Assimilating Multi-Satellite Snow Data in Ungauged Eurasia Improves the Asian Monsoon Seasonal Forecasts

The Asian monsoon affects more than 60% of the world’s population, yet its skillful subseasonal to seasonal (S2S) forecast remains an unresolved issue for global climate models. Current major climate modeling centers rely primarily on the ocean conditions and ocean-based data assimilation for monsoon seasonal forecasts. As monsoons are formed from the land–ocean thermal contrast, neglecting the land conditions and land-based data assimilation hinders accurate forecasts.

Effectively assimilating land observations especially in ungauged regions such as the Tibetan Plateau and Siberia using satellites is expected to improve the forecast skill. However, it has not been demonstrated how much skill can be derived from assimilating these observations from space. The Tibetan Plateau (TP) and mid- to high-latitude Eurasia (EA) are known to be two critical regions for modulating the Asian monsoon dynamics. While land gauged-based data are limited in these regions, space-based snow mass and coverage, especially from contrasting but complementary satellites, have not yet been assimilated in the dynamical climate models, which significantly limits our understanding of how these data improve monsoon forecast skill.

In this study, we innovatively tackle the above problem using multi-satellite land assimilation data that have been developed in our group. We focus on two contrasting but complementary satellites, namely the Moderate Resolution Imaging Spectroradiometer (MODIS) satellite and the Gravity Recovery and Climate Experiment (GRACE) satellite. The former uses optical sensors, so it detects snow coverage but cannot infer snow mass when the coverage approaches 100%. The latter uses gravity changes to detect underlying total water storage changes, and hence snow mass changes but it must rely on a land surface model to single out the snow contribution. The availability of our newly developed global snow mass data from the two satellites allows us to improve the snow initialization of a dynamic model forecast. Focusing on the memory of snow over the TP and the EA, we have carefully quantified the role of snow data assimilation (DA) in the monsoon seasonal forecast using hundreds of ensemble experiments.

Our results have demonstrated snow DA as an important but underutilized source of the Asian monsoon S2S forecast skill. Among all Asian monsoon sub-regions, central north India exhibits the most robust and pronounced improvement, which has profound agricultural and economic implications. More importantly, we show that different satellite observations (e.g., MODIS vs. GRACE) and different Eurasian regions (e.g., Tibet vs. higher-latitude Eurasia) were effective to the forecast skill at different lead times. These findings can potentially guide the future choice of satellites and region-of-focus tailored for the Asian monsoon forecast.

Currently multi-satellite land DA is a hot research topic, but most of the studies have been focused on DA methodological developments while the understanding of their linkages with the S2S atmospheric forecast has been surprisingly lacking. To the best of our knowledge, this study is the first to systematically understand the monsoon forecast skill from the perspective of snow DA, and it can potentially bridge the current knowledge gap.