Modern anthropogenic drought in Central Brazil unprecedented during last 700 years

A better understanding of the relative roles of internal climate variability and external contributions, from both natural (solar, volcanic) and anthropogenic greenhouse gas forcing, is important to better project future hydrologic changes. Changes in the evaporative demand play a central role in this context, particularly in tropical areas characterized by high precipitation seasonality, such as the tropical savannah and semi-desertic biomes. Here we present a set of geochemical proxies in speleothems from a well-ventilated cave located in central-eastern Brazil which shows that the evaporative demand is no longer being met by precipitation, leading to a hydrological deficit. A marked change in the hydrologic balance in central-eastern Brazil, caused by a severe warming trend, can be identified, starting in the 1970s. Our findings show that the current aridity has no analog over the last 720 years. A detection and attribution study indicates that this trend is mostly driven by anthropogenic forcing and cannot be explained by natural factors alone. These results reinforce the premise of a severe long-term drought in the subtropics of eastern South America that will likely be further exacerbated in the future given its apparent connection to increased greenhouse gas emissions.

for station locations) and regional precipitation from the Global Precipitation Climatology Centre (GPCC) dataset at 1 o resolution (17 o -13 o S -46.5 o -42.5 o W).
Figure S2 -Changepoint detection analysis using the algorithm 1 .Top panel: mean regional streamflow; middle panel: distribution of the z-score values of the speleothem proxies; bottom panel: change point detection analysis based on linear trend between proxy and depth ratio from Onça 2 speleothem covering the last 100 year from 1916 to 2016.Lines represent δ 18 O (red); δ 13 C (light blue); Mg/Ca (blue); Ba/Ca (yellow); Sr/Ca (purple); U/Ca (orange); deposition rate (green).Two change points in 1942 and 1994 are evident considering all the time series shown.S1) and PHYDA.Figure S5 -Comparison between cave atmosphere relative humidity and temperature with the outside environment: a) relative humidity measured inside (blue) and outside the cave (black) and potential evaporation (red, inverted scale).; b) Onça Cave daily mean temperature (black) and the outside environment daily mean temperature.The exterior measurements were obtained from the local meteorological station from the Instituto Nacional de Meteorologia (INMET -OMM: 88336) at Januária City, located 40 km southwest of the cave.The potential evaporation was measured using a Piche evaporimeter.Note that scaling of relative humidity and temperature differs between left and right yaxis.

Figure S1 -
Figure S1 -Comparison between precipitation record from Instituto Nacional de Meteorologia (INMET) and Agência Nacional de Água (ANA) meteorological stations covering an area of ~ 4 o × 4 o , centered over the study site (see TableS1for station locations) and regional precipitation from the Global Precipitation Climatology Centre (GPCC) dataset at 1 o resolution (17 o -13 o S -46.5 o -42.5 o W).

Figure S4 -
Figure S4 -Map of the Onça Cave, indicating the location of the speleothem sampling area.

Figure
FigureS8-Comparison between seasonal calcite Mg/Ca from monitoring experiment performed at Lapa da Onça and the cave relative humidity.Left: Time series of average monthly calcite Mg/Ca from Lapa da Onça cave experiment performed between June 2018 and October 2019 (red).The Mg/Ca of the dripping sites are presented as z-scored values.Right: comparison between the monthly mean Mg/Ca measured in the calcite deposition experiment (presented in the left panel) with cave relative humidity (blue lines), temperature (orange) and calcite deposition rate (black).Note that scales for mean calcite deposition, cave temperature and relative humidity are inverted.

Figure S10 -
Figure S10 -Comparison between the detrended Mg/Ca (z-score) data and local rainfall time-series (zscore).The times series were detrended using a linear regression between 1908 to 2016.Note that the Mg/Ca scale is inverted.

Figure S11 -
Figure S11 -Wavelet analysis performed with Onça2-4 δ 18 O.Black lines indicate the 95% significance level and the cone of influence (region over which record length is sufficient to interpret results).

Figure S13 -
Figure S13 -Regression lines with their corresponding scaling factors (β-coefficient) from Mg/Ca vs. PET ensemble medians for NAT, GHG+NAT and GHG experiment.The whisker plot shows the β-coefficient obtained from 20 Monte Carlo simulations estimated based on the regression between proxy and P-PET from each ensemble member.

Figure S14 -
Figure S14 -Regression lines with their corresponding scaling factors (β-coefficient) from δ 18 O vs P-PET ensemble medians for NAT, GHG+NAT and GHG experiment.The whisker plot shows the β-coefficient obtained from 20 Monte Carlo simulations estimated based on the regression between proxy and P-PET from each ensemble member.

Table S2 -U-Th ages 230
Th dating results.The listed error is 2σ.