Land surface temperature and rainfall changes in the Brazilian Amazon during the last two decades: variability in space, season, and by deforestation intensity 

Abstract

In the Brazilian Amazon, large-scale and spatially explicit evidence of observed climate change in the recent decades is still scarce but important for monitoring areas that require climate action if changes persist over a long-term. Many areas in this region have faced high levels of deforestation during the last decades, leading to changes in energy fluxes that may amplify the effects on climate caused by globally increasing greenhouse gas (GHG) emissions. In this study we assessed how land surface temperature and rainfall changed across space and seasonal intervals during the last two decades. We also aimed to understand if highly deforested areas were more likely to experience positive or negative changes at a particular seasonal interval. We focused on the Amazon rainforest domain over Brazil, dividing the region into 0.5° grid cells. We used forest cover mapping from MapBiomas to compute total deforestation from 2003 to 2021 relative to the cell’s area. For land surface temperature (LST), we used data from the MODIS MYD11A1 product, selecting daily daytime observations. For rainfall volume we used daily estimates from CHIRPS. For each year from 2003-2021, we performed aggregations for the annual, driest quarter (three consecutive driest months based on the climatology) and wettest quarter for each cell, averaging LST and summing rainfall volume. We then subtracted the 2013-2021 to 2003-2012 averages in each cell, excluding extreme drought years to reduce changes from internal climate variability such as extreme El Niño.  We found that, over the last two decades, annual LST and rainfall increased on average in the Brazilian Amazon by 0.31°C and 4% respectively. At least 80% of cells experienced increases in LST at any of the time intervals, with warming greater than 1.5°C in some locations. Areas with substantial warming were predominant in southern Amazon at the annual scale and driest quarter, but widely dispersed in the wettest quarter. In terms of rainfall, increases were found for 68% of cells at the annual scale and ~56.5% in the other time intervals. The driest quarter had the largest frequency of rainfall reductions and increases >20%. In all time intervals, substantial rainfall decreases were found in the states of Roraima and southwestern Amazonas, whereas substantial rainfall increases were found in eastern Acre. For all cells with deforestation >10%, more than 70% had LST warming >0.5°C annually and during the driest quarter, in contrast to only 40% in the wettest quarter; in cells with no long-term deforestation (<=0.5%), the proportion was less than 8%. In all time intervals, areas with >10% deforestation had a frequency of both rainfall increases and decreases about 10-15 percentage points greater, suggesting no clear pattern for a greater likelihood in rainfall changes. We find that regional climate change in the Amazon is very dynamic in space and season, especially in terms of rainfall changes, requiring careful examination. We also demonstrate how deforestation has been leading to a greater intensity of LST warming in the region, representing a regional climate forcing that may be as important as increased GHG emissions.