Project : USDA ARS

ARS Home » Pacific West Area » Tucson, Arizona » SWRC » Research » Research Project #442115

Research Project: Understanding Ecological, Hydrological, and Erosion Processes in the Semiarid Southwest to Improve Watershed Management

Location: Southwest Watershed Research Center

2025 Annual Report

Objectives
Objective 1. Quantify the magnitude and variability of the water balance components in semiarid landscapes and identify their controlling processes. 1.A: As an LTAR observatory, continue to collect and curate WGEW datasets including precipitation, runoff, sediment, pond runoff and sediment, meteorology, soil moisture, fluxes, vegetation, spatial datasets, and make datasets available under FAIR principles. 1.B: Quantify intra-storm variation in stable isotope values of precipitation over WGEW and identify relative influence of moisture source, season, local weather and sub-cloud processes. 1.C: Track daily watershed water balance components for rangeland ecosystems in the WGEW and SRER for improved assessment of water status and associated productivity. 1.D: Incorporate a variety of enhancements into watershed and erosion models maintained by the SWRC to add additional sub-processes, reduce predictive uncertainty, make them easier to use, enhance integration with land management agency workflows, and expand their use geographically. Objective 2: As part of the Long-Term Agroecosystem Research (LTAR) network, characterize and quantify impacts of water and agriculture/water management on semiarid watershed and agroecosystem processes. 2.A: Assess how novel remote sensing tools and low-cost, automated optical imagery can be used to quantify evapotranspiration and vegetation carbon uptake in water-limited regions. 2.B: Improve large-scale mapping of rangeland vegetation cover, lifeform, and biomass to classify rangeland ecological sites and states. 2.C: Quantify the long-term variability of riparian woodland evapotranspiration and CO2 exchange and their controls. 2.D: Assess impacts of altered temporal rainfall regime on semiarid grassland water and carbon cycling processes. Objective 3: Quantify and predict effects of climatic change, plant community transitions, and conservation practices on ecological, hydrological, and erosion processes. 3.A: Develop new conceptual and quantitative frameworks to assess the impacts of brush management on ecosystem structure and function and enhanced delivery of ecosystem services. 3.B: Assess impacts of climate change, wildfire, and vegetation management on hydrology and erosion processes across spatial scales within the rangeland-dry forest continuum. Two Goals are included for this Sub-objective. 3.C: Conduct field-based experiments on southwestern U.S. rangelands to assess the impact of woodland encroachment/infilling and tree removal conservation practices on vegetation, surface soils, and hydrology and erosion processes. 3.D: Evaluate the hydrologic, geomorphic, and ecologic impacts of failed soil and water conservation structures in Southwest rangelands. 3.E: Quantify how weather variability and potential changes in climate impact ecosystem net and gross carbon uptake in the water-limited Southwest. 3.F: Quantify how snowmelt amount and timing are impacted by vegetation structure under changing climate, wildfire, and vegetation management in the semiarid interior western U.S. 3.G: Estimate runoff and erosion risks over western U.S. rangelands.

Approach
Objective 1. A. Collect and make available Walnut Gulch Experimental Watershed (WGEW) datasets including precipitation, runoff, sediment, pond runoff and sediment, meteorology, soil moisture, fluxes, vegetation, spatial datasets. B. Quality-control and collate precipitation samples during summer rainfall events using a custom autosampler. C. Make measurements of precipitation, soil water content, runoff and evapotranspiration in the headwater watersheds of the WGEW and Santa Rita Experimental Range from the SECA network to track daily water balance components. D. Add functionality to existing runoff and erosion models to improve the applicability and ease of use for watershed management and assessments. Objective 2. A. Evaluate novel remote sensing spectral tools across the gradients of spatial and temporal dryland measurements. B. Use field measurements of cover, biomass and lifeform along with remotely sensed data to classify states on ecological sites. Structure from Motion will be used to estimate the distribution of cover and biomass by lifeform using machine learning (ML) and estimate erosion and runoff model parameters within the common site/state combinations. C. Use eddy covariance flux data from a riparian woodland site to better understand what controls annual ET and productivity. D. Utilize the Rainfall Manipulation facility in the SRER to fully control precipitation (using rainout shelters and irrigations) over hydrologically isolated plots with equal mixtures of multiple semiarid grassland plants and initiate hydroclimate disturbance treatments. Objective 3. A. Test for impacts on measured runoff after brush management treatments and demonstrate Rangeland Hydrology and Erosion Model (RHEM) capability to accurately simulate runoff and erosion processes for tree canopy and intercanopy areas on untreated and treated sites. B. Conduct a series of field studies quantifying impacts of fire on vegetation, ground cover, soil water repellency, infiltration, and runoff and erosion processes, and evaluate climatic and vegetation controls on surface water supplies using daily streamflow records in watersheds of the Colorado River Basin. C. Use artificial rainfall simulation and overland flow experiments to quantify infiltration, runoff, rainsplash, and erosion on tree-encroached sagebrush with tree-removal practices. D. Quantify the impacts of failed conservation structures using LiDAR data, aerial photographs, and satellite imagery. E. Use water and carbon flux data to better understand ecosystem responses to short and long term climate variability and improve models. F. Combine various datasets to quantify how snowmelt amount and timing are impacted by vegetation structure. G. Employ ML methods complemented by auxiliary data to develop relationships to field-collected variables from monitoring locations across the West and determine if ML techniques can predict RHEM parameters and runoff and erosion predictions directly.

Progress Report
This report documents progress for project 2022-13610-013-000D, titled, “Understanding Ecological, Hydrological, and Erosion Processes in the Semiarid Southwest to Improve Watershed Management”. In support of Sub-objective 1A, soil moisture and eddy flux data from two sites on the Walnut Gulch Experimental Watershed and two sites on the Santa Rita Experimental Range were updated and made available through AmeriFlux. The meteorological data from three sites on Walnut Gulch were also updated and available via the Unit’s website, and the fourth meteorological site on the Audubon Research Ranch still needs additional processing and quality control. Vegetation cover data were collected at the long-term transects. Under Sub-objective 1C, water and carbon balance figures on the Santa Rita Experimental Rangeland website are still being updated approximately monthly. For Sub-objective 1D, an Automated Geospatial Watershed Assessment (AGWA) tool has been successfully implemented in the ArcGIS Pro environment with the beta version released on July 30, 2024. Since then, several important milestones were achieved including adding complex slope functionality and RHEM2.4/2.5 (Rangeland Hydrology and Erosion Model) into AGWA Pro, optimizing the Gridded Soil Survey Geographic Database (gssurgo) query for soil parameters, and adding automatic clipping to reduce preprocessing requirements. Additionally, software bugs reported by users were resolved. In support of Sub-objective 2A, Red-Green-Blue (RGB) measures of greenness have proven to be more reliable, affordable and informative than sun-induced fluorescence (SiF). An aging network of automated RGB cameras at RainMan was replaced and newly programmed with data hubs to reduce data collection labor by about 80%. From these photos, daily greenness was calculated for 60 experimental rainfall plots to determine how different rangeland plant types respond at different speeds to summer rains, regulating the timing of forage production. Also, under Sub-objective 2A, thermal imagery was previously shown to be more informative than SiF. Time series of thermal images were processed for 25 experimental rainfall plots at RainMan and used to estimate how much of the water vapor lost is due to plant transpiration vs. non-biological evaporation. Additionally, thermal imagery collected at several conifer sites was compared with sap flow measurements from trees during drought and non-drought conditions to see if this remote sensing tool can monitor plant water use. For Sub-objective 2B, biomass estimates from the Rangeland Analysis Platform have been analyzed over time across Walnut Gulch. Under Sub-objective 2D, soils were evaluated for their relationship between volumetric water content and matric potential. These samples were collected from 36 locations in the RainMan experiment, and the results will help to quantify the location, depth, and severity of drought stress experienced by rangeland plants under rainfall treatments. Also related to Sub-objective 2D, data from an intensively monitored site in California were used to observe how landscapes recover from drought and which drought metrics capture recovery. The root-zone storage deficit (a highly effective remotely sensed metric for total subsurface moisture) was explored for robustness. Evaluation of root-zone deficit performance was performed across a set of large basins important for water supply in California. Models were then developed with and without the deficit metric included to forecast water supply at water supply basins in California. A range of model structures (linear regression, random forest, gradient boosting, etc.) were tested against the historically used regression models to establish performance gains of altering model structure vs. including a deficit term. To evaluate variations in surface water expression, machine learning models were developed to predict wet/dry state from PlanetScope satellite imagery. In support of Sub-objective 3A, statistical analysis of the change in runoff at flume one over time has been done, and some of the sub-watersheds have been modeled over time, though the overall analysis is incomplete. ARS researchers in Tucson, Arizona, have also been consulting with the University of Arizona and Arizona Department of Forestry to have them reduce the mesquite cover at the Santa Rita Grassland site to determine how brush management can alter ecosystem water and carbon fluxes. In support of Sub-objective 3B, vegetation and ground cover conditions, soil physical properties, spatial patterns in soil water repellency, and infiltration and erosion rates were quantified through rainfall simulation and overland flow experiments in unburned and recently burned areas of dry forests in the White Mountains in Arizona. Data from the field experiments contribute to improved understanding of the impacts of wildfire on (1) soil hydraulic properties, (2) the presence and persistence of soil water repellency, (3) runoff generation, and (4) erosion rates. The resulting database provides a basis for parameterizing respective fire effects in RHEM. For Sub-objective 3C, at Grand Staircase Escalante National Monument, a suite of vegetation, soil, and rainfall simulation experiments were applied to quantify the effects of pinyon and juniper removal on sagebrush vegetation, hydrologic function, and soil erosion. Field experiments at Pecos National Park employed infiltrometer and soil moisture sampling approaches to quantify impacts of pinyon and juniper removal on vegetation, infiltration, and soil water storage on grassland sites. Data from these field studies were compiled for subsequent syntheses to better understand the effects of tree removal (fuel reduction) on agroecosystem productivity and ecosystem functioning. Under Sub-objective 3D, analysis of data from the erosion-control berm database was done to identify which berm aspects are associated with a strong positive impact or with structural failure to inform future design. A water balance model was developed using data from Walnut Gulch to explore the landscape-scale impact of berms and stock ponds on water balance. For Sub-objective 3E, micrometeorological data collection for the Semiarid Ecohydrological Array (SECA) sites continued. Site data was quality-checked twice yearly and submitted to the AmeriFlux database where the dataset was published and made publicly available. Flux tower data was also used to quantify differences in evapotranspiration between burned and unburned forested areas in California. Also in support of Sub-objective 3E, satellite remote sensing and ground-based datasets were used to complete a global scale analysis of how changes in daily-scale precipitation variability alter plant productivity. For Sub-objective 3F, machine learning was developed for the future purpose of computationally efficient upscaling of snowpack and snowmelt information previously produced by a small-scale, hyper-resolution snow model. The upscaled snow maps assess how forest management alters the amount and timing of snowmelt water resources. One new Snowtography (snow-photography) station was installed to evaluate the impacts of mechanical thinning treatments in northern Arizona pine forest. The sites for two new stations were surveyed in a central Arizona shrub mastication project. Watershed spatial analyses were conducted to identify candidate sites for six new stations in Colorado and Wyoming. The Snowtography network operated by stakeholders under ARS scientific coordination reached 24 stations comprising daily snow measurement at about 700 points arrayed across gradients of elevation, forest type, and vegetation management. Additionally in support of Sub-objective 3F, improvements were made to SnowPix, a database of vegetation, snow, and soil moisture information from more than 20 Snowtography sites across the Colorado River Basin. Key improvements include development of automated scripts to query and update >40 dataloggers at remote sites (via cellular data connections), code development to make standardized figure sets across many sites, and code development to define ecohydrologic seasonality for each site and year based upon snowpack and soil moisture data. Also under Sub-objective 3F, a book chapter was written for a handbook of terrestrial ecohydrology on the topic of how forest disturbance and management alter hydrology. Under Sub-objective 3G, a neural network RHEM emulator and cover estimation model was recreated using PyTorch due to TensorFlow incompatibility issues on platforms such as Google Colab and SCINet. Several efforts were made to improve the Artificial Neural Networks (ANN) cover model, including downloading and processing Gridmet and Daymet climate data clipped to 66,432 National Resource Inventory sampling points. Largest Patch Index (LPI) vegetation transect data from Reynolds Creek and Walnut Gulch were collected for validating the model, and a new data collection effort was initiated with the University of Nevada, Reno, to collect additional LPI data at the Porter Canyon watershed for model validation. Additionally, under this sub-objective, a new project scientist examined the impact of groundwater pumping on water temperature in streams based on two different modeling approaches, and evaluated a new method for quantifying streamflow depletion from groundwater pumping in small catchments with fractured bedrock.

Accomplishments
1. A rangeland hydrology and erosion model for snow-dominated uplands. Runoff and erosion modeling for high elevation rangelands must account for water input as rainfall and snowmelt to effectively forecast hydrologic and erosion risks to resources, property, and human life. ARS researchers in Tucson, Arizona, and Boise, Idaho, the Natural Resources Conservation Service, in Davis, California, and the University of Arizona in Tucson, Arizona, developed a new approach to effectively represent rainfall and snowmelt contributions to hillslope runoff and erosion predictions by the Rangeland Hydrology and Erosion Model (RHEM). Assessment of the new approach found that more effective partitioning of water inputs from rain and snow can reduce predicted runoff and sediment responses at the annual time scale by more than 20% relative to treating all water input as rainfall. For rain-on-snow events, the new approach found runoff and sediment responses can either be enhanced or muted depending on variability in storm and snowpack characteristics. Overall, the new approach more appropriately partitions water input and provides more realistic prediction of the magnitude of cold-season runoff and erosion events from snow-dominated uplands relative to that of previous RHEM versions. The advances in this study will enhance the capacity of resource managers and other users to make more accurate predictions in high elevation rangelands.

2. Global precipitation intensity datasets for water erosion prediction. Simulating water erosion accurately requires data on how hard it rains (precipitation intensity). However, most precipitation measurements are at the daily scale, and subsequently, they do not capture the needed sub-hourly peak intensities that cause erosion events. As a solution, the CLImate GENerator (CLIGEN) tool can produce long-term precipitation time series as inputs to soil erosion models that are appropriate for prediction of annual average erosion rates. ARS scientists in Tucson, Arizona, and from the University of Arizona used CLIGEN to generate inputs for two erosion models, and results showed insignificant differences in erosion rates when compared to using observational-based inputs. The CLIGEN inputs provide modelers access to precipitation inputs appropriate for erosion prediction at the point and hillslope scale for application around the world.

3. Field-scale experiment reveals key responses of rangeland plant productivity to changes in daily rainfall timing. Five decades of daily rainfall data have shown that rainstorms are becoming larger, but with longer intervening drought intervals, across much of western U.S. rangelands. ARS scientists in Tucson, Arizona, conducted a five-year field experiment measuring how desert grasslands respond to fewer, larger rainfalls. Automated cameras aimed at 60 experimental plots revealed how the deeper-rooted perennial bunchgrasses, which are promoted by fewer, larger rainfalls are also slower to respond to rain, delaying the peak production of forage for grazing. Furthermore, year-round measurements showed that dry winters were followed by higher grass production the following summer, while controlling for summer rainfall amount, likely due to slower litter decomposition and accumulating nutrients during dry winters. Together, these results provide updated guidance for planning efficient grazing and rangeland management.

4. Availability and quality of nectar sources are critical to the survival of monarch butterflies. A change in the availability and quality of nectar plants during the monarch butterfly fall migration though the Great Plains Region, United States is thought to be a critical factor in the species’ population decline in North America. ARS researchers in Tucson, Arizona along with scientists from the Natural Resources Conservation Service in Fort Worth, Texas, utilized vegetation data from more than 8000 study sites sampled in 2021 to quantify and assess the quality and availability of monarch butterfly preferred nectar sources and associated ecosystem attributes and functioning along the autumn migration pathway through the Great Plains Region. The study found that preferred nectar sources for monarchs decrease in abundance from north to south along the autumn migration through the Great Plains Region, but that nectar sources at southern latitudes along this pathway remain diverse. Results indicate this nectar diversity may maintain components of key habitat in the wake of ongoing land use intensification. The study presents the most extensive quantification to date of preferred nectar sources for monarch butterflies along the Great Plains autumn migratory pathway and provides farmers, ranchers, and other producers and land managers vital information to aid voluntary conservation efforts of the monarch butterfly in North America.

U.S. DEPARTMENT OF AGRICULTURE

Southwest Watershed Research Center: Tucson, AZ