Arctic land is characterized by a high surface and subsurface heterogeneity on different scales. However, the effects of land surface model resolution on fluxes and soil state variables in the Arctic have never been systematically studied, even though smaller scale heterogeneities are resolved in high-resolution land boundary condition datasets. Here, we compare 210 km and 5 km setups of the land surface model JSBACH3 for an idealized case study in eastern Siberia to investigate the effects of high versus low-resolution land boundary conditions on simulating the interactions of soil physics, hydrology and vegetation. We show for the first time that there are differences in the spatial averages of the simulated fluxes and soil state variables between resolution setups. Most differences are small in the summer mean, but larger within individual months. Heterogeneous soil properties induce large parts of the differences while vegetation characteristics play a minor role. Active layer depth shows a statistically significant increase of +20% in the 5 km setup relative to the 210 km setup for the summer mean and +43% for August. The differences are due to the nonlinear vertical discretization of the soil column amplifying the impact of the heterogeneous distributions of soil organic matter content and supercooled water. Resolution-induced differences in evaporation fluxes amount to +43% in July and are statistically significant. Our results show that spatial resolution significantly affects model outcomes due to nonlinear processes in heterogenous land surfaces. This suggests that resolution needs to be accounted in simulations of land surface models in the Arctic.

Although detailed spatial and temporal distribution of soil moisture is crucial for numerous applications, current global soil moisture products generally have low spatial resolutions (25-50 km), which largely limit their application at local scales. In this study, we developed a high-resolution soil moisture retrieval framework based on ensemble learning by integrating Landsat 8 optical and thermal observations with multi-source datasets, including in-situ measurements from 1,154 stations in the International Soil Moisture Network, the Soil Moisture Active Passive (SMAP) soil moisture product, the ERA5-Land reanalysis dataset, and auxiliary datasets (terrain, soil texture, and precipitation). Two widely used ensemble learning models were explored and compared using ten-fold cross-validation. The extreme gradient boosting (XGBoost) model performed slightly better than the random forest (RF) model, with a root mean square error (RMSE) of 0.047 m(3)/m(3) and correlation coefficient (R) of 0.952, respectively. Further validation using data from four independent soil moisture networks demonstrated that the prediction accuracy of the XGBoost model was comparable to the SMAP soil moisture product, but with a much higher spatial resolution. The model was finally used to map soil moisture over the high-altitude Tibetan Plateau, which is especially sensitive to climate change, from May to September of 2015. The comparison between our fine-scale soil moisture map at 30 m resolution and the coarse-scale SMAP soil moisture product (36 km) revealed high spatial consistency. These results suggest that there is potential to generate accurate soil moisture products globally at 30 m spatial resolution from the long-term Landsat archive. This finding has practical implications in scenarios requiring fine-scale soil moisture maps, such as climate change and permafrost modeling, hydrological and land surface modeling, and agriculture monitoring.

Climate warming is causing permafrost thaw and there is an urgent need to understand the spatial distribution of permafrost and its potential changes with climate. This study developed a long-term (1901-2100), 1-km resolution daily meteorological dataset (Met1km) for modeling and mapping permafrost at high spatial resolutions in Canada. Met1km includes eight climate variables (daily minimum, maximum, and mean air temperatures, precipitation, vapor pressure, wind speed, solar radiation, and downward longwave radiation) and is suitable to drive process-based permafrost and other land-surface models. Met1km was developed based on four coarser gridded meteorological datasets for the historical period. Future values were developed using the output of a new Canadian regional climate model under medium-low and high emission scenarios. These datasets were downscaled to 1-km resolution using the re-baselining method based on the WorldClim2 dataset as spatial templates. We assessed Met1km by comparing it to climate station observations across Canada and a gridded monthly anomaly time-series dataset. The accuracy of Met1km is similar to or better than the four coarser gridded datasets. The errors in long-term averages and average seasonal patterns are small. The error occurs mainly in day-to-day fluctuations, thus the error decreases significantly when averaged over 5 to 10 days. Met1km, as a data generating system, is relatively small in data volume, flexible to use, and easy to update when new or improved source datasets are available. The method can also be used to generate similar datasets for other regions, even for the entire global landmass.