Machine learning (ML), as an artificial intelligence tool, has acquired significant progress in data-driven research in Earth sciences. Land Surface Models (LSMs) are important components of the climate models, which help to capture the water, energy, and momentum exchange between the land surface and the atmosphere, providing lower boundary conditions to the atmospheric models. The objectives of this review paper are to highlight the areas of improvement in land modeling using ML and discuss the crucial ML techniques in detail. Literature searches were conducted using the relevant key words to obtain an extensive list of articles. The bibliographic lists of these articles were also considered. To date, ML-based techniques have been able to upgrade the performance of LSMs and reduce uncertainties by improving evapotranspiration and heat fluxes estimation, parameter optimization, better crop yield prediction, and model benchmarking. Widely used ML techniques used for these purposes include Artificial Neural Networks and Random Forests. We conclude that further improvements in land modeling are possible in terms of high-resolution data preparation, parameter calibration, uncertainty reduction, efficient model performance, and data assimilation using ML. In addition to the traditional techniques, convolutional neural networks, long short-term memory, and other deep learning methods can be implemented.

Some of the most intense convective storms on Earth initiate near the Sierras de Córdoba mountain range in Argentina. The goal of the RELAMPAGO field campaign was to observe these intense convective storms and their associated impacts. The intense observation period (IOP) occurred during November–December 2018. The two goals of the hydrometeorological component of RELAMPAGO IOP were 1) to perform hydrological streamflow and meteorological observations in previously ungauged basins and 2) to build a hydrometeorological modeling system for hindcast and forecast applications. During the IOP, our team was able to construct the stage–discharge curves in three basins, as hydrological instrumentation and personnel were successfully deployed based on RELAMPAGO weather forecasts. We found that the flood response time in these river locations is typically between 5 and 6 h from the peak of the rain event. The satellite-observed rainfall product IMERG-Final showed a better representation of rain gauge–estimated precipitation, while IMERG-Early and IMERG-Late had significant positive bias. The modeling component focuses on the 48-h simulation of an extreme hydrometeorological event that occurred on 27 November 2018. Using the Weather Research and Forecasting (WRF) atmospheric model and its hydrologic component WRF-Hydro as an uncoupled hydrologic model, we developed a system for hindcast, deterministic forecast, and a 60-member ensemble forecast initialized with regional-scale atmospheric data assimilation. Critically, our results highlight that streamflow simulations using the ensemble forecasting with data assimilation provide realistic flash flood forecast in terms of timing and magnitude of the peak. Our findings from this work are being used by the water managers in the region.

Sub-seasonal to seasonal (S2S) forecasts are critical for planning and management decisions in multiple sectors. This study shows results from dynamical downscaling using a regional climate model at a convection-permitting scale driven by boundary conditions from the global reanalysis of the Climate Forecast System Model (CFSR). Convection-permitting modeling (CPM) enhances the representation of regional climate by better resolving the regional forcings and processes, associated with topography and land cover, in response to variability in the large-scale atmospheric circulation. We performed dynamically downscaled simulations with the Weather Research and Forecasting (WRF) model over the Upper and Lower Colorado basin at 12 km and 3 km grid spacing from 2000 to 2010 to investigate the potential of dynamical downscaling to improved the modeled representation of precipitation the Southwestern United States. Employing a convection-permitting nested domain of 3 km resolution significantly reduces the bias in mean (∼2 mm/day) and extreme (∼4 mm/day) summer precipitation when compared to coarser domain of 12 km resolution and coarse resolution CFSR products. The convection-permitting modeling product also better represents eastward propagation of organized convection due to mesoscale convective systems at a sub-daily scale, which largely account for extreme summer rainfall during the North American monsoon. In the cool season both coarse and high-resolution simulations perform well with limited bias of ∼1 mm/day for the mean and ∼2 mm/day for the extreme precipitation. Significant correlation was found (∼0.85 for summer and ∼0.65 for winter) for both coarse and high-resolution model with observed regionally and seasonally averaged precipitation. Our findings suggest that the use of CPM is necessary in a dynamical modeling system for S2S prediction in this region, especially during the warm season when precipitation is mostly convectively driven.

Since the 1970s, land cover in central Argentina has shifted away from perennial crops and grasses toward annual crops, largely soy. In this study, we use observations and modeling to understand how this shift in land use has affected the sub‐surface, surface, and atmospheric fluxes of moisture and energy in a flat agricultural area. We analyze the flux tower data from a paired site at Marcos Juarez in central Argentina during the period of the RELAMPAGO field campaign (2018–2019). When compared to perennial alfalfa, the observations over soy show lower evapotranspiration (ET) and specific humidity, higher sensible heat, higher outgoing shortwave radiation, and soil temperature. Water table (WT) depth is shallower below the soy than the alfalfa sites. To better understand the long‐term temporal behavior from 1970s to present, the Budyko framework is used to show that the trends in ET cannot be explained by climate variables alone. We then use the Noah‐MP land surface model calibrated at both soy and alfalfa sites. Long‐term simulations of the calibrated model suggests that ∼95% of precipitation is evaporated in the alfalfa site with negligible recharge and runoff. Contrarily in the case of soy, ET is about 68% of precipitation, leaving nearly 28% for recharge and 4% for runoff. Observed increases in streamflow and decreases in WT depth over time are likely linked to shifts in land cover. Furthermore, the partitioning of net radiation shifts from latent heat to sensible heat resulting in a 250% increase in Bowen ratio (from 0.2 to 0.7).

Processes mediating the coupling of terrestrial, aquatic, biospheric, and atmospheric systems influence weather, climate, and ecosystem dynamics via transfer of energy, momentum, water, and carbon (or other species). These exchange processes are quantified by measurements of near-surface turbulent fluxes. Understanding processes at these interfaces provides insight toward understanding and predicting current and future states within the Earth system. The Atmospheric Radiation Measurement (ARM) user facility has been conducting measurements of near-surface turbulent fluxes since the early 1990s at long-term fixed locations and shorter-term mobile deployments across the Earth. ARM has utilized two established methods for conducting these measurements: energy balance Bowen ratio (EBBR) and eddy covariance (EC). Primary measurements from the former include sensible and latent heat flux, while the latter also measures fluxes of momentum and carbon (primarily carbon dioxide, with methane fluxes measured at two locations to date). The EBBR systems have been deployed at 22 locations, and, to date, the EC systems have been deployed at over 50 sites, with plans for additional novel site locations in the future. Herein, the history, evolution, and key aspects of these instrument systems are documented, along with information on data quality assurance and post-processing, as well as best use practices. Additionally, three data validation experiments were recently conducted, and their key findings are summarized. Finally, ancillary datasets acquired by ARM, which can contextualize and aid interpretation of the near-surface turbulent flux measurements, are discussed. The datasets described herein include the eddy correlation flux measurement system: 30ECOR (https://doi.org/10.5439/1879993, Sullivan et al., 1997), 30QCECOR (https://doi.org/10.5439/1097546, Gaustad, 2003), ECORSF (https://doi.org/10.5439/1494128, Sullivan et al., 2019a), and associated AmeriFlux and Methane Value-Added Product, AMCMETHANE (https://doi.org/10.5439/1508268, Billesbach, 2011); the energy balance Bowen ratio system: 30EBBR (https://doi.org/10.5439/1023895, Sullivan et al., 1993) and 30BAEBBR (https://doi.org/10.5439/1027268, Gaustad and Xie, 1993); and the carbon dioxide flux measurement system: CO2FLX (https://doi.org/10.5439/1287574, https://doi.org/10.5439/1287575, https://doi.org/10.5439/1287576, Koontz et al., 2015a, b, c; https://doi.org/10.5439/1989774, https://doi.org/10.5439/1989776, https://doi.org/10.5439/1992202, Biraud and Chan, 2002a, b, c). These data can be found by searching the above data stream names at https://adc.arm.gov/discovery/#/results/ (last access: 8 September 2025).