Subseasonal to seasonal (S2s) precipitation forecasts provide great potential for hydrological forecasting at an extended range. The study proposed a support vector machine (SVM) regression‐based method to improve S2s precipitation forecasts from the European Center for Medium‐Range Weather Forecasts (ECMWF) across the globe (60°N to 60°S). Results suggested that the SVM‐based method significantly improved ECMWF daily precipitation forecasts in representing the spatiotemporal variation of precipitation with higher consistency and reduced errors when compared against observations. Furthermore, the SVM‐based method enhanced the probabilistic skill of ECMWF forecasts, providing improved ranked probability skill score (RPSS) for real‐time forecasts in 2020 (e.g., RPSS_ECMWF = −0.03 and RPSS_rg3 = 0.08 for lead week 1). The most substantial improvement from the SVM‐based method is witnessed in regions with complex terrains where ECMWF yielded the worst skill, such as the Andes mountain range, Congo River Basin, and the Tibet Plateau. However, the SVM‐based post‐processing method did not alter the characteristics of precipitation forecasts regarding climate zone and lead time. Both ECMWF and post‐processed forecasts showed higher skill in the temperate and continental climate zones when the lead time is shorter than 2 weeks. In comparison, low‐latitude regions exhibited higher predictability when the lead time is longer than 5 weeks, which is attributed to the slow variation of boundary conditions such as the El Niño‐Southern Oscillation (ENSO).

Geostationary meteorological satellites provide precipitation estimates with a high spatio-temporal resolution, which is important for near real-time precipitation monitoring. This study systematically evaluated geostationary orbit (GEO) satellite-based quantitative precipitation estimates (QPEs) from Chinese Fengyun-2 G (FY-2 G), Fengyun-4A (FY-4A), and South Korean Geo-KOMPSAT-2A (GK-2A) across Northeast Asia in 2020. Compared against ground-based rainfall gauges at a 6-hourly scale, FY-2 G provided the highest accuracy in the China region with a high correlation coefficient (R = 0.53) and a low bias (−0.26 mm) due to the ground calibration process in FY-2 G. Conversely, GK-2A provided more accurate precipitation estimates for South Korea and Japan stations. FY-4A QPE generally showed a large positive bias throughout different seasons, although it provided satisfactory R and categorical statistics. FY-based QPEs slightly overestimated summer precipitation, especially over South Korea and Japan region, while GK-2A tended to underestimate summer precipitation. All examined QPEs showed poor accuracy during the winter season due to the frozen particles and ice clouds. Intensity analysis revealed that FY-based QPEs tended to overestimate the occurrence of no rain and heavy rain cases, whereas GK-2A underestimated no rain and heavy rain cases and overestimated light rain occurrence. It is also found that all examined QPEs captured the temporal variation of precipitation during storm events, while FY-based products overestimated heavy precipitation peaks and GK-2A underestimated peak precipitation. The findings in the study provided valuable information to further improve current infrared precipitation retrieval algorithms.

Spatiotemporal variability in Earth's terrestrial water storage (TWS) causes changes in surface deformation. The potential for using ground‐based Global Positioning System (GPS) vertical displacement observations for estimating TWS is explored through a comparison of vertical displacements derived from space‐based gravimetric retrievals, ground‐based GPS, and model‐based hydrologic estimates. The study presented here focuses on two snow‐dominated basins in the Western United States for the years 2003–2016. Seasonal variations are observed in the vertical displacements derived from all three data sets, and the variation is coherent with the changes in hydrologic loading. Good consistency is observed between any two of the three data sets with gravimetric retrievals and hydrologic model estimates providing the highest level of agreement (i.e., all examined stations with correlation coefficient R > 0.70). Vertical displacements derived from gravimetric retrievals and ground‐based GPS yielded R > 0.70 for more than 89% of the stations. In addition, it is found that both GPS‐derived and space‐based, gravimetry‐derived vertical displacements clearly reflected the impact of climate variation (i.e., heavy precipitation during 2010–2011 winter followed by prolonged drought). Vertical displacements derived from the hydrologic model highlighted the relatively large precipitation convergence phase during late 2010 to early 2011 at some stations but not the prolonged drought that followed. The results indicate that ground‐based GPS observations of vertical displacement have the capability to capture variations in TWS changes, which can be systematically merged in conjunction with Gravity Recovery and Climate Experiment (GRACE) into a land surface model to improve TWS estimates in a follow‐up study.

Ground‐based Global Positioning System (GPS) observations of vertical surface displacement can be used to study terrestrial water storage (TWS) change after properly accounting for nonhydrological loading effects. This study systematically merged ground‐based GPS observations of vertical displacement into a land surface model in order to better estimate TWS. Assimilation was conducted across two snow‐dominated watersheds in the western United States using a one‐dimensional ensemble Kalman filter (EnkF). Modeled estimates were compared against TWS retrievals derived from the Gravity Recovery and Climate Experiment (GRACE) mission and in situ measurements of snow water equivalent (SWE), soil moisture, and runoff. The GPS data assimilation (GPS DA) technique improves prediction skill of TWS anomalies relative to the Open Loop (OL; model run without assimilation) when compared against GRACE TWS retrievals, especially during an extended drought period post‐2011 (e.g., correlation coefficient R_OL = 0.46 and R_GPSDA = 0.82 in the Great Basin). Furthermore, GPS DA improves SWE estimation with improved R values found over 76% and 69% of all pixels collocated with in situ stations in the Great Basin and Upper Colorado watersheds, respectively. GPS DA estimates of surface soil moisture and runoff, however, exhibit degraded performance relative to the OL due to the limited sensitivity of GPS observations to surface soil moisture variations and the lack of a dynamic surface routing scheme as well as other missing model physics (e.g., river and dam regulation, irrigation‐related water withdrawals, surface water impoundments) that are not included in the Catchment Land Surface Model.

Empirical baseflow filters are widely used for baseflow separation. These filters rely on ad‐hoc parameters that introduce significant uncertainties in the calculation. A recent study by Mei et al. (2024, https://doi.org/10.1029/2023wr036386 ) optimized these parameters using environmental tracer data for 1,100 catchments across the Contiguous United States (CONUS). However, optimized parameters are unavailable for most CONUS catchments lacking tracer data. To address this gap, we developed regionalization models for the filter parameters using the random forest (RF) algorithm and 82 catchment‐scale characteristics, including geomorphology, climate, soil properties, and human activities. We demonstrated this approach for the block length parameter of the smooth minima baseflow filter, one of the optimized filters in Mei et al.’s study, across 855 catchments. Our results show that the prediction of achieves an of 0.80. Predictor importance analysis identified catchment area as the most influential factor, followed by climate, hydrology, soil, and water usage characteristics. Using the RF‐predicted in baseflow separation improves daily baseflow accuracy, with the median Kling‐Gupta Efficiency increasing from 0.62 to 0.80 compared to the literature‐suggested area‐based power function. This study enhances the accuracy of baseflow separation, providing a robust foundation for understanding streamflow partitioning and supporting improved hydrological modeling.