Abstract Soil moisture (SM) retrievals from the NASA Cyclone Global Navigation Satellite System (CYGNSS) are assimilated into the NASA Goddard Earth Observing System (GEOS) land data assimilation system from August 2018 to June 2024. The system uses a 24-member ensemble Kalman filter to integrate CYGNSS 6-hourly SM retrievals alongside, or in place of, established datasets from the Advanced Scatterometer (ASCAT), the Soil Moisture Active Passive (SMAP) mission, and the Soil Moisture Ocean Salinity (SMOS) mission. Assimilating CYGNSS alone (CYG_DA) reduces the standard deviation of CYGNSS observation-minus-forecast (OmF) residuals by about 6% relative to a model-only control simulation (CNTL). OmF residuals for monitored-only (that is, not assimilated) observations (ASCAT, SMAP, and SMOS) are also smaller in CYG_DA, indicating that assimilating CYGNSS improves the model background more broadly. When all four sensors are assimilated jointly (ALL_DA), OmF standard deviations are reduced by up to 11%, indicating enhanced cross-sensor consistency. Validation against independent in situ measurements confirms that CYG_DA improves surface-layer SM correlations and reduces unbiased RMSE relative to CNTL, with the largest skill gains seen in the multi-sensor ALL_DA experiment. A quasi-global triple-collocation analysis of CYG_DA using SMOS and ASCAT SM products shows widespread skill increases (median ΔR 2 ≈ 0.04) and reductions in error variance (mean ≈ −22%) relative to CNTL, particularly across semi-arid and seasonally vegetated regions. These results demonstrate that CYGNSS provides complementary SM information that enhances the temporal realism and stability of GEOS land-surface analyses. CYGNSS SM should thus be included in future multi-sensor reanalysis systems.
Abstract The Z – R relationships are widely used in radar meteorology and hydrology, as rainfall estimation plays a key role in understanding hydrometeorological processes and supporting operational applications such as flood forecasting. Although many studies determine coefficients for these Z – R relationships using regression-based methodologies with direct reflectivity observations and rainfall intensity measurements, the determination process is often not described in detail. In particular, the choice of the dependent variable is frequently left implicit, despite its influence on the outcome. This study investigates how the choice of the dependent variable affects rainfall estimate performance using X-band radar data, rain gauges, and disdrometers. We conducted analyses over 30 events (22 convective and 8 stratiform) in 2023 over the Seveso–Olona–Lambro River basin and the Milan metropolitan area (northern Italy). The results show that treating rainfall intensity as the dependent variable in least squares regressions with logarithmic transformation consistently improves rainfall estimate performance, as assessed with multiple skill scores, including bias, mean absolute error, root-mean-square error, coefficient of determination, Nash–Sutcliffe efficiency, and ratio of valid pairs. We also compared the resulting relationships with commonly used standard Z – R relationships in terms of mean areal accumulated rainfall. The relationships determined using rainfall intensity as the dependent variable exhibit better performance, highlighting the relevance of properly determining coefficients in hydrological modeling. The findings underscore the importance of explicitly reporting regression setups and variable roles in Z – R coefficient determination studies. Such transparency enables the reassessment of standard operational coefficients, enhances rainfall estimate reliability, and improves reproducibility across instrumentation configurations and event types. Significance Statement We highlight the importance of properly selecting the dependent variable when determining the relationship between radar reflectivity and rainfall intensity using radar, disdrometer, and rain gauge data. This remains relevant as least squares regression is widely used in operational rainfall estimation. Our results show that treating rainfall intensity as the dependent variable significantly improves estimation performance, as it is the target variable inferred from indirect radar observations in real-world applications. These findings emphasize the need for transparency in calibration methodologies and support more reliable radar-based hydrological applications.
Abstract Monitoring continental runoff is important for the global water cycle. Here, we update downstream monthly streamflow for the world’s 936 largest rivers to 2024 using multiple data sources for quantifying variations and long-term changes in continental discharge from 1950 to 2024. The update added new data for 697 rivers and extended the record length for 512 rivers. Still, many large rivers in South and East Asia, Africa, and some in South America have short records, and stations with observations have decreased since 1978. Streamflow in the world’s major rivers exhibits large interannual-to-decadal variations that are correlated with basin-mean precipitation, with insignificant long-term trends since 1948 for most of the rivers. Out of the 200 rivers with the largest discharges, only 71 (35.5%) show statistically significant streamflow trends during 1948–2024, with 34 being positive and 37 being negative. Global continental discharge decreased slightly from 1949 to 1993 and then rebounded to levels slightly above the 1950–90 mean in recent years. The streamflow-inferred runoff trends during 1950–2024 show coherent spatial patterns that are broadly consistent with observed precipitation trends, both showing increases over the central and Northeast United States, northern Europe and most of Russia, Argentina and Uruguay, and northwestern Australia, but decreases over most of Africa, East and South Asia, eastern Australia, the western United States, western and eastern Canada, and parts of Brazil. Both internal climate variability (especially from the tropical Pacific) and externally forced changes [e.g., due to carbon dioxide (CO 2 ) increases] likely contributed to these regional trends in precipitation and runoff, with below (above)-normal land precipitation and runoff during warm (cold) phases of the tropical Pacific sea surface temperatures.
Abstract Understanding changes in extreme precipitation is critical for enhancing climate resilience, particularly in regions such as West Africa, where communities are highly vulnerable to hydroclimatic extremes. This study examines the spatial and temporal variations in extreme precipitation across West Africa over 1998–2024 using high-resolution satellite data from NASA’s Integrated Multi-satellitE Retrievals for GPM (IMERG) V07B Final Run product. Extreme precipitation is analyzed using daily accumulated precipitation indices, including absolute and threshold-based indices. The Theil–Sen slope estimator and the Mann–Kendall test are applied to identify statistically significant trends at three significance levels. To assess the robustness of the detected trends, results from IMERG are compared with those from the Global Precipitation Climatology Project (GPCP) version 3.3 climate data record over the same period. The comparison shows generally similar spatial patterns and directions of change between the two datasets across West Africa. Results reveal distinct regional patterns of change. The Guineo-Congolian and Guinean regions, particularly Nigeria’s eastern coast, show significant declines in annual precipitation, along with significant decreases in the frequency and intensity of extreme events. In contrast, the western Sudanian and Sahelian regions, notably southern Mali and Burkina Faso, exhibit increasing trends in both total precipitation and extreme precipitation indices, indicating a shift toward wetter conditions driven by more frequent and intense extreme events. Furthermore, the contribution of extreme precipitation to annual totals varies across climatic zones, with smaller contribution from extreme days at lower latitudes. These findings highlight the importance of regionally tailored climate adaptation strategies for effective flood risk management and sustainable water resources.
Abstract Canopy interception remains one of the most uncertain components of the hydrometeorological water balance, particularly in gridded land surface modeling products where interception is parameterized through canopy water storage and wet-canopy evaporation. This study benchmarks event-scale interception loss estimates from three widely used operational North American Land Data Assimilation System phase 2 (NLDAS-2) land surface models from the NLDAS-2 configuration [Mosaic, Noah, and Variable Infiltration Capacity (VIC) model] against direct observations from 22 forested sites in the National Ecological Observatory Network. NLDAS-2 interception estimates reflect the coupled influence of gridded meteorological forcing and land surface model parameterizations, with storm characteristics prescribed through the forcing. Using 1787 storm events, we evaluate model − observation differences in interception loss as a total depth ( I L ) and as a percent ( I % ) of precipitation amount, with and without restricting events to cases where modeled and observed rainfall totals agreed within ±20%. The R 2 between event-scale modeled I L and National Ecological Observatory Network (NEON) I L ranged from 0.05 to 0.24, and all three models show lower I L magnitude and variability across sites. The mean percent bias in event-scale I % relative to NEON ranges from −6% to −18%. Precipitation amount, wind speed, and potential evaporation emerged as the strongest predictors of I L errors, consistent with a dominant influence of meteorological forcing and wet-canopy evaporation parameterizations on event-to-event differences. Errors in I % were more strongly influenced by energy-related variables and showed greater variability. These results point to opportunities for improving gridded land surface modeling products, including reconsidering how wet-canopy evaporation depends on wind and evaporative demand variability, and improving canopy storage parameterizations. Significance Statement Accurately representing how forests intercept and evaporate rainfall is important for predicting runoff, streamflow, and ecosystem water use. This study compares the prediction of interception loss from three land surface models to detailed rainfall and throughfall data collected from ecosystems across the United States. The results show that current models underestimate how much rainfall is held and evaporated by the forest canopy. By identifying which weather conditions contribute most to this bias, the findings point to new ways to improve interception model accuracy. Better representation of forest rainfall interception can help support water-resource forecasting and climate prediction in forested regions.
