Precipitation, the primary freshwater source, is crucial for water availability and effective adaptation and mitigation strategies against natural disasters. While temperature projections show agreement, precipitation projections exhibit significant regional uncertainties and lack coherence among climate models. This complexity stems from diverse factors within GCMs, such as their sensitivity to radiative forcing, warming rates, and aerosol radiative cooling. Sea surface temperature (SST) also plays a pivotal role, influencing regional water deficits and surpluses. Internal climate fluctuations, such as ENSO, IOD, PDO, NAM, SAM, and AMO, contribute to precipitation variability, and their interactions and teleconnections with precipitation do not align across GCMs, increasing projection heterogeneity. The increasing number of GCM ensembles expands the spread of precipitation's climate change signal and uncertainty, necessitating new approaches to reconcile the wide range of projections. Temporal aggregations are inadequate for heterogeneous variables like precipitation; excessive averaging obscures insights into significant changes. This study presents a novel approach analyzing trends in continuous, long-term time-series from multiple GCM ensembles, quantifying agreement in wetter or drier conditions and identifying hotspots with potential global human impacts. The study focuses on precipitation totals, defining global warming-induced drying or wetting as statistically significant and substantial continuous decreases or increases in precipitation capable of altering local regimes under intermediate and high emission scenarios. By using non-parametric trends and an ensemble of 146 CMIP5 and CMIP6 climate model runs, the study identifies regions where wetting and drying patterns converge globally. The trend-based approach aligns with the continuous nature of radiative forcing, providing flexibility and robustness in detecting and quantifying global warming-induced changes and effectively controlling for natural variability. This innovative approach evaluates the entire time-series without interannual averaging, combining information from the fullest range of GCM projections to determine their agreement and the extent of precipitation impacts without ensemble aggregation. Each set of simulations provides a plausible storyline of future precipitation patterns, and this approach offers an impact-based framework with country-scale analysis of drying and wetting agreement, the magnitude of change, and the exposed population, informing climate adaptation policies.

Existing literature highlights the challenges in projecting future precipitation patterns due to inherent uncertainties and inconsistencies across various Global Climate Models (GCMs). Studies such as those by Milly et al. (2002) and Hirabayashi et al. (2013) have addressed the increasing risk of floods and the global flood risk under climate change, respectively, showcasing the importance of accurate precipitation projections. However, the lack of coherence in precipitation projections across models, as noted by McSweeney and Jones (2013), hampers the development of effective adaptation strategies. Previous research has explored various aspects of precipitation variability and change, including the influence of sea surface temperature (SST) patterns (Good et al., 2021) and the roles of atmospheric circulation (Bony et al., 2013; Liu et al., 2013), but a comprehensive approach integrating these factors with a large ensemble of GCMs and a focus on long-term trends was lacking. Studies like those by Cook et al. (2014) and Dai (2013) examined drying trends and the increase in drought under global warming but often relied on aggregated data or shorter timeframes. The current study builds upon this existing literature by employing a novel methodology that addresses the limitations of previous approaches, providing a more robust and comprehensive assessment of future precipitation changes and their impact on the global population.

This study analyzed 146 GCMs (67 for intermediate and 79 for very high GHG emission scenarios) from the CMIP5 and CMIP6 archives. Data included 120-year (1980-2099) seasonal time-series of precipitation. The Mann-Kendall test and Theil-Sen Slope were used to detect statistically significant (p<0.05) trends and assess their magnitude. These non-parametric methods capture long-term monotonic trends without sensitivity to start/end points or outliers. The long time series helps constrain the influence of natural variability. A new metric quantifies drying/wetting agreement across GCMs. It assesses: (i) statistically significant trends (p<0.05); (ii) trend direction (slope); and (iii) whether the cumulative 120-year trend shifted by at least 10% of the local mean regime. The number of models meeting these criteria determines multi-model agreement at the grid-cell scale. Two agreement thresholds were used: 50% (majority) and 66% (2/3). No ensemble aggregation was used; individual runs were assessed, summarized as agreement. The analysis was repeated for CMIP5, CMIP6, and the full ensemble, across seasons and annually, for both emission scenarios. Seasonal dominance of annual wetting/drying was determined by identifying the season with the greatest trend magnitude in regions with >50% annual agreement. The population affected by global warming-induced wetting/drying was estimated using gridded datasets for current and future population projections, with 1km resolution. Agreement thresholds defined dry and wet masks, used to quantify the population within these regions on a grid-cell, country, and state-level basis. Two thresholds were used: 50% and 66%, with 50% considered more suitable given the approach's robustness and the large ensemble size. Country and state boundaries were obtained from the GADM database. Further details on the R code are available on Zenodo.

Analysis of 146 GCM runs revealed clear hotspots of wetting and drying agreement, highlighting regions where models agree on the direction, magnitude, and significance of long-term precipitation trends. At the country level, Greece, Spain, Palestine, Portugal, and Morocco showed substantial drying agreement (≥85% of models with robust decreasing trends), with potential rainfall reductions up to -21% (intermediate emissions) and -55% (very high emissions). Conversely, Finland, North Korea, Russia, and Canada exhibited high wetting agreement (≥90%), with some countries showing cumulative annual regime changes exceeding +35% (intermediate) and +48% (high) by 2100. Higher model agreement correlated with larger projected changes. Regions with no substantial annual agreement included central Europe, Southwest Asia, Australia, and parts of the African west coast and South America; some regions exhibited significant seasonal changes not apparent annually. Agreement was greater under very high emissions scenarios, with drying changes more pronounced than wetting changes from intermediate to very high emissions. The Caribbean and Mediterranean regions are most impacted by the expansion of the drying zone. For countries with heterogeneous spatial patterns (e.g., US, Brazil, Chile, Indonesia, South Africa), state-level regionalizations are recommended. Using a 50% agreement threshold, 38% of the current global population (3 billion) is projected to be affected by significant rainfall changes under intermediate emissions and 65.6% (5 billion) under very high emissions. This increases to 35.5% and 61.4% respectively when considering future population projections. A more conservative 66% threshold reduces these numbers but still reveals significant impacts. Many populated regions with substantial drying agreement (e.g., Mediterranean Europe, North Africa, Central America, Caribbean, southern South America, eastern Brazil, western Australia) already face water scarcity. Conversely, highly populated areas in Asia, northern Europe, north-western US, and central Africa show substantial wetting agreement; many experienced major floods and extreme precipitation recently. Seasonal analysis revealed a drying-to-wetting gradient in the US from MAM to JJA, similar gradients over Africa, and strong JJA and SON drying over southwestern Australia and northeastern Brazil. These patterns are consistent with previous studies but with more comprehensive analysis. No globally dominant season for wetting or drying was observed; however, seasonal dominance varied regionally (e.g., SON dominance for drying in southwestern Australia and the Indian Ocean, JJA for Iberian Peninsula, SON for wetting in China, India, and Central Africa, MAM in the United States).

This study provides a novel estimation of the global population affected by significant long-term precipitation changes due to global warming, offering an intermediate-to-high emissions envelope of projected impacts based on multi-model agreement. The findings of robust wetting and drying agreement across numerous regions highlight the spatial extent of projected precipitation changes. Several global-scale controls influence these changes, including Hadley Cell expansions, Walker circulation weakening, changes in the Intertropical Convergence Zone (ITCZ), land-ocean warming contrasts, poleward precipitation transport, and vegetation responses to increased CO2. Model deficiencies, such as ITCZ biases, might also influence drying patterns in specific regions. The study's significant contribution lies in its comprehensive analysis of GCM agreement on the direction and magnitude of long-term precipitation changes, providing an impact envelope for the global population under different emission scenarios. The regions identified as experiencing significant wetting or drying encompass a large fraction of the globe's landmass, underscoring the potential for widespread impacts. The study also addresses regions where model agreement is limited, highlighting areas of greater uncertainty. The detailed analysis incorporating both current and future population projections enhances the study's practical relevance and informs climate adaptation policies. Observational data supports the projected spatial patterns, showing consistency with observed increasing temperature and extreme events. Regional characteristics of changes in seasonal precipitation, such as longer dry spells, lower dry-season precipitation, and changes in wet-season precipitation, provide context for the study's findings. While this study focuses on changes in total precipitation amounts, changes in intensity, frequency, and duration of extreme events are also expected.

This study advances understanding of how increased GHG emissions affect precipitation regimes and populations globally, providing insights into the direction of precipitation changes under different emissions scenarios. The novel approach detects agreement in future wetting and drying trends across multiple models, offering robust quantification of change. The country-scale analysis and quantification of potentially exposed populations directly assist in designing climate adaptation policies. Future research should focus on investigating the influence of specific climate drivers on regional precipitation changes, further refining the understanding of model agreement and uncertainty across various climate scenarios and improving the ability to predict impacts on water resources and population vulnerability.

While this study utilizes a large ensemble of GCMs, the diversity of simulations is limited by the number of unique GCMs included. The analysis focuses on changes in total precipitation amounts, not considering changes in intensity, frequency, or duration of extreme events. Although the study accounts for some forms of natural variability, it cannot fully capture the range of possible fluctuations or long-term variations in climate patterns. Model biases and limitations in simulating regional precipitation patterns might affect the accuracy of the projections in certain regions. Lastly, the projections are based on the scenarios considered, and the actual emission pathway and resultant climate change might differ.