Three sets of numerical experiments utilizing the Brazilian Earth System Model - BESM version 2.535,36,37 are done to estimate the impacts of the Amazon forest cover change, those originating from atmosphere and ocean circulation changes induced by global warming, and the combination of the two. The model experiments were labeled: for GHG scenarios (Historical - Hi, and RCP8.5 - R8); and land use scenarios (Fo – Forested, and Sa – Savannah)(see Supplementary Fig. S1), therefore generating a total of four simulation runs. For instance, HiSa represents the historical GHG scenario with the savannah land use pattern over the Amazon Basin, whereas R8Fo indicates the RCP8.5 scenario run with the forested land use pattern. The results presented are based on 28 simulation years, after a two years spinup run. Daily total precipitation, 2 m maximum and minimum air temperature, surface runoff, soil moisture, and atmospheric profiles of wind and specific humidity are used for the analyses and discussions that follow.
Annual precipitation change
The annual precipitation differences of the savannization and global warming scenarios relative to the HiFo control run are shown in Fig. 1. While the effects of the radiative forcing scenario for the forested simulation (R8Fo) depicts negative departures of annual precipitation mostly over the northern-northeastern border of South America (Fig. 1a), the pattern of rainfall reduction expands inland when land use changes are considered, both for current climate conditions HiSa (Fig. 1b) and future climate scenario R8Sa (Fig. 1c). The rainfall control run (HiFo) is shown in Supplementary Fig. S2. The severe rainfall reduction of up to 70% over areas in central Amazon for the HiSa shown in Fig. 1b is noteworthy. However, when both atmospheric radiative forcing and vegetation cover change are applied together, not only the area under 70% total annual reduction over the interior of the Amazon basin expands, but expressive rainfall reduction extends further south, over the Cerrado (Fig. 1c).
Dry season length change
The impact of land use and climate change on the length of the dry season over Brazil was evaluated by the distribution of consecutive dry days events (droughts length) within the 28 years of numerical simulations (described in the Methods). To gauge the effects of both global warming and Amazon land cover change on the length of the dry seasons, a severe drought index is defined by the length of consecutive dry days representing the longest 10% occurring over each grid point. Fig. 2 shows the spatial distribution of extreme drought events for the departures of the R8Fo, HiSa, and R8Sa scenarios relative to the HiFo simulation. It is noteworthy that while the larger effects of global radiative forcing in enlarging the length of dry spells are confined to the north-northeast portions of South America (Fig. 2a), the impacts of land use change over the Amazon are to enlarge the duration of dry spells southward, over the interior portion of the Amazon, and to southern Brazil (Fig. 2b). More so when the radiative forcing is combined with the savannah conditions imposed over the Amazon. The savannah condition intensifies the dry spell's length, increasing their duration by 60 days in the Amazon basin, and extending the length of the driest spells to the central-southeastern-southern portions of Brazil (Fig. 2c). The spatial distribution of extreme drought events for control run (HiFo) is shown in Supplementary Fig. S3. According to previous studies, the intensification of dry spell duration over deforested areas in Amazonia is connected to the lack of moisture recycling mechanisms, affecting cascade moisture recycling along the way38,39,12. Also, 33 found that deforestation over the Amazon basin increases the frequency of occurrence of longer dry seasons in the central-southern Amazon (by between 29 and 57%), depending on the deforestation scenario considered.
Hydrological cycle change
The bar diagram in Fig. 3a shows the area averaged annual rainfall departures for each Region of Brazil relative to the HiFo control run. It is noteworthy that the negative annual rainfall anomalies over the Amazon and Cerrado are enlarged when both land use change and global climate change are considered, with the R8Sa presenting the largest rainfall reduction over both regions (see the orange and red bars for the North and Midwest regions in Fig. 3a). These results also suggest that in the Northeast region, climate change reduces rainfall, while only vegetation change increases it. However, in the South and Southeast regions there was an increase in rainfall in all scenarios compared to the current climate, notwithstanding accompanied by the increase of dry season length (Fig. 3b).
The Fig. 3b shows the percent variation of the length of extreme drought events relative to the HiFo experiment for the land use and radiation forcing scenarios, averaged over each of the five regions of the country. The increase in the frequency of longer periods of consecutive dry days is noteworthy for all regions and all R8 scenarios, with the largest dry season increase for the R8Sa experiment (red bars in Fig. 3b) for all regions. The drying effect of deforestation can be further evidenced by the comparison of R8Fo and R8Sa indices (the green and red bars in Fig. 3b, respectively); most notably over the North, but also over the Midwest region.
Meridional moisture advection
As shown in the previous section, the replacement of the Amazon rainforest by savannah vegetation type has produced a pronounced increase in the duration of dry spell length, primarily in the Amazon region, but also extending to the central-southeastern parts of the continent. The duration of dry spells was invariantly enhanced by the effects of global warming. Other studies have presented evidence that the reduction of moisture in the basin affects remote regions on the continent, through the reduction of moisture transport by the southward-flowing low-level jets39,40,41. This process was evaluated here by the low troposphere meridional moisture advection, vertically integrated (from 975 to 700 hPa), over the area, averaged between the longitudes 55W-65W (dashed rectangle in the Supplementary Fig. S1) depicted in Fig. 4. For the current climate, forested experiment (HiFo), the region north of 3N gains moisture from the Atlantic Ocean (indicated by the positive advection in Fig. 4a) during the summer season (DJFM). Between 3N and 12S, delimiting the Amazon Basin (indicated by the vertical dashed lines in Fig. 4a) negative advection indicates that the region loses moisture to its neighboring southern region. This moisture is carried southward, beyond the Amazon Basin, to the latitude belt between 12S and 20S. Global warming scenarios, without vegetation change (experiment R8Fo), did not show a significant impact on meridional moisture advection relative to the HiFo control case (dashed green line in Fig. 4b), despite the increase in the specific humidity in the atmosphere due to the increase in global temperature (Supplementary Fig. S4). However, when vegetation type changes are considered, a pronounced reduction of southward moisture advection relative to the HiFo control case takes place. Over the Amazon Basin region, there is a significant reduction in negative advection (positive difference, continuous red line in Fig. 4b), a result of a lesser supply of soil humidity (discussed below). Beyond the south of the Amazon basin, the reduction in moisture positive advection (negative difference) indicates that in this region, the supply of moisture by the forest has decreased. This lower moisture advection was even more intense with the joint effect of global warming (dashed red line in Fig. 4b), suggesting that the joint effect of savannization and global warming may induce a deficit in rainfall in the rainy season in the regions south of the Amazon. The lack of rainfall in the Brazilian Pantanal during the summers of 2019 and 2020 was accompanied by reduced transport of warm and humid summer air from Amazonia42.
Air temperature change
To gauge the impacts of atmospheric radiative forcing and land use change over the Amazon on the occurrences of air temperature extremes, the daily maxima temperature predicted at every grid point, for each scenario, was considered for the 28 years long simulations. Fig. 5 shows the maximum air temperature anomalies for the warmest months over South America for experiments R8Fo, HiSa, and R8Sa relative to the HiFo control run. Similarly to other results43,44, the R8Fo experiment shown in Fig. 5a depicts positive anomalies, with the warmest departures occurring over the north-northeast and southern South America. On the other hand, the effects of the land use change scenario for the current radiative forcing experiment (HiSa) showed positive air temperature anomalies over the Amazon and negative anomalies eastward, over the Northeast region (Fig. 5b). The combined effects of radiative forcing and land use change scenario (Fig. 5c), depicted the largest temperature anomalies, reaching values as high as 14 °C above the HiFo control run in central Amazon, which represents actual maximum temperatures surpassing the value of 46 °C. The maximum air temperature for the control run (HiFo) is shown in supplementary Fig. S5. Experiment R8Fo produced an increment of 3.3 °C (average for the period 2073-2100) over the Amazon basin (shown in Supplementary Fig. 1c), and with the R8Sa simulation, the temperature increased by 5.4 °C relative to the HiFo control. These values are in the same range as shown by other results9.
The annual maximum thermal amplitude at each gridpoint over South America, as measured by the difference between the daily maximum temperature of the warmest month and the daily minimum temperature of the coldest month (Supplementary Fig. S6), increases with the warmer scenario, and adding land cover change, the amplitude increases even more. The occurrence of temperature extremes, together with the increase in the period of droughts, in the Amazon Basin and remote regions, can represent adverse consequences for agriculture and human health.
Soil wetness and runoff changes
Climate change can affect both groundwater content and recharge, which can threaten water availability for rural communities and cities45,46,47. The combined changes of precipitation and temperature patterns discussed above reduced surface soil moisture throughout South America (Fig. 6). The effects of global radiative forcing decreased the soil moisture by 10% most prominently in the northern part of South America and the Amazon basin (Fig. 6a). The land use change in current climate reduces the soil moisture to 15% over the Amazon basin (Fig. 6b). However, when considering the effects of global radiative forcing and Amazon land use change together, the soil dryness is even more prominent, spreading to almost the entire Brazil and neighboring countries (Fig. 6c). Other studies have shown that global warming and precipitation reduction are the main reasons for the reduction of total soil moisture content48. Such attenuation of soil moisture, on the other hand, will further aggravate the global water cycle disruption and enhance the variability of extreme meteorological disasters48.
The impacts of rainfall and temperature changes resulting from the atmospheric and land use change scenarios are inspected next. When climate change alone is considered, runoff decreases over northern Brazil and increases over South Brazil (Fig. 7a), as shown in other works49,50,51. However, coherently with the reduction of southward moisture advection due to the Amazon land use change discussed above, surface runoff reduction extends from the Amazon region to the midwest and northeast Brazil (Fig. 7b) and when both drivers are considered (Fig. 7c). The reduction of runoff in the Midwest region with the Amazon savannization is notable, in both radiative forcing conditions, with possible downstream adverse consequences in several services such as hydropower electric generation. The spatial distribution of soil moisture and runoff for control run (HiFo) are shown in Supplementary Figures S7 and S8.