification system (2018). Again, we plot full effects, where this time the interaction terms are over country income groupings. A full regression table is presented in Table 18 in the Annex. The results show that while across all groupings we estimate effects of temperature variability on wealth to be negative, the magnitude of the effect is largest among low-income countries. Because poverty levels are highest among this group of countries, and our findings imply that temperature variability negatively affects wealth most significantly here, we interpret this as evidence that suggests that temperature variability also likely increases poverty.

Effect of day-to-day temperature variability by World Bank income classification

Heterogeneity Analysis by Koppen-Geiger Climate Classification

In addition to heterogeneity by region, we show how the effects of temperature variability on household wealth vary by climatic zones into which grid-cells fall, which might cross country and regional boundaries. We use the Koppen-Geiger climate classification, which has five main groups: A (tropical), B (arid), C (temperate), D (continental), and E (polar) (Beck et al. 2018 ). As before, we plot estimated combined coefficients on temperature variability and the climate classification interaction terms from specification Eq. ( 4 ) – i.e. \(\widehat{{\beta }_{1}+{{\varvec{\beta}}}_{5}}\) – together with the 95% confidence interval, in Fig.  10 . The full regression table is shown in Table 14 in the Annex.

Effect of day-to-day temperature variability by Koppen-Geiger climate classification

The results show that the negative effect of temperature variability on household wealth is largest for group A in the Koppen-Geiger climate classification. This means that households in tropical areas are most impacted by temperature variability. After group A, households in group B (arid) and group C (temperate) experience the largest magnitude of the negative effect of temperature variability on wealth, although the confidence intervals of the two estimates overlap – meaning that there is no statistically significant difference between the effects estimated for arid and temperate areas. In the case of groups D (continental), and E (polar), the estimated effect is not statistically different from zero. Footnote 10

Robustness Checks

We also implemented a ‘double-difference’ transformation to our data to then implement the SDD estimation as a robustness check. We present results in Table  5 , together with the original SFD results. The SDD results are presented in the rows with the tag (2nd diff). The key finding here is that the significant negative relationship between temperature variability and relative wealth persists. Estimated coefficients across our models are now smaller (-0.11 to -0.14), but this is expected, given possible attenuation bias that results from the double-difference transformation (Druckenmiller & Hsiang 2018 ). Again, these results strengthen our headline findings of the significant negative effects that increases in temperature variability have on household wealth.

Limiting the Wealth Data to ‘High-Quality’ Predictions

We also re-estimate our main specifications for observations with ‘high-quality’ RWI predictions only. We determine ‘high-quality’ observations based on model error at two levels of disaggregation: 1) country level, and 2) 2.4 km grid-cell level.

Table 6 presents the results for the Method 1 (‘snake’ SFD) approach and SDD using ‘high-quality’ predictions at the country level. This involves re-running our main regression models for a sub-set of countries that have average model error in the bottom 50% of the distribution. Despite the lower number of observations resulting from this, our key coefficient estimates are in line with previous findings and remain highly statistically significant. Across specifications including first- and second-order differences and a set of control variables, we obtain a negative effect of day-to-day temperature variability on household wealth in the range of -0.15 to -0.19.

The results from applying Method 2 (100 km radius SFD) to the sub-set of ‘high-quality’ country level observations is shown in Table  7 . The coefficients on day-to-day temperature variability have values between -0.16 and -0.19, similar to those obtained with Method 1 (‘snake’ SFD) in Table  6 . Our second approach to accounting for model error is to re-estimate our main specifications for a sub-set of 2.4 km grid-cells that have model error in the bottom 50% of the distribution. Results for Method 1 (‘snake’ SFD) and Method 2 (100 km radius SFD) are shown in Table 11 and Table 12 in the Annex, respectively. Using only ‘high-quality’ observations at the grid-cell level, we obtain key coefficients largely in line with our previous findings, in the range of -0.09 to –0.13. This suggests that our headline results were unlikely to be significantly influenced by uncertainty related to predictions of RWI, as measured by the grid-cell level model error.

One concern with the interpretation of our earlier results relates to reverse causality. As described above, we deal with this by regressing the RWI on a long-term measure of day-to-day temperature variability for two time periods which precede the household wealth data. We use the average day-to-day temperature variability in two periods: 1) 1989–2001; and 2) 1989–2017. For the temperature variability measure averaged for 1989–2001, the results from an SFD regression using the ‘snake’ method, as well as the spatial double difference, are shown in Table  8 . The coefficient on day-to-day temperature variability remains negative and highly statistically significant. Full regression tables, including for the average between 1989–2017 and using the 100 km radius approach are shown in Table 15 , Table 16 , Table 17 in the Annex. Our results remain unchanged across different specifications. We interpret this to mean that this relationship is unlikely to be driven by reverse causality.

In column (4) of Table  9 , we present results of our main SFD specification (using Method 1 ‘snake’) when including population density. Our main coefficient of interest on temperature variability and related significance levels do not change meaningfully when compared to the specification in column (3), which excludes population density as an explanatory variable. This can be interpreted in three ways. First, we argue that this indicates that our results are not sensitive to the inclusion of and are unlikely to be driven by a set of predictors included in the original algorithm used by Chi et al. ( 2022 ). Basically, this is evidence that our results are not resurfacing a relationship between household wealth and temperature variability that was previously introduced by the algorithm used to predict the RWI across LMICs. Second, we argue that this is evidence that our results are not driven by a simple correlation between temperature variability, wealth, and high population density – and by implication urbanisation – which our SFD approach might not have been able to deal with appropriately. Finally, these results indicate that location choice cannot explain the mechanism by which temperature variability might be affecting household wealth: even when holding population density levels constant, and after the SFD transformation, we find that higher temperature variability is associated with lower household wealth.

We leverage a newly available high-resolution ‘big data’ dataset on relative household wealth and poverty from Chi et al. ( 2022 ) and remote sensing temperature data from the ERA5 global climate time-series data in a SFD framework to explore a key question around the causal impact of day-to-day temperature variability on household wealth in LMICs. In the age of climate change, a key concern is understanding the effects of increasing temperature variability on social and economic outcomes in the regions that are projected to be most affected. Our research uses an innovative approach which shows that high-resolution estimates of household wealth, derived from ML algorithms, as well as novel analytical methods in causal inference, can be used in downstream research to investigate policy-relevant questions – particularly around climate-responsive social protection, as discussed below.

Our findings show a highly statistically significant negative effect of increases in day-to-day temperature variability on household wealth in LMICs, which is remarkably robust across different econometric specifications. This result provides evidence that households in LMICs will likely be negatively impacted by increases in temperature variability. In light of this, our findings highlight the importance of building climate resilience amongst vulnerable populations in LMICs. One potential channel is through climate-responsive social protection programmes, such as anticipatory cash transfers triggered by climatic shocks, or agricultural insurance schemes indexed to weather conditions.

Climate projections predict an unequal effect of climate change on temperature variability trends globally, with LMICs expected to see the largest increase in temperature variations (Bathiany et al. 2018 ). This highlights the need to identify geographies where the impact of day-to-day temperature variability on household wealth is strongest – and conversely, weakest. Our heterogeneity analysis shows that the magnitude of negative effects of day-to-day temperature variability on household wealth is largest in Sub-Saharan Africa, followed by Middle East & North Africa, East Asia & Pacific, and South Asia, when compared to Europe & Central Asia. In addition, our results show that negative effects are largest among low-income countries, compared to lower-middle- or upper-middle-income countries. We also conduct heterogeneity analysis by climate zones and show that households in tropical climates will be most affected by the negative effects of temperature variability on household wealth, followed by arid and temperate climates. This implies that the importance of climate-responsive social protection is largest in those geographies and climate zones. Future studies could undertake a more spatially disaggregated heterogeneity analysis. The high resolution of household wealth data could be used to indicate small areas that will be most impacted, up to a 2.4 km spatial resolution. This disaggregation in the data could inform geographical targeting of social policy, allowing for identification of populations most affected by climatic shocks.

Finally, it is important to keep in mind that our analysis remains cross-sectional in nature. Therefore, it provides only a snapshot of the relationship between day-to-day temperature variability and asset-based household wealth at a particular point in time, rather than an insight into poverty dynamics over time. This is because the micro-estimates of wealth that we rely on were only produced for the latest available ‘ground truth’ DHS data. Such small area estimates of poverty, produced either using machine learning or statistical modelling techniques, often rely on non-traditional ‘big data’ sources of information as its predictors. This offers an opportunity to leverage high-resolution datasets that are frequently updated, such as geo-spatial and mobile phone data, in order to produce dynamic high-resolution and high-frequency ‘poverty maps’, which change the spatial distribution of poverty as new information becomes available. Our analysis, as well as any future studies looking at the relationship between climate change and poverty using high-resolution data, could then investigate poverty dynamics. Small area estimation techniques could also be used to estimate high-resolution consumption or income data, in addition to asset-based wealth measures used in this study.

Data Availability

Data is provided within the manuscript.

This is the case for Indonesia, where the latest DHS data available to the authors is from 2002–03.

The detailed list of all ground truth data sources can be found in Table S1 of the online Appendix . The latest DHS data comes from 2018. However, in some countries the survey from that time period was not available. In that case, an earlier DHS was used.

These are: number of mobile devices, number of WiFi access points, number of cell towers, number of Android devices, number of iOS devices. The remaining 5 features are: radiance, road density, population, if urban or built up, and one out of over 100 satellite images.

Specifically, the ML model explains 72% of variation when data from 15 census countries is pooled together and 86%, on average, when countries are considered individually.

The lower figure is obtained when aggregating individual geo-referenced data points from independent data sources to the 2.4 km grid-cell level and the higher figure comes from aggregation at the lowest administrative unit in those countries – cantons for Togo and local government areas for Nigeria.

The data is publicly available to download from the Humanitarian Data Exchange site: Relative Wealth Index—Humanitarian Data Exchange (humdata.org) . At the time we accessed the data, the latest version was from 8th April 2021. All the analysis in this paper is based on this version. Since then, new micro-estimates for countries not originally available for download were added to the website. We will continue accessing new micro-estimates of wealth as they become available for public download and update our results.

Full list of countries and regions under the World Bank classification can be found here: World Bank Country and Lending Groups – World Bank Data Help Desk .

In practice, we estimate the combinations of these coefficients using the ‘lincom’ command in R, after running specification (4). See https://search.r-project.org/CRAN/refmans/biostat3/html/lincom.html .

A full list of countries by income classification, including for 2018, can be found here: https://datahelpdesk.worldbank.org/knowledgebase/articles/906519 .

Note that most of our grid cells fall into groups A-C, with only a small number of grid cells classified as either group D or group E. The exact numbers are: 9,260 (50%) in Group A, 6,069 (33%) in Group B, 2,658 (14%) in Group C, 276 (2%) in Group D, and 315 (2%) in Group E.

Andalón M, Azevedo JP, Rodríguez-Castelán C, Sanfelice V, Valderrama D (2014) Weather shocks and health at birth in Colombia. Policy Research Working Papers. The World Bank. https://doi.org/10.1596/1813-9450-7081

Asfaw S, Maggio G (2018) Gender, weather shocks and welfare: evidence from Malawi. J Dev Stud 54(2):271–291

Article   Google Scholar  

Banerjee A, Duflo E (2011) Poor economics. PublicAffairs

Baquie S, Fuje H (2020) Vulnerability to poverty following extreme weather events in Malawi. Working Paper. World Bank, Washington, DC. https://doi.org/10.1596/1813-9450-9435

Bathiany S, Dakos V, Scheffer M, Lenton T (2018) Climate Models Predict Increasing Temperature Variability in Poor Countries. Sci Adv. https://doi.org/10.1126/sciadv.aar5809

Beck HE, Zimmermann NE, McVicar TR, Vergopolan N, Berg A, Wood EF (2018) Present and Future Köppen-Geiger Climate Classification Maps at 1-Km Resolution. Sci Data 5(1):180214. https://doi.org/10.1038/sdata.2018.214

Brunckhorst B (2020) Rural mobility and climate vulnerability: evidence from the 2015 drought in Ethiopia. 2020–17. CSAE Working Paper Series. Centre for the Study of African Economies, University of Oxford. https://ideas.repec.org/p/csa/wpaper/2020-17.html

Cachon GP, Gallino S, Olivares M (2012) Severe weather and automobile assembly productivity. SSRN Scholarly Paper ID 2099798. Social Science Research Network, Rochester, NY. https://doi.org/10.2139/ssrn.2099798

Chi G, Fang H, Chatterjee S, Blumenstock JE (2022) Microestimates of Wealth for All Low- and Middle-Income Countries. Proc Natl Acad Sci 119(3):e2113658119. https://doi.org/10.1073/pnas.2113658119

Chonabayashi S, Jithitikulchai T, Qu Y (2020) Does agricultural diversification build economic resilience to drought and flood? Evidence from poor households in Zambia. African Journal of Agricultural and Resource Economics 15(1):65–80. https://doi.org/10.53936/afjare.2020.15(1).05

Dell M, Jones B, Olken B (2008) Climate change and economic growth: evidence from the last half century. Am Econ J Macroecon 4 (July). https://doi.org/10.3386/w14132

Druckenmiller H, Hsiang S (2018) Accounting for unobservable heterogeneity in cross section using spatial first differences. Working Paper. Working Paper Series. National Bureau of Economic Research. https://doi.org/10.3386/w25177

Foltz JD, Traore N (2018) Temperatures, productivity, and firm competitiveness in developing countries: evidence from Africa. In: 2018 annual meeting, August 5–7, Washington, D.C. Agricultural and Applied Economics Association, p 274434.  https://ideas.repec.org/p/ags/aaea18/274434.html

Ge J, Liu Qi, Zan B, Lin Z, Sha Lu, Qiu Bo, Guo W (2022) Deforestation Intensifies Daily Temperature Variability in the Northern Extratropics. Nat Commun 13(1):5955. https://doi.org/10.1038/s41467-022-33622-0

Haile B, Signorelli S, Azzarri C, Johnson T (2018) Welfare Effects of Weather Variability: Multi-Country Evidence from Africa South of the Sahara. PLoS ONE. https://doi.org/10.1371/journal.pone.0206415

Harris I, Osborn TJ, Jones P, Lister D (2020) Version 4 of the CRU TS Monthly High-Resolution Gridded Multivariate Climate Dataset. Sci Data 7(1):109. https://doi.org/10.1038/s41597-020-0453-3

Hersbach H, Bell B, Berrisford P, Hirahara S, Horányi A, Muñoz-Sabater J, Nicolas J et al (2020) The ERA5 Global Reanalysis. Q J R Meteorol Soc 146(730):1999–2049. https://doi.org/10.1002/qj.3803

Hirvonen K, Sohnesen TP, Bundervoet T (2020) Impact of Ethiopia’s 2015 Drought on Child Undernutrition. World Dev 131:104964. https://doi.org/10.1016/j.worlddev.2020.104964

Hovdahl I (2020) Deadly variation: the effect of temperature variability on mortality. No 01/2020. Working Papers. Centre for Applied Macro- and Petroleum economics (CAMP), BI Norwegian Business School. https://ideas.repec.org/p/bny/wpaper/0084.html

Ibanez AM, Romero J, Velasquez A (2021) Temperature shocks, labor markets and migratory decisions in El Salvador.  https://conference.iza.org/conference_files/ClimateChange_2021/velasquez_a26669.pdf

Ishaya (2014) Post-Adaptation Vulnerability of Cereals to Rainfall and Temperature Variability in the Federal Capital Territory of Nigeria. Ethiop J Environ Stud Manag 7:532. https://doi.org/10.4314/ejesm.v7i5.7

Jafino BA, Walsh B, Rozenberg J, Hallegatte S (2020) Revised estimates of the impact of climate change on extreme poverty by 2030. The World Bank

Kotz M, Wenz L, Stechemesser A, Kalkuhl M, Levermann A (2021) Day-to-Day Temperature Variability Reduces Economic Growth. Nat Clim Chang 11(4):319–325. https://doi.org/10.1038/s41558-020-00985-5

Letta Marco, Montalbano Pierluigi, Tol Richard (2018) Temperature Shocks, Short-Term Growth and Poverty Thresholds: Evidence from Rural Tanzania. World Dev 112:13–32

Linsenmeier M (2021) Temperature variability and long-run economic development. 110499. LSE Research Online Documents on Economics. London School of Economics and Political Science, LSE Library. https://ideas.repec.org/p/ehl/lserod/110499.html

Markham S, Markham I (2005) Biometeorological Effects on Worker Absenteeism. Int J Biometeorol 49:317–324. https://doi.org/10.1007/s00484-004-0246-y

Melitz MJ (2003) The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity. Econometrica 71(6):1695–1725

Mendelsohn R, Nordhaus W, Shaw D (1994) The Impact of Global Warming on Agriculture: A Ricardian Analysis. Am Econ Rev 84(4):753–771

Google Scholar  

Moberg A, Jones PD, Barriendos M, Bergström H, Camuffo D, Cocheo C, Davies TD et al (2000) Day-to-Day Temperature Variability Trends in 160- to 275-Year-Long European Instrumental Records. J Geophys Res Atmos 105(D18):22849–22868. https://doi.org/10.1029/2000JD900300

Narloch U, Bangalore M (2016) Environmental risks and poverty: analyzing geo-spatial and household data from Vietnam. World Bank Policy Research Working Paper No. 7763. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2814397

Nkegbe PK, Kuunibe N (2014) Climate variability and household welfare in Northern Ghana. Working Paper 2014/027. WIDER Working Paper. https://doi.org/10.35188/UNU-WIDER/2014/748-6

Pape U, Wollburg P (2019) Impact of drought on poverty in Somalia. World Bank Working Paper, January.

Randell H, Gray C (2016) Climate Variability and Educational Attainment: Evidence from Rural Ethiopia. Glob Environ Chang 41:111–123. https://doi.org/10.1016/j.gloenvcha.2016.09.006

Rutstein SO, Johnson K (2004) The DHS wealth index August. https://dhsprogram.com/publications/publication-cr6-comparative-reports.cfm

Skoufias E, Rabassa M, Olivieri S (2011) The poverty impacts of climate change: a review of the evidence. World Bank Policy Research Working Paper 5622

Tam BY, Gough WA, Mohsin T (2015) The Impact of Urbanization and the Urban Heat Island Effect on Day to Day Temperature Variation. Urban Clim 12:1–10

The World Bank (2022) Poverty and shared prosperity: correcting course. https://www.worldbank.org/en/publication/poverty-and-shared-prosperity

WorldPop, WorldPop ( www.worldpop.org -School of Geography and Environmental Science, University of Southampton, Department of Geography and Geosciences, University of Louisville, Departement de Geographie, Universite de Namur, and Columbia University (2018) Global high resolution population denominators project - Funded by The Bill and Melinda Gates Foundation (OPP1134076)

Yevessé D (2021) Effects of temperature and rainfall variability on the net income of cereal crops in Togo: semi-parametric approach. AERC | African Economic Research Consortium. Research Paper 435. https://aercafrica.org/aerc-doc/effects-of-temperature-and-rainfall-variability-on-the-net-income-of-cereal-crops-in-togo-semi-parametric-approach/

Download references

Acknowledgements

This paper was funded by the Data and Evidence to End Extreme Poverty (DEEP) project. DEEP is a consortium of the Universities of Cornell, Copenhagen, and Southampton led by Oxford Policy Management, in partnership with the World Bank’s Development Data Group and funded by the UK Foreign, Commonwealth & Development Office.

Author information

Authors and affiliations.

Oxford Policy Management Limited, Ground Floor, 40-41 Park End Street, Oxford, OX1 1JD, United Kingdom

Ida Brzezinska & Paul Jasper

You can also search for this author in PubMed   Google Scholar

Contributions

IB and PJ conceived the concept for this project. IB processed the data and produced most analytical code. PJ provided project supervision and reviewed all analytical code. Both authors contributed to interpreting the results. Both authors contributed to writing and approved the final manuscript.

Corresponding author

Correspondence to Ida Brzezinska .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

See ( Table 10 )

See (Table  11 )

See (Table  12 )

See (Table  13 )

See (Table  14 )

See (Table  15 )

See (Table  16 )

See (Table  17 )

See (Table  18 )

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/ .

Reprints and permissions

About this article

Brzezinska, I., Jasper, P. The Negative Effect of Temperature Variability on Household Wealth in Low- and Middle-Income Countries. EconDisCliCha (2024). https://doi.org/10.1007/s41885-024-00160-6

Download citation

Received : 28 March 2024

Accepted : 26 July 2024

Published : 28 August 2024

DOI : https://doi.org/10.1007/s41885-024-00160-6

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Temperature variability
• Climate change

JEL Classification

• Find a journal
• Publish with us
• Track your research