Progress 10/01/23 to 09/30/24
Outputs
PROGRESS REPORT Objectives (from AD-416): Objective 1: Conduct long-term, big-data oriented, network scale research (e.g., LTAR, NWERN, PhenoCam, and SCINet) and data synthesis to innovate applications and solutions for adaptive land and watershed management. Sub-objective 1A: Develop and evaluate unoccupied aerial systems (UAS) remote sensing and other big data methods, analysis procedures, and applications for solving rangeland resource management problems. Sub-objective 1B: Develop databases and web-tools to extract and optimize North American Multi-Model Ensemble (NMME) historical seasonal forecasts for multi-site and regional forecasting applications across the LTAR site network. Sub-objective 1C: Evaluate Great Basin community and individual human dimension responses to cheatgrass, wildfire, and other challenges through expansion of spatially explicit formats integrated between ecological landscape layers with primary sociological data pertinent to management decision-making, perceived risk of annual grass invasion scales, and ranch-scale adaptation capacity behaviors. Sub-objective 1D: Continue to collaborate in LTAR and other network cross- site research projects contributing leadership, expertise, and data to address the regional and national scale problems concerning environmental health, agricultural productivity, and human dimensions of U.S. agroecosystems. Objective 2: Develop methods and tools to facilitate successful restoration outcomes in sagebrush-steppe ecosystems in the Great Basin. Sub-objective 2A: Assess the efficacy of prescriptive cattle grazing for restoring degraded sagebrush-steppe rangelands currently dominated by invasive annual grasses. Sub-objective 2B: Utilize seedlot and seedbed models to identify restoration opportunities in the highly variable and changing weather environment of the Intermountain western U.S. Objective 3: Assess weather and climate impacts on soil health, seedling establishment and rangeland productivity given the complex soil, vegetation, and topography typical of the western U.S. Sub-objective 3A: Utilize the Bureau of Land Management Land Treatment Digital Library (LTDL) and gridMET historical climate database to characterize what type of weather year is required to yield a positive restoration outcome after wildfire. Sub-objective 3B: Determine whether seasonal climate forecasting has sufficient skill for predictions of positive and negative restoration outcomes for post-fire seeding projects in the Great Basin. Approach (from AD-416): Goal 1A: Develop UAS remote sensing for monitoring rangeland fuel load, height, and continuity. We will use a combination ongoing field data and UAS imagery collections (initiated in 2015) to develop tools and workflows for characterizing fuels in 3 vegetations types which dominate much of the Great Basin region. Hypoth. 1B: The North American Multi- Model Ensemble (NMME) can be used to develop forecasting applications across a wide sector of U.S. agricultural. We will conduct hindcast assessments of all current NMME models for the period 1982-2022. Hindcast skill will be evaluated by comparing predictions to a gridded historical weather database, gridMET spanning the contiguous U.S. Goal 1C: Develop a socio-ecological adaptation capacity index for the northern Great Basin region. About 50-60 interviews will be collected from rural communities of the region. Focus Groups will conduct participatory analyses of adaptation drivers and challenges identified from the interviews, and then weight these factors across geographic/social contexts to develop an applicable index. Goal 1D: Develop long-term vegetation datasets in support of the LTAR network. We will continue ongoing collections (begun in 2015) of foliar cover, biomass, species richness/abundance, and other vegetation field data as well as phenology camera imagery along an elevational and precipitation gradient with the Great Basin LTAR site. Hypoth. 2A: High Intensity Low Frequency (HILF) beef cattle grazing will more effectively promote restoration of cheatgrass-invaded rangelands than lower intensity, BLM-permitted cattle grazing. We will continue analysis and publication of vegetation response data from the previous 9 years of this experiment. This experiment will then be replicated at a new study area. Hypoth. 2B: Seedbed microclimatic indices are correlated with native species distributions and the persistence and spread of invasive species over space. Time-series estimates of seedbed temperature and water potential will be developed for multiple plant materials, locations and temporal scenarios at selected field sites in the western U. S. using the Simultaneous Heat and Water (SHAW) model. Hypoth. 3A: Postfire seedling establishment success is correlated with winter/spring precipitation and winter temperature conditions in the year after planting. We will screen the USGS Land Treatment Digital Library (LTDL) records of post-fire rehabilitation treatments in the Great Basin to identify those with sufficient post-treatment data to indicate the level of both seed-mix and individual seedlot success in the first one to three years after seeding. We will then identify seedbed-microclimatic profiles that are correlated with relative seeding success or non-success at all selected LTDL field sites. Hypoth. 3B: Postfire seeding success in the Great Basin can be predicted using climate forecasts and historical weather and restoration data. We will identify and optimize seasonal forecasting models for all sites from Sub-objective 3A parametric/ nonparametric analyses will be used to determine whether climate metrics identified in Sub-objective 3A are associated with establishment success. This report documents progress for project 2052-21500-001-000D, "Disturbance Mitigation and Adaptive Restoration of Sagebrush-Steppe Ecosystems", which started in February 2024, and continues research from project 2052-13610-014-000D, "Assessment and Mitigation of Disturbed Sagebrush-Steppe Ecosystems". In support of Objective 1, ARS researchers in Boise, Idaho, used unmanned aircraft systems (UAS) to collect remote sensing imagery in the Johnston Draw study area, where a prescribed fire took place on October 6, 2023, within the Reynolds Creek Experimental Watershed (RCEW) near Murphy, Idaho. A total of 35 plots were surveyed with multispectral and visual band sensors pre- and post-fire. The plots were selected to be representative of the vegetation and fuel types of northern Great Basin at moderate elevations. Herbaceous fuel heights, loads, and continuity consequently varied by fuel type among plots. Seven of these plots were also surveyed post-fire using a UAS-borne hyperspectral sensor. A workflow for georectifying UAS imagery using coded ground control point targets was developed and published as an online tutorial in the USDA Scientific Computing Initiative (SCINet) Geospatial Workbook. Phenology cameras (PhenoCams) located at the Nancy Gulch, Lower Sheep Creek, and Reynolds Mountain sites within the RCEW were enhanced and maintained. All three automated cameras successfully contributed imagery to the nation- wide PhenoCam and Long-Term Agroecosystem Research (LTAR) networks. Field data for the RCEW Long-Term Vegetation Research (LTVR) program and the Great Basin LTAR site were collected as planned. ARS researchers in Boise, Idaho; Temple, Texas; Burns, Oregon, and collaborators at the University of California, Merced, and California Polytechnic State University, San Luis Obispo, assessed the potential utility of using seasonal climate forecasts to predict plant production in the sagebrush/bunchgrass rangeland of southeastern Oregon. Historical data were used to develop predictive models relating seasonal climate to annual yield, and for assessing seasonal forecasting skill at estimating monthly temperature and precipitation during the upcoming growing season. It was determined that seasonal climate forecasts in this region have sufficient skill to justify developing web-based technology to assist local land managers in making stocking decisions, anticipating wildfire fuel loads, and for restoration planning and wildlife management. ARS researchers in Boise, Idaho, funded collaborative research with University of Idaho, including hiring a Postdoc, to continue rural sociology and other human dimension research concerning the invasive grass/wildfire syndrome and in support of the LTAR network. In support of Objective 2, ARS researchers in Boise, Idaho, established sampling sites and collected field data on the new LTAR Common Experiment (CE) study area selected during the previous project plan cycle. These field data will serve to establish the initial, pre-treatment vegetation and soil conditions for the new application of the CE at this study area. For Objective 3, ARS researchers in Boise, Idaho; Temple, Texas; and Burns, Oregon, and collaborators at the University of California, Merced, conducted a test-case study to assess the potential for seasonal climate forecasting of soil favorability for rangeland restoration in the Great Basin sagebrush-steppe. Seasonal forecast models from the North American Multi-Model Ensemble (NMME) were determined to have significant skill at providing precipitation and temperature forecasts in October for the upcoming establishment season (October � May). Previous research has established likely patterns of seasonal temperature and precipitation that either facilitate or inhibit seedling establishment in this region. The current study results were sufficiently positive to warrant additional evaluation of historical Bureau of Land Management seeding projects contained in the Land Treatment Digital Library. Artificial Intelligence (AI)/Machine Learning (ML) Artificial Intelligence (AI), including Machine Learning (ML) and Deep Learning (DL) methods were used for this project. Transformer neural networks and self-supervised anomaly detection with masked autoencoder methods were used in LTAR and SCINet-funded research for detecting and removing spatial error from global positioning system (GPS) livestock tracking datasets. This work was done on both local and the Scientific Computing Initiative Network (SCINet) computing hardware. Conventional research approaches to the GPS error problem have been applied for many years but have always fallen short given the dynamic complexity of factors influencing the presence and size of error. DL methods provided a means to reduce residual error in tracking datasets to the point these data were suitable for high-resolution, movement characterization and foraging optimization analyses. Both ML and DL methods were used in developing tools to classify and count animals evident in camera trap data sets under a SCINet-funded project. This work was done on both local and SCINet computing hardware. Use of AI methods has yielded a vast improvement in efficiencies over manual, human-classification and census of imagery. Application scope has been clearly increased by the use of AI methods. Additionally, random forests were used for landscape classification mapping. The work was done on local computing hardware. This ML approach provided much better classification accuracy than the more traditional unsupervised and supervised approaches. ACCOMPLISHMENTS 01 Developing rangeland fuels mapping for prescribed fire planning. Western juniper/pinyon pine woodlands occupy over 22.8 million hectares of North America and are encroaching on sagebrush steppe at a rate of 4, 600 square kilometers per year. Sagebrush steppe is a critical source of forage for livestock and habitat for sagebrush obligates like the greater sage grouse. Developing safe and effective juniper control strategies using prescribed fire are often challenged by a lack of spatial information on pre-fire vegetation fuel conditions. ARS researchers in Boise, Idaho, investigated the use of commercial high- resolution satellite imagery and machine learning (ML) artificial intelligence (AI) analysis methods to create pre-fire maps of vegetation fuel types with the purpose of informing prescribed fire planners of spatially relevant, rangeland fuel conditions. Seven separate fuel types, e.g., western juniper, bitterbrush, sagebrush, etc. , were mapped to very high resolution (50-cm) with a user�s accuracy of 83% using random forest ML methods. These validated maps were presented to the public via the Ag Data Commons repository and received 710 downloads in their first two months of availability. These fuels maps will improve the efficiencies associated with the 1.5-billion-dollar cost of fire prevention, suppression, and restoration in the United States and, more importantly, will ultimately help protect human lives, property, and natural and cultural resources across extensive areas of the western United States. 02 Assessing regional differences in the seasonal germination response of bottlebrush and big squirreltail. Bottlebrush and big squirreltail are high priority species for restoration of millions of hectares of rangeland in the western United States that have been degraded by wildfire and introduced annual grasses. ARS researchers in Boise, Idaho, modeled germination responses of each of these species as if the species had been planted in its habitat of origin and alternatively, as if it was planted in the other species' habitat. The experiment was simulated across the past 40 years of climatic variability. Boise researchers confirmed germination responses differed enough between species to signal clear advantages for seeding these species into their respective habitats of origin. Virtual simulations of this type can be used to identify establishment traits that are associated with specific environments and could inform land managers of the most suitable plant materials for restoration of plant communities that are under threat from invasive weeds and climate change.
Impacts
(N/A)
Publications
- Acciaro, M., Pittarello, M., Decandia, M., Sitzia, M., Giovanetti, V., Lombardi, G., Clark, P. 2024. Resource selection by Sarda cattle in a Mediterranean silvopastoral system. Frontiers in Veterinary Science. 11. Article 1348736. https://doi.org/10.3389/fvets.2024.1348736.
- Hardegree, S.P., Richards, C.M., Sheley, R.L., Reeves, P.A., Jones, T.A., Walters, C.T., Schantz, M.C., Flerchinger, G.N. 2024. Virtual reciprocal garden assessment of germination syndromes for Elymus elymoides ssp. brevifolius and Elymus multisetus. Rangeland Ecology and Management. 96:1- 11. https://doi.org/10.1016/j.rama.2024.04.013.
- Schantz, M.C., Hardegree, S.P., Sheley, R.L., Abatzoglou, J., Hegewisch, K. , Elias, E.H., James, J., Moffet, C. 2024. Forecasts for rangeland management applications in the western United States. Rangeland Ecology and Management. 94:207-214. https://doi.org/10.1016/j.rama.2024.03.008.
- Sonnier, G., Augustine, D.J., Paudel, S., Porensky, L.M., Silveira, M., Toledo, D.N., Azad, S., Boughton, R., Browning, D.M., Clark, P., Fay, P.A., Kaplan, N.E., Thibault, K., Swain, H.M., Veum, K.S., Boughton, E. 2024. Impact of plant diversity and management intensity on magnitude and stability of productivity in North American grazing lands. Applied Vegetation Science. 27(2). Article e12776. https://doi.org/10.1111/avsc. 12776.
|