Search

Offline land surface models (LSMs) require atmospheric forcing data sets for simulating water, energy, and biogeochemical fluxes. However, available forcing data sets remain highly uncertain and can introduce additional differences in LSM simulations. This study explored the impact of various forcing data sets, ranging from widely used to newly developed, on hydrological simulations using the Common Land Model 2024 (CoLM2024). We conducted 12 global experiments using different forcing data sets to force CoLM2024. We evaluated the model's performance against plot‐scale observations and globally gridded reference data. We examined the uncertainties in forcings and their impact on output variables such as latent heat, sensible heat, net radiation, and total runoff. Globally, precipitation has the highest degree of uncertainty at 4.4%. The forcing uncertainties propagate to the model simulations and cause significant differences in simulated variables. Runoff uncertainty is about 15.7% globally, with a greater impact in low latitudes. Our evaluation shows that the newly developed data sets, such as CRUJRA and ERA5LAND, generally outperform the others. However, the optimal forcing data set varies depending on the variable of interest and the targeted region. Partial Least Squares Regression analysis reveals that different simulated variables are associated with dominant forcing variables, highlighting the importance of selecting forcing data sets for specific applications and regions. This study confirms the importance of improving the quality and consistency of meteorological data. This would help reduce simulation biases and guide the improvement of the model structure and parameterization of CoLM2024. Plain Language Summary: Atmospheric forcing data sets, including precipitation, air temperature, humidity, radiation, wind speed, and air pressure, are necessary for driving offline land surface models, for simulating water, energy, and carbon fluxes. However, atmospheric forcings deviate from each other and can lead to varying results in land surface model (LSM) simulations. Here we investigated to what extent the forcing data set can affect the simulation using the CoLM2024. To achieve this, we compared 12 different forcing data sets, from widely used to newly developed. Our results confirm that the uncertainties among these data sets, particularly in precipitation, can lead to significant differences in simulations such as runoff and surface fluxes. Our results show that the uncertainty in runoff is about 15.7% globally, with more pronounced in tropical regions. We find that using CRUJRA and ERA5LAND data sets improved CoLM2024 simulations compared to traditional ones. Our analysis reveals that different output variables are sensitive to different forcing variables, and the best data set varies depending on the region and the simulated variable. For example, runoff is mainly influenced by precipitation, while net radiation is more sensitive to shortwave radiation. This study underscores the need for improved meteorological data quality to enhance simulation accuracy, guiding CoLM2024 refinement and reducing biases. Key Points: 12 forcing data sets show large uncertainties and great effects on land surface model simulationsNewly data sets for CoLM2024 simulations generally outperformed others, but optimal choice variesPartial Least Squares Regression revealed the mechanism of uncertainty propagation [ABSTRACT FROM AUTHOR]

Estimates of the van Genuchten (1980, abbreviated as VG) parameters and saturated hydraulic conductivity (Ks) were made for the contiguous United States at a resolution of 100 m and seven soil depths by combining the SoilGrids+ (SG+) soil property maps of Ramcharan et al. with the R3H3 member of the Rosetta3 hierarchical pedotransfer functions (PTFs) of Zhang et al. To this end, we developed multi‐threaded code that significantly speeds up computation (up to a factor 25) depending on the level of parallelism. We verified estimates first by calculating simple summary statistics of estimated basic properties of SG+ with actual measured soil properties for 14,113 pedons in the National Cooperative Soil Survey (NCSS) (2023) labsample database. Next, we computed summary statistics of PTF‐estimated moisture contents for NCSS and SG+ data. The results show estimation errors are dominated by intrinsic errors of the PTF, and that (potentially correctable) systematic errors are present in SG+ soil properties and PTF estimates. The resulting hydraulic property maps contain well over 750 million points for each of the seven layers and show considerable horizontal and depth variation for each VG parameter and Ks, except the VG "n" parameter, which is dominated by values between 1.25 and 1.6. The hydraulic property maps are 99.9% complete, and we demonstrate that plausible profiles and uncertainty information can be generated for virtually each point. The maps are available as two multi‐channel GeoTIFF maps per SG+ layer: one with the five hydraulic parameters and one with the corresponding covariances. Core Ideas: van Genuchten parameters and saturated hydraulic conductivity were estimated for the contiguous United States and seven depths.Retention estimates were verified with National Cooperative Soil Survey soil data and analyzed for systematic errors.Parallelized code was developed to speedup future repeat analyses.Uncertainty in hydraulic characteristics can be expressed individually for every point.Data available as GeoTIFF files at 100‐m resolution, two files (parameters and covariances) for each depth. Plain Language Summary: Soil hydraulic properties are needed to calculate the infiltration of water into soils and to estimate flooding potential. Such properties can be measured in the laboratory which is usually an expensive and labor‐intensive task. Pedoytransfer functions are statistical methods that can estimate soil hydraulic properties from primary soil properties such as soil texture and bulk density. Recently, maps of primary properties were published for the contiguous United States at a resolution of 100 m. The present work applies artificial neural network pedotransfer functions to these maps and evaluates the accuracy of the estimates by soil depth and other criteria. The results are encouraging with some suggestions for future improvements. The developed code and estimated data are publicly available. [ABSTRACT FROM AUTHOR]

Hyper‐resolution land surface modeling provides an unprecedented opportunity to simulate locally relevant water and energy cycles. However, the available meteorological forcing data is often insufficient to fulfill the requirements of hyper‐resolution modeling. Here, we developed a comprehensive downscaling framework based on topography‐adjusted methods and automated machine learning (AutoML). With this framework, a 90 m and hourly atmospheric forcing data set was developed from ERA5 data at a 0.25° resolution, and the Common Land Model (CoLM) was then forced with the developed forcing data over two complex terrain regions (the Heihe River Basin and Upper Colorado River Basin). We systematically evaluated the downscaled forcing and the CoLM outputs against both in situ observations and gridded data. The ground‐based validation results suggested consistent improvements for all downscaled forcing variables with mean RMSE improved by 6.362%–95.86%. The downscaled forcings, which incorporated detailed topographic features, offered improved magnitude estimates, achieving a comparable level of performance to that of regional reanalysis forcing data. The downscaled forcing driving the CoLM model showed comparable or better skills in simulating water and energy fluxes, as verified by in situ validations. The hyper‐resolution simulations provided a detailed and more reasonable description of land surface processes and attained similar spatial patterns and magnitudes with high‐resolution land surface data, especially over highly elevated areas. Additionally, this study highlighted the benefits of using mountain radiation theory‐based shortwave radiation downscaling models and AutoML‐assisted precipitation downscaling models. These findings emphasized the significance of integrating topography‐based downscaling methods for hillslope‐scale simulations. Plain Language Summary: High‐resolution atmospheric forcing data is required for accurate hyper‐resolution land surface modeling. However, the currently available forcing data is not sufficient due to its coarse spatial resolution. In this study, we developed a downscaling framework that employs physical realism equations and machine learning to account for the impact of topography on atmospheric conditions. We tested the framework using the Heihe River Basin and Upper Colorado River Basin as examples and found that the downscaled forcing provided more realistic representations than those from the coarse‐resolution data. The downscaled forcing‐based hyper‐resolution simulation captured topographic heterogeneity and improved water and energy variable simulations. These findings emphasize the efficacy of the proposed topographic‐based downscaling framework and offer a valuable way to enhance predictions over regions with complex terrain. Key Points: A downscaling framework for atmospheric forcing based on topography‐adjusted methods and automated machine learning is developedThe downscaled forcing offers better estimates of atmospheric conditions in terms of accuracy, magnitude, and spatial detailed featuresThe CoLM driven by the downscaled forcing significantly improved the simulation of water and energy variables over complex terrain areas [ABSTRACT FROM AUTHOR]

Terrestrial evapotranspiration (ET) is a vital process regulating the terrestrial water balance. However, significant uncertainties persist in global ET estimates. Focusing on the area between 60°, we performed an intercomparison of 90 state‐of‐the‐art ET products from 1980 to 2014. These products were obtained from various sources or methods and were grouped into six categories: remote sensing, reanalysis, land surface models, climate models, machine learning methods, and ensemble estimates. It is shown that global ET magnitudes of categories differ considerably, with averages ranging from 518.4 to 706.3 mm yr−1. Spatial patterns are generally consistent but with significant divergence in tropical rainforests. Global trends are mildly positive or negative (−0.10 to 0.37 mm yr−2) depending on categories but with distinct spatial variability. Evaluation against site measurements reveals various performances across land cover types; the ideal point error values range from 0.45 to 0.83, with wetlands performing the worst and open shrublands the best. Using the three‐cornered hat method, there are spatial differences in ET uncertainty, with lower uncertainty for ensemble estimates, showing less than 15% relative uncertainty in most areas. The best global ET data set varies depending on the intended use and study region. Distinct spatial patterns of controlling factors across categories have been identified, with precipitation driving arid and semi‐arid regions and leaf area index dominating tropical regions. It is suggested to include advancing precipitation inputs, incorporate vegetation dynamics, and employ hybrid modeling in future ET estimates. Constraining estimates using complementary data and robust theoretical frameworks can enhance credibility in ET estimation. Plain Language Summary: Evapotranspiration (ET) is an essential process where water transfers from the land to the atmosphere through evaporation and plant transpiration. However, substantial uncertainties exist across available global ET estimates produced by various models and methods. To date, we created the most comprehensive compilation of over 90 state‐of‐the‐art ET data products, including satellite‐based, land surface model, climate model, reanalysis, machine learning, and ensemble‐based estimates, covering 1980–2014 and ranging between 60°N and 60°S. These estimates show large differences in global ET magnitude, trends, and spatial patterns. We evaluated ET estimates against in situ measurements and found diverging performance across land cover types. We used statistical methods to quantify uncertainties in ET data products and found discrepancies in uncertainty distributions across methods. We also identified key factors driving spatial ET variability in each product category. Our analysis underscores the importance and opportunities for reducing the uncertainty in terrestrial ET and provides valuable insights for improving hydroclimatic predictions. Key Points: Global ET estimates show large discrepancies in magnitude, trends and spatial patterns across 90 state‐of‐the‐art data setsThere is no single best global ET data set suitable for all applications and locations, the choice depends on the intention and region of studyDominant controls on ET differ across categories spatially; constraining estimates via Budyko and novel hybrid modeling can enhance confidence [ABSTRACT FROM AUTHOR]

Northeast China (NEC) is the most prominent grain‐producing region in China. However, it is currently facing significant impacts from climate change. Since the climate‐related impacts on crop yield in this region are a major concern for society in the future, quantifying climate change impacts on crop yields in NEC is essential to ensure future food security. This study aimed to quantify the effects of future climate change on crop yields in NEC and explore adaptation strategies using the Crop Growth Model (PCSE) driven by downscaled CMIP6 climate projections under four Shared Socioeconomic Pathways (SSPs) scenarios during 2015–2100. Results showed that there could be average reductions in crop yields of 21.4% for maize and 4.2% for soybean by the year 2100 under SSP585 compared to the 2015 baseline. The increasing temperature was the dominant factor in reducing yields, although elevated CO2 and precipitation offered partial compensation. The optimized planting date brought noticeable benefits for rice and soybean but had limited effects on maize due to heat stress. Relocating rice expansion eastward and implementing earlier planting increased yields by up to 50% but adversely decreased soybean and maize due to competition. This study enriches our comprehension of climate change impacts on NEC agriculture, while also quantifying potential benefits and constraints of evaluated adaptations. The proposed adaptations may help mitigate projected yield declines in other key agricultural regions across the globe. Adjusting crop management practices to capitalize on changing climate factors shows promise as a strategy for sustaining production globally. Plain Language Summary: Northeast China (NEC) is essential for achieving food accessibility and food security in China, but its agricultural production is seriously threatened by climate change. This study investigated how climate change could impact maize, rice, and soybean crop yields in this region from 2015 to 2100, using the PCSE crop model and forcing from the international Coupled Model Intercomparison Project 6 (CMIP6) climate projections. Simulation results suggested rising temperatures would negatively affect crop yields, especially for maize. Increased carbon dioxide and rainfall partially offset these losses. By 2100, average declines were projected for maize (−21%) and soybeans (−4%), respectively. Altering planting dates to match crop needs was an effective adaptation to boost yields, with rice benefiting the most. Relocating rice production eastward and expanding its area substantially increased simulated rice yields (+62%) but decreased other crops due to competition (−32% maize, −28% soybeans). Even with adaptations, some areas still showed yield declines, indicating extra measures may be needed to maintain production. Overall, this study suggests significant risks to crop yields from climate change in Northeast China, with adaptations like optimized planting and strategic crop relocation able to mitigate some but not all of the projected impacts. Key Points: Rising temperatures negatively impacted simulated future crop yields, especially for maizeAdjusting planting dates boosted yields for rice and soybeans but had limited benefits for maizeRelocating and expanding the rice area increased yields substantially [ABSTRACT FROM AUTHOR]