Evidence of human influence on Northern Hemisphere snow loss

Seasonal snow is a crucial indicator of climate change, with warmer temperatures favoring rain over snow and increasing snowmelt. This leads to reduced snow water storage and significant hydrologic risks. However, despite consistent warming trends at various scales, observational data on snowpack loss is inconsistent. While the IPCC acknowledges a general decline in Northern Hemisphere springtime snow water equivalent (SWE) since 1981, the extent of anthropogenic influence remains unclear, particularly at decision-relevant scales. This lack of robust attribution hinders identifying vulnerable regions and developing effective water security strategies. Several factors contribute to the inconsistent response of snowpacks to warming: observational uncertainties in SWE estimates (agreement on long-term snow change is found in only one-third of major river basins); high snowpack variability influenced by climate oscillations (Pacific Decadal Oscillation, Atlantic Multidecadal Variability); and the complex, non-unidirectional relationship between forcing and snowpack (warming can enhance precipitation and snowfall extremes, potentially offsetting losses). Existing attribution studies, often relying on limited climate models and realizations, may conflate internal variability and model uncertainties, further complicating the attribution of human-forced snowpack changes. This study aims to address these uncertainties by combining observational data with empirical and climate models to attribute snowpack changes to anthropogenic warming at hemispheric and river-basin scales, assessing the impact of temperature and precipitation changes on snow water storage, and generalizing snowpack and runoff responses to future warming.

Previous research has highlighted the potential impacts of warming climates on water availability in snow-dominated regions (Barnett et al., 2005; Immerzeel et al., 2020; Mankin et al., 2015; Qin et al., 2020). Studies have attempted to attribute declining snowpack to human influence, focusing on specific regions (Barnett et al., 2008; Pierce et al., 2008; Najafi et al., 2017; Jeong et al., 2017). However, these studies often faced challenges due to observational uncertainties and the complexity of disentangling natural variability from anthropogenic effects (Gottlieb & Mankin, 2022; Mortimer et al., 2020). The IPCC's assessment of snow trends also reflects the existing uncertainties in the field (Fox-Kemper et al., 2021). Other relevant research explored the influence of temperature and precipitation variability on snow trends (Mankin & Diffenbaugh, 2015), the value of large ensembles for adaptation decision-making (Mankin et al., 2020), and the role of climate models in capturing regional climate dynamics (Lehner, 2020; Mudryk et al., 2020; Kouki et al., 2022). Studies also investigated human influence on winter precipitation (Guo et al., 2019) and the response of snow-dependent hydrologic extremes to global warming (Diffenbaugh et al., 2013). The limitations of previous attribution studies due to model uncertainties, internal climate variability, and the complex relationship between precipitation and temperature changes are discussed, leading to the development of a novel approach in this study.

This study employs a data-model fusion approach combining observations-based ensembles of snowpack, temperature, precipitation, and runoff data with empirical and climate models. The approach addresses uncertainties by utilizing multiple datasets and integrating empirical models with climate model simulations to attribute snowpack changes to anthropogenic warming at different scales. The research leverages an ensemble of five long-term gridded SWE products (ERA5-Land, JRA-55, MERRA-2, Snow-CCI, and TerraClimate), along with in-situ observations, to capture the spatial pattern of historical trends in March SWE (1981-2020). The study utilizes Coupled Model Intercomparison Project Phase 6 (CMIP6) historical simulations (HIST) with and without anthropogenic emissions (HIST-NAT) to assess the influence of human-caused climate change. A pattern correlation analysis is applied to compare observed trends with those from the CMIP6 simulations to quantify the probability that the observed pattern is attributable to human influence. A random forest machine-learning algorithm is used to construct basin-scale SWE reconstructions, combining empirical models with climate model simulations. This accounts for observational uncertainty, isolates the impact of temperature and precipitation changes, and identifies forced snowpack changes above the noise level. The ensemble approach encompasses various combinations of observations and climate models to capture and quantify uncertainties. The counterfactual, no-anthropogenic-climate-change snowpack is estimated by removing the forced response of temperature and precipitation from the observed time series. Bias in CMIP6 models regarding historical warming trends and precipitation changes is analyzed in relation to model internal variability. Finally, the relationship between climatological winter temperatures and the marginal sensitivity of snowpack to warming is analyzed to reveal the nonlinear response of snow to temperature increase.

Despite substantial uncertainty in spatially distributed snowpack estimates, gridded snow products share a spatial pattern of historical trends consistent with in-situ observations. Over the past 40 years, March SWE has declined sharply in the southwestern USA and western/central/northern Europe (10-20% per decade). Conversely, central North America and northern Eurasia have seen increasing spring snowpacks. Coupled climate model simulations forced with historical human and natural forcing capture some features of the observed spatial pattern, particularly the large snow loss over Europe and modest gains over northern Eurasia. Simulations excluding anthropogenic emissions fail to reproduce the observed pattern. A pattern correlation analysis revealed that it is virtually certain (>99% probability) that human emissions have contributed to the observed pattern of March snowpack trends in most observational datasets. At the river-basin scale, empirical reconstructions show spring snowpack declines in many mid-latitude basins and modest increases in cold, high-latitude basins. The largest decreases (around 10% per decade) were observed in the southwestern USA and Europe. Consistent directional trends are seen in about half of all major river basins. Human-forced changes in temperature and precipitation have altered spring snowpack trends in 31 major river basins. Anthropogenic climate change reduced spring snowpacks in mid-latitudes (south of 60°N) by 4.1 ± 3.4% per decade and enhanced them in cold, high-latitude basins by 2.5 ± 1.8% per decade. The analysis revealed a highly nonlinear sensitivity of snowpack to warming. Below -8°C, snowpack is little affected by warming, but above this threshold, each additional 1°C of warming results in accelerating losses. This nonlinearity explains the lack of clear snow trends at the hemispheric scale despite substantial warming. The majority of the Northern Hemisphere's snow mass is found in regions less sensitive to warming, while most of the population resides in regions where even modest warming can cause significant snow loss. Under SSP2-4.5, highly populated basins are projected to experience strong declines in spring runoff due to nonlinear snow loss. The Western USA and Europe are particularly vulnerable.

The findings address the research question by demonstrating a clear human influence on Northern Hemisphere snow loss, overcoming previous limitations in attribution studies. The highly nonlinear temperature sensitivity of snowpack explains the seemingly paradoxical lack of widespread snow loss despite warming and highlights the potential for rapid future declines. The results demonstrate that natural variability has masked significant anthropogenic impacts in some regions, emphasizing the need for sophisticated methods to detect and attribute forced changes. The study's significance lies in its improved understanding of the complex interplay between climate change, snowpack dynamics, and water security at multiple scales. The identification of a nonlinear temperature sensitivity of snowpack is particularly important as it highlights the potential for sudden and widespread changes in water resources in the near future. The combination of multiple data sources and innovative analytical approaches strengthens the confidence in the results, providing a more robust scientific basis for decision-making related to water resource management.

This study provides compelling evidence of human influence on Northern Hemisphere snow loss, explaining the elusive nature of observed trends and forecasting significant future declines. The discovery of a nonlinear temperature sensitivity of snowpack highlights the imminent risk to water security in populated regions. Further research should focus on improving climate model skill in capturing regional climate processes to reduce uncertainties in projecting future snowpack changes. Refining observational estimates of SWE, particularly in data-sparse regions, is also crucial. These findings underscore the urgent need for adaptation and mitigation strategies to manage the water security risks associated with accelerating snow loss.

The study acknowledges that the CMIP6 models may overestimate historical warming in some regions and underestimate multidecadal drying/wetting trends. However, the biases are generally within the range of model internal variability, suggesting that they don't significantly undermine the attribution findings. The nonlinear snowpack response to warming is highly dependent on the average winter temperatures, which vary spatially due to factors like topography and proximity to moisture sources. The analysis focuses primarily on March SWE, and results may vary for other months. The reliance on multiple datasets introduces challenges and complexities in integrating the various data sources, potentially influencing the results. The study focuses on the Northern Hemisphere, and findings may not be directly generalizable to the Southern Hemisphere.