Research Project: Improving Resiliency of Semi-Arid Agroecosystems and Watersheds to Change and Disturbance through Data-Driven Research, AI, and Integrated Models

Location: Water Management and Systems Research

2025 Annual Report

Objectives
Objective 1: Improve biophysical and ecohydrologic components of crop and ecosystem services models ranging in spatial scale from sub-field areas to watersheds by linking process-based modeling, data assimilation, and artificial intelligence (AI). Sub-objective 1.A: Develop ecophysiological model components for croplands. Sub-objective 1.B: Enhance modeling of crop phenology, yield, and ET in semi-arid conditions at daily to seasonal and plot to watershed scales. Objective 2: Inform precision agriculture (water and nutrient management in crop systems) and precision conservation within fields, across farms, and at regional watershed scales, using high-resolution process modeling and machine learning. Sub-objective 2.A: Improve the Agricultural Ecosystems Services (AgES) process model using components from Objective 1; develop and test a subdaily version of AgES. Sub-objective 2.B: Apply AgES to simulate on-farm precision conservation; train surrogate models for users; publish long-term data and results. Objective 3: Quantify current and future impacts of climate variability, land-use change, land disturbance (e.g., wildfire, insect infestation), and rehabilitation on water resources from source-water catchments in snow-dominated agricultural watersheds. Sub-objective 3.A: Develop and implement snow-process model components and ecosystem × hydrometeorology interactions. Sub-objective 3.B: Develop geospatial methods to analyze and model hydrologic function and response to change, from “fire to farm”. Sub-objective 3.C: Predict the ecohydrological impact of precision-conservation treatments in source-area catchments. Objective 4. Develop management practices incorporating the latest technology developments for a field-size aspirational four-year dryland crop rotation system with precision nutrient, agrichemical, and weed control and crop population management. 211 C1 PS1C, C3 PS3A. Objective 5. Compare yields, economic returns, and environmental impacts of the aspirational dryland rotation system to the system that is currently used by producers of the region. 211 C1 PS1C, C3 PS3A.

Approach
Agricultural productivity and ecosystem services are inextricably linked to water resources that are facing dual pressures of decreasing supply and increasing demand. In the western US, water resources are predominantly derived from the melting of seasonal high-elevation snowpack where disturbance (e.g., wildfire, insect infestation) and hydrologic and ecosystem functioning can directly impact water availability for agricultural production, i.e., “fire to farm”. Additionally, shifting precipitation patterns and increasing air temperatures are resulting in smaller and earlier peak snowpack water equivalents and advancing the timing of snowmelt and peak streamflow. Subsequent impacts of these changes to moisture availability affect natural ecosystem functioning, plant ecophysiological responses, and vegetation contributions to water cycling, in turn affecting ecosystem services and downstream water availability and quality. This project aims to improve the understanding of ecohydrological processes in semi-arid western US agricultural watersheds by considering the continuum of water resources from streamflow generation in the mountains through ecosystem controls of water cycling to impacts of farm-level water limitations on crop growth and productivity. To address these needs, we will use a variety of data-assimilation tools, process-based models, and artificial intelligence (AI) to better characterize the soil-plant-atmosphere pathway across landscape types. Model components and improvements resulting from this research will inform precision agriculture and conservation across spatiotemporal scales and improve quantification of water supply responses to climate, disturbance, and management (Fig. 1). By focusing on agricultural watersheds, this project will develop holistic tools and build broad and spatially resolved datasets that will improve crop production under limited water, operational forecasting of water supplies and ecosystem services, precision conservation of water quality, and broader earth-systems research to inform land surface models.

Progress Report
Objective 1a: Extensive instrumentation has been installed and paired with remote sensing data at the Limited Irrigation Research Farm (LIRF), and at multiple sites within high-elevation source-water catchments to measure plant responses to climate variability and to parameterize ecophysiological models. Data collection at LIRF has focused on measuring plant transpiration using sapflow, remotely sensed vegetation indices, and micrometeorology. Instrumentation at source-water locations has focused on quantifying the transport and storage of water within tree stems, which has the potential to represent a substantial portion of total water within a watershed. These data will be analyzed and used to develop model components in the Ages hydrologic model. Objective 1b: The Ages model with the Unified Plant Growth Model (UPGM) crop component has been used to simulate soil-water dynamics and crop phenology in experimental plots near Greeley, Colorado. Data include replicated irrigation treatments at full evaporative demand and limited irrigation for corn, wheat, sorghum, dry bean, and sunflower. Simulated soil moisture was calibrated to fit measured data at multiple depths. The calibrated model for soil hydrology was then used to refine crop phenology parameters at multiple developmental stages. Objective 2a: New features were developed in Ages to allow simulation of hydrology at subdaily time steps (Ages 1.0 is a daily model). Users may specify model inputs based on available data to drive the model at any time step down to 1 minute intervals. The subdaily version of Ages also provides multi-temporal time steps by switching from daily when no runoff is detected to subdaily when precipitation occurs. These features have been tested on a 56 hectare agricultural watershed (wheat-fallow cropping). Results were reported at special session of the Soil Science Society of America annual meeting in San Antonio on November 12, 2024. This presentation illustrated the space-time dynamics of rainfall-runoff patterns using animated graphics where surface runoff accumulated going downslope to the outlet. In addition to these results, certain model parameters were scaled within the model based on the time step to guide future applications. Objective 2b: Previous simulations of streamflow and nitrate loads have been extended in time to address temporal changes in wastewater discharges to the Big Dry Creek in Colorado. Reduced nutrient concentrations of point sources due to improved water treatment are changing the relative importance of agricultural non-point sources, which makes simulation of precision management more impactful. The model is also being improved to account for direct pumping from the stream and groundwater. New methods of model calibration will fit statistical distributions of flow and load in comparison to previous calibrations to daily timeseries data. The new results were presented to the Big Dry Creek Watershed Association and will be submitted for publication in a peer-reviewed journal. Objective 3a: To improve understanding of spatial variability in snowpack processes, ten monitoring locations and sensor networks have been established across an elevation gradient (5000-11,200 ft) to monitor hydrometeorology and streamflow within the Cache la Poudre and Big Thompson river basins. These data have been analyzed paired with field measurements of snowpack water equivalent to parameterize and calibrate the Ages hydrologic model. In addition, 4 km modeled climate data (Weather Research and Forecasting model) and 30 m spatially explicit modeled snowpack data (SnowModel) have been used as calibration data for Ages in the Blue River Basin in central Colorado, identifying climatic factors driving snowpack dynamics and streamflow from high-elevation watersheds. In 2025, additional weather stations were added as well as instrumentation added to burned trees in wildfire areas to assess temperature differences and impacts on snow melt. Cosmic ray sensors were installed at two sites (persistent and transitional snowpack) to detect both snow water equivalent and soil moisture where other sensors have been installed, including multiple sensors for spatial variability of soil moisture on a steep burned slope. A new laser flow velocity sensor was installed downstream of the burned transitional snow site to improve estimation of rapidly changing stream flow rates. Objective 3b: Wildfire severity plays an important role in river basin functioning. To better understand factors driving wildfire burn severity, a uniquely large and comprehensive dataset has been developed of landscape, climate, and forest structure predictors associated with pixel-level burn severity for all wildfires within in the Southern Rockies ecoregion from 1990-2022. This dataset was analyzed using a multivariate machine learning model to determine which predictors have the greatest influence on burn severity, finding that forest structure (canopy height, canopy density) and climate (minimum daily air temperature, relative humidity) are key to predicting burn severity across various forest types within a large regional area. Objective 3b: To reduce risk from future wildfire, land management agencies commonly perform fuel reduction treatments within forests. To improve understanding of fuel treatment effects, ARS researchers compiled a vast dataset of 284 fuel reduction treatments paired with untreated controls completed > 10 years ago. These locations have been sampled for the recovery of fuel and vegetation structure before and after treatments and will be analyzed to assess climate and landscape factors driving fuel treatment effectiveness and longevity. Note, this is a departure from original objectives to establish hydrometeorological monitoring stations and future fuel treatment locations due to complexities in organizing instrumentation and treatment timing with land managers. Objective 4a: The data of four years from Aspirational (ASP) and Business-As-Usual (BAU) management scenarios were compiled. A manuscript regarding yield in relation to different management practices and weather patterns was submitted to the journal for peer review publication. A second publication regarding Ages model of soil water in ASP-BAU fields is under development. Data was presented for the scientific communities at the European Geosciences Union (EGU). At the meeting, ideas were exchanged regarding the site management, crop rotations, and nutrient management to cope with climate change. Objective 4b: The journal manuscript from Objective 4a included assessment of precision management zones in the Aspirational (ASP) management scenario, and results related to the precision management study were presented at an ARS Field Day event. Additional data collection (year 2) for the split-N precision management study was completed, and year 3 data collection is ongoing (need for year 4 will be assessed after year 3 is complete). Dryland precision management scenario fertilizer requirements (based on yield projection) are driven by water availability, so seasonal weather (timing and amount of in-season soil moisture as it relates to crop phenology) is a critical factor in the experimental design. Objective 4c: Year 1 of corn population study was completed, but Year 2 was postponed for various reasons. Objective 4e: Normalized difference vegetation index (NDVI) data were collected with an Unmanned Aerial Vehicle concurrent with soil moisture data along topographical field transects. Additional data are needed to develop a robust NDVI-yield relationship covering a range of in-season soil water availability conditions (as discussed in Objective 4b). Objective 5a: Data for both the ASP-BAU management scenarios as well as long-term Alternative Crop Rotation (ACR, 30-year) studies have been compiled for inclusion in the economic analyses. Analysis has been expanded to include data from all 196 ACR plots over the entire period of record. Preliminary data analyses, completed in collaboration with Colorado State University economists, were presented at two Customer Focus Group meetings and a field day. Objective 5b: Analyses of several soil chemical properties were included in the journal paper from Objective 4a. More-comprehensive analysis was partially accomplished and is ongoing.

Accomplishments
1. Targeted forest restoration treatments manage invasion risk of non-native species. ARS researchers in Fort Collins, Colorado, working with Colorado State University faculty, have improved precision management in forested source-water systems by providing long-term empirical evidence on how mechanical forest restoration treatments influence both forest structure and understory plant dynamics. Through a robust before/after control/impact design, the research demonstrates that restoration treatments effectively reduced tree density and increased native understory cover and species richness, aligning with ecological resilience goals. Importantly, it reveals that while these treatments support native biodiversity, they also slightly increase vulnerability to non-native species, with moisture deficit identified as a key driver of invasion risk. Identifying stand structure and topography as primary influencers of native plant responses offers valuable insight for spatially targeting restoration efforts. These findings enable land managers to more precisely weigh ecological trade-offs and tailor interventions that enhance native biodiversity while mitigating potential invasions under changing climatic conditions.

2. Improving post-fire recovery with biochar and mulch soil amendments. ARS researchers in Fort Collins, Colorado, in collaboration with the U.S. Forest Service and Colorado State University faculty, have provided land managers with powerful, long-term insights into how post-fire soil amendments such as biochar and wood mulch can influence forest recovery and resilience in high-severity burn areas. Across more than a decade, these treatments altered soil moisture, nutrient availability, and microbial communities, indirectly shaping plant functional traits and species composition. Biochar increased soil carbon, microbial diversity, and ammonium levels, promoting shrub growth, while wood mulch improved soil moisture and moderated nitrogen dynamics. Together, the amendments produced additive, lasting benefits that exceeded individual effects, pointing to their potential to improve reforestation outcomes where natural regeneration is limited. By integrating trait-based plant responses and soil–microbe interactions, land managers are equipped with science-based strategies to fine-tune restoration treatments for better ecosystem function and drought resilience.

3. Linking water age and wildfire effects improves streamflow predictions. ARS researchers in Fort Collins, Colorado, in collaboration with other federal scientists and faculty at universities across the western United States, have uncovered how wildfire, snowpack persistence, and groundwater storage interact to shape streamflow in mountain watersheds. This work shows that post-fire runoff responses differ by elevation, with persistent snow zones more prone to streamflow due to higher soil moisture, while low-elevation areas face increased overland flow risk when severely burned. Using tritium age dating, the team also found that snowmelt-driven streamflow is primarily composed of older groundwater, not recent precipitation as previously assumed. These findings provide land and water managers with critical insights to refine post-fire hydrologic models and improve water supply forecasting in snow-dominated regions.

4. Optimizing hydrological model forcing data selection. ARS researchers in Fort Collins, Colorado, conducted a comprehensive review and synthesis of 63 gridded climate datasets to guide hydrologic modelers in selecting the most appropriate data for their analyses. They evaluated datasets built from ground observations, satellite imagery, and reanalysis products, highlighting key differences in accuracy across terrain types and station densities. Drawing from 29 recent intercomparison studies, they found that ground-based datasets typically offer greater accuracy, especially for temperature and precipitation, though not always better streamflow predictions. Their results identify critical factors such as spatial resolution, coverage, latency, and variable interdependence that should inform dataset choice. These science-based recommendations will help hydrologic investigators justify and optimize model data selection for more accurate modeling.

5. Incorporating topography improves spatial predictions of dryland crop yields. ARS researchers in Fort Collins, Colorado, have developed an understanding of sub-field yield variability by analyzing multi-year data for wheat, corn, and proso millet across northeastern Colorado dryland farms. Utilizing high-resolution topographic, soil, and weather data, they identified key drivers impacting crop yield, with topographic position index (TPI) and soil texture exerting significant negative effects, while nitrogen application and soil carbon boosted yields. Their work revealed TPI as a surprisingly powerful predictor—rivaling nitrogen rates and outperforming traditional topographic indicators such as wetness index and slope. Using random forest models, they explained about 25% of yield variability, underscoring the complexity of spatial yield drivers and the need for ongoing research. This breakthrough provides critical insights to optimize cropland management and enhance precision agriculture strategies.

6. Soil moisture and strength vary within a satellite grid cell in the foothills of northern Colorado. ARS researchers in Fort Collins, Colorado, collaborated with Colorado State University researchers to explore how soil moisture and soil strength vary spatially within a 9 km by 9 km satellite grid cell. Extensive new spatial data were collected for soil moisture and soil strength within four geophysically distinct regions. The results contradict the typical assumptions used to downscale soil moisture to a practical resolution (10 m in this case), which may lead to suboptimal performance. Because soil strength is influenced by soil moisture, modeled soil strength predictions may improve when soil moisture estimates from downscaling procedures consider the variability in the underlying relationships revealed here. The findings support efforts to predict near real-time soil strength needed to traverse similar landscapes with large vehicles for a variety of purposes including transportation, agricultural management, and security.

7. Fine-scale soil moisture is estimated based on topography, soils and vegetation on a large ranch. ARS researchers in Fort Collins, Colorado, collaborated with researchers at Colorado State University to explore how to scale down coarse soil moisture from satellites to finer scales (3, 10, or 30 m) needed for practical decisions. The Equilibrium Moisture from Topography Plus Vegetation and Soil (EMT+VS) downscaling model that includes diverse topography, vegetation, and soils was applied to a 4,000 ha (10,000 ac) ranch in Northern Colorado. EMT+VS outperformed two widely used optical downscaling methods. The model also estimates the contributions of evapotranspiration, lateral water flow, and deep drainage to the spatial patterns of soil moisture on different dates related to landscape topography, soils and vegetation. Demonstration of the model skill at these scales has practical implications for agricultural management and mobility across large areas of rangelands and forested lands.

8. Soil-moisture thresholds influence soil CO2 emissions. ARS researchers in Fort Collins, Colorado, collaborated with Colorado State University researchers to quantify factors affecting CO2 emissions and their mitigation using smart agriculture. Machine learning statistical models were used to analyze the impacts of agricultural practices on environmental sustainability, including the contribution to greenhouse gas emissions. Scientists predicted the short-term soil CO2 emissions from organically amended systems (biochar and chicken and dairy manure) using soil moisture and weather variables as predictor variables. Applying biochar at a rate of 5 Mgha-1 reduced the soil CO2 emissions by 14.5% compared to the control plots. The classification and regression tree (CART) model was also applied and identified a soil moisture threshold of 10% volumetric water content for increased CO2 emissions. Overall, application of smart management procedures significantly reduces the short-term trend in CO2 emissions during the crop growing season.

9. Particle Swarm Optimization (PSO) gets a face-lift. ARS researchers in Fort Collins, Colorado, led university partners from Colorado State University in the development of a new graphical user interface (GUI) for a model calibration service. The Multi-Group Particle Swarm Optimization (MG-PSO) tool is a computational service developed to perform many model runs in parallel while the “swarm” of model parameters (“particles”) converge on an optimal solution. A new GUI was developed and deployed to help users adopt MG-PSO as an efficient tool for model calibration and sensitivity analysis, particularly for application of the spatially distributed ARS Ages watershed model, but potentially any model could leverage the service. Compared with previous sequential calibration tools, MG-PSO is at least ten times faster, which in practical terms can reduce a full model calibration from weeks to days or hours. As the tool is adopted, researchers in government and academia will save time and money getting model results to inform decisions in agriculture and water management.

10. Soil health indicators for water-limited regions: Evaluating relative sensitivity to compost and cropping intensification. Increased temperatures and shifting precipitation patterns present challenges to producers and scientists alike. Under this context, it is imperative to adopt management strategies that balance productivity and sustainability. ARS scientists in Fort Collins, Colorado, and Lubbock, Texas, in collaboration with scientists from Colorado State University and New Mexico State University, evaluated soil physical and biological parameters to assess the effects of management practices. Specifically, soil health was evaluated in two long-term studies – one in Akron, Colorado, under dryland conditions with compost addition and one in Clovis, New Mexico, in a limited irrigation system with various cover crops. Results indicated that compost was the main driver of changes to soil health indicators. Intensifying cropping system also impacted some soil health indicators, but the crop phase at the time of sampling often had a stronger effect than the treatment. These findings are particularly valuable for dryland and limited-irrigation farmers, as well as land managers aiming to improve soil health under variable climate conditions. They highlight the importance of tailoring management strategies by considering both long-term inputs like compost and the timing of crop phases to sustain productivity and resilience in the face of increasing climate variability.

Review Publications
Balch, J.K., Iglesias, V., Mahood, A.L., Cook, M.C., Amaral, C., Decastro, A., Leyk, S., McIntosh, T.L., Nagy, R.C., St. Denis, L., Tuff, T., Verleye, E., Williams, A.P., Kolden, C.A. 2024. The fastest growing and most destructive fires in the U.S. (2001-2020). Science. 386:425-431. https://doi.org/10.1126/science.adk5737.
Mahood, A.L., Barnard, D.M., Green, T.R., Macdonald, J., Erskine, R.H. 2024. Soil climate underpins year effects driving divergent outcomes in semi-arid cropland to grassland restoration. Ecosphere. 15. e70042. https://doi.org/10.1002/ecs2.70042.
Mankin, K.R., Mehan, S., Green, T.R., Barnard, D.M. 2025. Review of gridded climate products and their use in hydrological analyses reveals overlaps, gaps, and need for more objective approach to selecting model forcing datasets. Hydrology and Earth System Sciences. 29(1):85-108. https://doi.org/10.5194/hess-29-85-2025.
Fischer, S.C., Niemann, J.D., Scalia, J., Bullock, M., Proulx, H.E., Kim, B., Green, T.R., Grazaitis, P.J. 2024. Assessing the influence of model inputs on performance of the EMT+VS soil moisture downscaling model for a large foothills region in Northern Colorado. Journal of Hydrology. 650. Article e132397. https://doi.org/10.1016/j.jhydrol.2024.132397.
Kaiser, S., Fegel, T.S., Barnard, D.M., Mahood, A.L., Sparks, K., Wilkins, M., Rhoades, C.C. 2024. Long-term soil nutrient and understory plant responses to post-fire rehabilitation in a lodgepole pine forest. Forest Ecology and Management. 575(1). https://doi.org/10.1016/j.foreco.2024.122359.
Brooks, P.D., Solomon, D.K., Kampf, S., Warix, S., Bern, C., Barnard, D.M., Barnard, H.R., Carling, G.T., Carroll, R.W., Chorover, J., Harpold, A., Lohse, K., Meza, F., McIntosh, J., Neilson, B., Sears, M., Wolf, M. 2025. Groundwater dominates snowmelt runoff and controls streamflow efficiency in the western United States. Communications Earth & Environment. 6. Article e341. https://doi.org/10.1038/s43247-025-02303-3.
Wells, R., Mankin, K.R., Niemann, J.D., Kipka, H., Green, T.R., Barnard, D.M. 2024. Estimating changes in streamflow attributable to wildfire in multiple watersheds using a conceptual watershed model. Ecohydrology. 17(7). Article e2697. https://doi.org/10.1002/eco.2697.
Mahood, A.L., Macdonald, J., Muthukrishnan, R., Barnett, D.T., Sokol, E.R., Simkin, S. 2024. neonPlantEcology: An R package for preparing NEON plant data for use in ecological research. Ecological Modelling. 493. Article e110750. https://doi.org/10.1016/j.ecolmodel.2024.110750.
Bindner, J.R., Proulx, H., Wickham, K., Niemann, J.D., Scalia IV, J., Green, T.R., Grazaitis, P. 2025. Dependence of soil moisture and strength on topography and vegetation varies within a SMAP grid cell. Hydrology. 12(2). Article 34. https://doi.org/10.3390/hydrology12020034.
Mikha, M.M., Mankin, K.R., Khan, S.B., Barnard, D.M. 2024. Precision management influences productivity and nutrients availability in dryland cropping system. Agronomy Journal. 116(6):3325-3343. https://doi.org/10.1002/agj2.21686.
Miller, Q., Barnard, D.M., Sears, M., Hammond, J., Kampf, S. 2025. Variability in hydrologic response to wildfire between snow zones in forested headwaters. Hydrological Processes. 39(5). Article e70151. https://doi.org/10.1002/hyp.70151.
MacDonald, J., Barnard, D.M., Mankin, K.R., Miner, G.S., Erskine, R.H., Poss, D.J., Mehan, S., Mahood, A.L., Mikha, M.M. 2025. Topographic position index predicts within-field yield variation in a dryland cereal production system. Agronomy. 15(6). Article e1304. https://doi.org/10.3390/agronomy15061304.
Mahood, A.L., Stevens-Rumann, C., Rhea, A., Ritter, S., Barrett, K., Fornwalt, P.J., Barnard, D.M. 2025. Forest restoration treatments increase native plant diversity but open the door to invasion in the Colorado Front Range. Forest Ecology and Management. 593. Article e122881. https://doi.org/10.1016/j.foreco.2025.122881.
Ma, W., Cui, X., Han, W., Zhang, H., Zhang, L. 2025. Improved soil salinity estimation in arid regions: Leveraging bare soil periods and environmental factors. iScience. Article e113020. https://doi.org/10.1016/j.isci.2025.113020.
Doskocil, L.G., Fassnacht, S.R., Barnard, D.M., Pfohl, A.K., Derry, J.E., Sanford, W.E. 2025. Twin-peaks in streamflow timing: Can we use forest alpine snow melt-out response to estimate? Water. 17(13). https://doi.org/10.3390/w17132017.
Li, J., Zhang, H., Barnard, D.M. 2025. Transfer learning-based accurate detection of shrub crown boundaries using UAS imagery. Remote Sensing. 17(13). Article e2275. https://doi.org/10.3390/rs17132275.
Barnard, D.M., Mahood, A.L., Macdonald, J., Amundson, K., Fegel, T., Gleason, S.M., Guimond, M.M., Kaiser, S., Sparks, K., Wilkins, M., Rhoades, C. 2025. Indirect effects of soil amendments on plant traits and the microbiome in post-wildfire forest recovery. Forest Ecology and Management. 594. Article e122953. https://doi.org/10.1016/j.foreco.2025.122953.