Integrated field and satellite based decision support system for climate-resilient and sustainable ranches and rangelands across California

This project focused on California’s rangelands,  where climatologically, ecologically,  and topographically diverse landscapes pose the biggest challenge for sustainable ranching (Fig. 1).  During Year 1 of the project, we made significant progress to address the four interconnected research objectives (Fig. 2). This laid the foundational work for the development of  monitoring growth status, forage production, and RDM, to facilitate adaptive management and decision-making.

Research Objective 1. Build an extensive field-based geospatial database on the aboveground biomass metrics over California’s rangeland, to be used to calibrate and evaluate the remote sensing-based monitoring algorithm.

UCCE and FSA collaborated (FSA funded field supplies and Loss Adjuster labor cost) to identify and establish approximately 40 monitoring sites (700 m2) that will provide a representative distribution of annual rangelands across all sub-ecoregions statewide (Fig. 1). In late spring of 2022, we reconvened our partners, revisited the exclosure site selection, and revised sampling and measurement protocol, based on the experiences during previous field season (September 2011 to March 2022). A total of 42  sites were finalized,  representing the diversity of the annual grasslands across all sub-ecoregions statewide (Fig. 1). The exact locations of each sites were recorded by high precision Trimble GPS. To match the footprint size of available moderate to high spatial resolution satellite remote sensing data (~10 to 30 m), each site was situated in a large open grassland (> 45 x 45 m, a.k.a., Macro plots) and includes one large exclosure (14.63 x 14.63 m) to accommodate remote sensing and field clip-plot quantification. A total of 17 county based UCCE farm advisors and 5 regional FSA “Loss Adjusters” collected field data in 2021/22 and 2022/2023.

Forage clippings were done once or twice a month from January to March to capture the rapid growth period, once in April/May to capture peak standing crop, and once from late September or October to capture RDM. Plant functional group composition (percent grass vs forb cover) was recorded at each sampling event (2021/2022 and 2022/2023 seasons); additionally, dominant species were recorded via the dry-weight-rank method at each site at peak standing crop (2022/2023 season). For the 2021/2022 season, clippings were specifically collected within non-grazed exclosures and in adjacent grazed areas. A subset of samples were sorted to separate new growth from old-growth (i.e., RDM) to measure the current year’s forage biomass to match satellite based estimations. This approach will allow us to assess the ability of the remote sensing platform to quantify forage biomass in paired grazed vs non-grazed areas, which can differ in build-up of RDM. For the 2022-2023 season, we continued to clip forage samples within grazing exclosures, field sorting samples with greater than 15% RDM cover.

For each monitoring site, 10-15 samples were clipped during each clipping event at each site. The clipping samples were averaged by date for each site, resulting in a total of 150 site-date data points for further analysis. By following the same protocols for sampling, field data collection, and preprocessing, we built a large geospatial database of forage production with high quality and consistency, which will continue to grow as new clipping data comes in.

Research Objective 2. Develop a well-calibrated remote sensing algorithm scalable statewide for consistent forage production estimates in near real-time. 

We will improve upon the remote sensing approach that we developed for forage production estimation using MODIS daily 250-m and Landsat 16-day 30-m satellite observations (Liu et al., 2021). Our prior approach was calibrated with limited exclosure site data. Moreover, it was applied only to herbaceous-dominated rangelands, to avoid the mixed woody-herbaceous pixel issue at 30-m resolution. In order to provide daily estimates for the majority of rangelands without the masking effects from shrubs and trees, higher spatial resolution imagery is needed as a main data stream. Since PlanetScope's open California data policy phased out, we  relied on Sentinel 2A&B at 10 m every 5 days (plus Landsat), both publicly-available and free, for operational near real-time forage monitoring (Fig. 2). Our focus will be on all 10 ×10-m pixels without trees or shrubs. A first version of the base mask for herbacious cover at 10m was generated by first combining herbaceous or grasslands pixels from California vegetation map, the European Space Agency (ESA)  World Cover, and the National Land Cover Database (NLCD). Only pixels with NDVI lower than 0.3 in mid-August were retained for further analysis since the presence of trees and other perennial vegetation within a 10m pixel would increase the NDVI value compared to senescent grasses.  In the next step, we will further refine the layer with the meter scale NAIP aerial imagery, available every other year, and updated with Sentinel summer/fall-time imagery when grass is senescent.

To estimate forage production, we adopted the similar biophysical light use approach, developed from our previous studies (Liu et al., 2019, 2021), and worked on further improvement of accuracy and robustness, especially with higher spatial and temporal resolution of Sentinel 2 satellite imagery than previously used Landsat 30m imagery, and leveraging the substantial field data from the newly established network of the exclosure sites for model training and testing.

We assembled weather data,  including incoming solar radiation, temperature, precipitation, were also downloaded from CIMIS and PRISM. We also assembled the topographical indices such as elevation and soil data such as soil organic matter. We have downloaded Sentinel 2A & 2B imagery at 10m to 20m with repeated observations every 5 days from 2022 to May 2023, with a total of 16 tiles covering all 42 field sampling sites. An automatic preprocessing workflow has been developed to filter clouds and shadow for each of the 12 spectral bands.

Light interception and light-use efficiency optimization will be further improved by incorporating the three unique red-edge bands from Sentinel-2, which Landsat does not include. Reflectances in the red-edge spectrum is more sensitive to chlorophyll absorption and thus reduces the saturation issue of the previously widely used NDVI at higher biomass levels (Lin et al., 2019). Moreover, the red-edge-based vegetation indices are more sensitive to the spatio-temporal variations of pigment degradation and therefore serves as a proxy for plant stress related with the light-use efficiency reduction (Zarco- Tejada et al., 2018). During Year 1, we calculated a set of vegetation indices for monitoring phenology and growth and plant stress, including the Normalized Difference Vegetation Index (NDVI), the red-edge bands based NDVI, the Simple Ratio (SR), the red edge bands based SR, the Soil Adjusted Vegetation Index (SAVI), and the Normalized Burn Ratio (NBR).  The Savitzky-Golay filter was applied to the NDVI time series to fill the gap and remove the abnormal values; a linear interpolation was then performed to generate the daily time series of NDVI. We will evaluate the relationships of various red-edge related indices with measured production and include the most significant ones for the refined approach.

For each individual 10 m grass pixel, we have already identified the phenological metrics in California annual grasslands such as the start of the growing season (SOS), end of the growing season (EOS), and length of the growing season (LOS).  The Sentinel-2 NDVI time series of each pixel was fitted with two sigmoidal curves and the rate of change was then used to identify the SOS and EOS and estimate the growing season length, which was fed into the daily accumulation of forage production.

We started to explore the light use efficiency optimization approach with the available field data and extracted Sentinel-2 data. The baseline model was based on a constant light use efficiency determined by the accumulated absorbed photosynthetically active radiation (APAR) and the field measured forage production before and at the peak production. The improved optimization of LUE factored the impact of environmental stressors. We are in the process of including three Sentinel 2 unique red-edge bands that can indicate the moisture stressors constraining plant growth for LUE optimization.

We will use 70% of the field data, randomly selected, for the optimization process, and use the remaining 30% of the data for validation. Uncertainty will be quantified in terms of annual forage production, peak standing biomass and phenology. This remote sensing algorithm will be further optimized as new field measurements become available. The validated remote sensing tool will be implemented over all of the herbaceous-dominated rangeland areas that are within the Mediterranean climate zone.

Research Objective 3. Develop a locally-optimized RDM estimation tool with multi-sensor imagery. 

 Recent work on variability and loss in RDM over the summer dry period (Larsen et al., 2020) suggests important implications for using remote sensing estimates of both growing and non-growing season above-ground biomass. To map RDM operationally during the dry season, we will develop an innovative and scalable approach using locally optimized machine learning algorithms. Our approach involves two steps:  (1) scaling up the field measured RDM with unique hyperspectral measurements from PRISMA imagery acquired over selected large areas, and (2)  wall to wall cost-effective RDM mapping with the calibrated machine earning models to integrate improved growing season production estimates, factors related to decomposition, and coincident Landsat/Sentinel imagery. 

During Year 1, we focused on preparing for developing a calibrated remote sensing algorithm to estimate RDM using PRISMA hyperspectral imagery at 30-m resolution. We submitted tasking order  to acquire PRISMA imager from August 2022 to to cover our monitoring sites. A total of 10 scenes were acquired at a 30 × 30 km swath. Work is on the way to explore the relationship between cellulose absorption features at ~2100 nm and many other spectral features, such as Cellulose Absorption Index and the traditional broadband indices (NDRI) with field measured RDM. Once more field data is collected this fall,  a statistical model will then be built  to identify the most important hyperspectral features and indices for RDM estimation across sites and estimate RDM with PRISMA data. The calibrated model will then be applied to all herbaceous-dominated 30-m pixels to map RDM for each PRISMA scene. We will build a separate model to incorporate only those broadband indices available from Landsat and Sentinel for RDM estimation, and evaluate the improvement of PRISMA-based RDM method over existing approaches. 

Next, in order to upscale the PRISMA approach for operational RDM estimates across the whole state, we will utilize a machine learning approach to integrate forage production in the growing season (Objective 2) with the fall-season Landsat/Sentinel imagery, preceding weather, solar energy, and topography for a more robust RDM estimation. This will leverage the large RDM datasets generated above from PRISMA as training data. Models will be developed for each individual region to account for potential impacts of regional variations in soils, species and decomposition rate. We will estimate RDM for all grass-dominated Sentinel pixels and assess the accuracy with ground measurements.

Research Objective 4. Prototype an online interactive app to help stakeholders improve ranch management and risk mitigation decision-making.

We brainstormed the designing ideas for the API functions and dashboard. While waiting for the field data to be completed, we also assessed the historical annual forage production maps (2001-2018 from Landsat imagery at 30m) produced from our previous studies to quantify and understand potential trend of year to year forage production variability, an important aspect related to ranch management and risk mitigation decision-making.  The coefficient of variation (CV), defined as the ratio of the standard deviation divided by the mean,  was used to represent the interannual variability in forage production as well as other vegetation and climatic factors. The 8-year moving window from 2001 to 2018 was applied to reduce the uncertainties caused by the disturbances such as drought. We applied the Mann-Kendall and the Theil-Sen methods to test the significance of  and quantify the trends of the variability of forage production, for each 30m pixel, and then summarized over each sub-ecoregion. To further  understand the trends of variability in forage production, the random forest model was built using a suit of remote sensing indices, climate indices as well as soil and topographical indices were used as predictors. Around 20,000 samples, of which the minimum distance between the points was set as 300 meters, were randomly selected for model building, due to the extremely large samples at 30m. For the random forest modeling, 70% of the sampling points were used for training models, and the left 30% of the points were used for validation. The Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and Variance Explained by the independent variables were used to quantify the accuracy of the model. The performance of the input variables was evaluated by the percent increase in Mean Square Error (%IncMSE). The partial dependence plot was used to quantify how the trends of variations of forage production vary with each independent variable. 

We will be prototyping an easy to use API  accessible in both desktop and mobile devices, to enhance data access and interactive visualization of near real-time monitoring results. All of the calibrated algorithms for statewide applications will be implemented with Python API on Google Earth Engine (GEE). This workflow automatically ingests the imagery from Landsat series at 30m every 8 days and Sentinel 2 at 10m every 5 days, weather data, as well as the less frequent NAIP imagery at 0.6m to 1m, and topography all available for free on the GEE platform. The imagery stream will be further  augmented with available Planet 3m imagery via Planet and GEE cloud integration tools (via Jin’s Ambassador partnership with Planet). Imagery fusion and routine estimates of growing season forage production will be produced at 10m resolution on the fly and updated every 3 to 5 days. The RDM estimates will be generated every 10 days in late season, using the calibrated ML algorithms to integrate the growth curve, and preceding weather, concurrent multispectral imagery, and topographical attributes.

We will work closely with ranchers, local UCCE advisors, and the Advisory Group to co-produce the interactive app, which will guarantee the ultimate product will serve the specific needs of California rangeland managers. A user-friendly dashboard will be designed to allow users to query, visualize, and summarize forage production and RDM at various scales. Various tabs will also permit users to examine spatial heterogeneity within and across management units, which will aid them in designing optimized livestock management and grazing strategies across the landscape. By combining with our previous historical daily forage production estimates at 30m resolution (Liu et al., 2020), end users can visualize deviations from average production (anomalies) at ranch, region, or state-wide scales, and make adaptive decisions. We will also incorporate the machine learning based analysis and models on site-specific forage production sensitivity to weather (Liu et al., 2023). The RDM monitoring will assist producers and land managers to evaluate how closely they are achieving RDM objectives before the end of the season, and adapt their grazing activities accordingly. Examples include determining how many grazing days they would have if they turned cattle into a pasture mid-summer.

We expect this interactive app will profoundly affect in-season on-ranch decision-making. Having access to near-real-time forage production data during the growing season will allow producers who cannot regularly access lands to make critical decisions regarding stocking rates, pasture rotations, and shipping dates with substantially greater precision and accuracy. Additionally, all producers will benefit from features allowing them to query forage production data by date in order to compare production across months and years. Impact and effectiveness of the interactive app will be regularly evaluated in collaboration with project Advisory Group and via extension education workshops (see Education Plan and Evaluation and Producer Adoption sections).