Monitoring standing herbaceous biomass and thresholds in semiarid rangelands from harmonized Landsat 8 and Sentinel-2 imagery to support within-season adaptive management

Authors: Kearney, Sean; Porensky, Lauren; Augustine, David; Gaffney, Rowan; Derner, Justin

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/17/2022
Publication Date: 2/1/2022
Publication URL: https://handle.nal.usda.gov/10113/7638449
Citation: Kearney, S.P., Porensky, L.M., Augustine, D.J., Gaffney, R.M., Derner, J.D. 2022. Monitoring standing herbaceous biomass and thresholds in semiarid rangelands from harmonized Landsat 8 and Sentinel-2 imagery to support within-season adaptive management. Remote Sensing of Environment. 271. Article e112907. https://doi.org/10.1016/j.rse.2022.112907.
DOI: https://doi.org/10.1016/j.rse.2022.112907

Interpretive Summary: Ranchers and other rangeland managers require information about forage biomass (i.e., quantity) to make decisions and learn from past management strategies. Since rangelands cover such vast areas and conditions change quickly in response to precipitation and temperature, it is challenging to measure forage biomass frequently and in a way that reflects different conditions in different areas. New satellite-based tools are emerging to estimate total forage production in a given season, but they do not account for forage ‘losses’ during the season from events such as grazing, trampling, insects, fire and weather (e.g., hail). We developed a satellite-based model to estimate standing forage biomass (i.e., the quantity of non-woody vegetation present at a given point in time) in shortgrass prairies with low error. The model was able to detect the week in which pasture biomass dropped below key decision-making thresholds (e.g., 400 lbs./ac) with about 70 – 87% accuracy, depending on the threshold. We also demonstrated how our model can be used to predict the probability that forage biomass in a given pasture or location has dropped below a defined threshold. This will give rangeland managers the flexibility to incorporate their own levels of risk and perception into decision making. The satellite data we used to develop our model will allow us to produce ‘near-real time’ biomass maps going forward, specifically a new map every 2-3 days showing conditions 7-10 days in the past. We can also look back across the entire satellite time series dating back to 1986 to generate historical maps and compare current conditions to long-term averages for a given site. These near-real time and historical time series maps will provide rangeland managers with new tools to support adaptive management by allowing them to make more informed decisions and monitor outcomes of management strategies across space and time.

Technical Abstract: Adaptive management often requires rangeland managers to respond to changing forage conditions (e.g., standing herbaceous biomass) within the grazing season at the scale of individual pastures. While within-season biomass can be directly measured or estimated in the field, it is often impractical to do this in extensive rangeland systems with adequate frequency and spatial representation for responsive decision-making by rangeland managers. We sought to develop a single model to accurately predict daily herbaceous biomass across seasonally- and annually-varying conditions from satellite imagery, specifically the Harmonized Landsat-Sentinel (HLS) time series. We also sought to assess how information about plant community composition derived from a high-spatial resolution map would improve model performance. We used herbaceous biomass data from 1,764 sample plots collected over 8 years in North American shortgrass steppe for training in a cross-validated model selection approach to evaluate (1) predictive performance of candidate models both spatially and temporally, (2) relative variable importance of individual spectral bands, vegetation indices (VI), and recently developed broadband spectral angle indices (BAIs), and (3) the degree to which including plant community composition improved model performance. All 11 candidate models identified in the model selection procedure contained a BAI and an individual spectral band, and 6 contained one of each input feature type, demonstrating the benefit of combining these HLS features for predicting herbaceous biomass across varying conditions. The spatial and temporal cross-validation and selection procedures produced the same top model with similar performance (mean absolute error [MAE] = 151 – 182 kg ha-1; R2 = 0.75 – 0.79), suggesting that a single model performs well over space and time. Including plant community composition in the model reduced MAE by 11 ~ 13%. Bootstrapping revealed that 6 - 7 years of training data were required to achieve consistent model performance across years with varying conditions (e.g., wet, average, and dry). The top model could accurately detect (70 – 87% accuracy) the week that biomass dropped below biomass thresholds as low as 450 kg ha-1 with modest commission error (10 – 26%). Developing near-real time temporal maps showing the probability that herbaceous biomass is below a given threshold supports more widespread adoption potential for adaptive management in extensive rangelands across differing scenarios of risk perception and avoidance for decision-making by rangeland managers. These maps are an integral component for the development of precision conservation, livestock, and rangeland management strategies for the sustainable provision of multiple ecosystem services from extensive rangelands.