Mapping Diversification Practices on the Landscape

Mapping diversification practices

Kangogo Sogomo & Tim Bowles, UC Berkeley

 

  • Background context on the activity

Changes in agricultural production practices are one element of agroecological transitions (Gliessman et al., 2018).  In particular, incorporating “planned biodiversity”—crop and non-crop vegetation and animals at multiple scales—is considered the key for supporting biodiversity and ecosystem functions that in turn allow for reductions in inputs and other agroecosystem benefits (Altieri, 1999; Kremen et al., 2012; Tamburini et al., 2020). Thus, one aspect of understanding where agroecological transitions are happening or aren’t happening is to map where and to what extent farms are applying such diversification practices. This understanding complements the mapping of organic agriculture, since conventional and organic systems vary widely in their use of diversification practices (Carlisle et al., 2022). For example, farm fields may be managed organically but with little use of the diversification practices that would allow for a greater degree of ecosystem functions like nutrient cycling and retention, pest and disease control, and water conservation.

Previous research from the Berkeley Agroecology Lab mapped cover crops and hedgerows in the northern central coast, namely Monterey, Santa Cruz, and San Benito Counties (Thompson et al., 2023), finding that only ∼6% of farmland had winter cover crops in 2021 and 0.26% of farmland had hedgerows or windbreaks in 2018. Leveraging hyperspectral and multispectral remote sensing data sources such as satellites and drones presents an opportunity to identify field practices such as cover crops but also foliar biochemical contents such as nitrogen, water and canopy-level indicators of carbon. Therefore expanding understanding of cover crop trends, low  adoption and high adoption areas and nutrient cycling benefits occurring across the state per time.

 

  • What is the current state of the project

The first phase of our project with COAR focuses on three objectives:

  • Objective 1: Map field level cover crop presence and absence
    • Spatial scope: Santa Cruz, San Benito and Monterey counties
    • Temporal scope: 5 years
    • Data Sources Identified:
      • DWR crop map, Fields of the world (FTW) field boundary data, Continuous Living Cover (CLC) 
    • Work in Fall 2025 includes:
      • Analyze crop maps from central coast (Fig 1)
      • Determine the cover crop signal trend
  • Objective 2: Estimate the nitrogen content available in plant biomass as a proxy for water quality protection. Work in Fall 2025 includes:
    • Literature review of technical methods for estimating nitrogen content via Remote Sensing in Agroecosystems  (Kangogo)
    • Meeting with regional policy stakeholders (Kangogo and Tim)
    • Collect and solicit ground truth data (Kangogo, Tim, and BAE lab)
    • Evaluating cover crop nitrogen calculator as an alternative to estimating nitrogen content directly (Kangogo)
  • Objective 3: Map continuous living cover (CLC) across the northern central coast and expand to the entire state. Work in Fall 2025 includes:
    • Connect with preexisting work on CLC from Data Science and Environment (Kangogo)
    • Develop productivity scores based on CLC to support meaningful field level agroecological assessment
    • Investigate potential funding options for continuing this work

 

  • The direction the work is planned to go

We would like our work to become part of a more integrated, quantitative / top-down and qualitative / bottom-up, research project with others on the CATs team, for example with adapting the TAPE tool (Amelie et al.) and characterizing agroecological typologies (Crystele et al.). We would like to devise this project together with COAR collaborators. 

Leveraging hyperspectral and multispectral remote sensing data sources such as satellites and drones presents an opportunity to identify field practices such as cover crops, intercropping, crop rotation but also foliar biochemical contents such as nitrogen, water and canopy-level indicators of carbon. Furthermore, integrating remote sensing of foliar biochemical traits, such as leaf water status, and canopy-level indicators of carbon, can improve our understanding of how field management practices influence key hydrological processes, including soil water retention, runoff, and water quality. More than likely these would both be pending additional funding support from other grants.

 

  • With whom the work will engage and how
  • Technical partners
    • Center for Data Science and Environment, UC Berkeley
      • Partner for continuous living cover
    • Guzman Lab, Stanford University
      • Potential partner for spatial crop diversity
  • Policy advocates
    • Sustainable Conservation
    • CalCAN
    • EJ organizations focused on water quality
  • State agencies
    • Central Coast Regional Water Quality Control Board 
      • Mapping cover crop biomass and nitrogen content is related to Ag Order 4.0 policy for water quality
    • CARB
      • Mapping practices is relevant to state’s goals in the Climate Scoping Plan
    • CDFA
      • Mapping practices is relevant to goals for regenerative agriculture

  • Farmers

    • Thoughts on data privacy
    • Interpreting remote sensing data in the context of actual farm management

     

  • Methods for a general audience

Varies by practice:

  1. Mapping cover crops: 
    1. Machine learning based methods leveraging the multispectral and hyperspectral remote sensed sources i.e Sentinel, drone - strong focus on Red Edge, Red, NIR bands for identification
    2. Determination of field level plant and harvest dates of cover crops to guide biomass estimation useful in estimating nitrogen content
  2. Mapping plant available nitrogen content in cover crops:
    1. Data sources similar to above and Thermal: ECOSTRESS
    2. By using chlorophyll-nitrogen spectral related indices from Red Edge and NIR  we are able to more reliable phenology related information in dense, mature canopies where NDVI saturates 
    3. Ground truth estimations from cover crop nitrogen calculator and field plot experiments occurring in Santa Cruz
    4. A machine learning based model used to map 
  3. Continuous living cover (CLC): 
    1. DSE developed a continuous living cover dataset using NDVI for the 3 counties of interest
    2. Expand the CLC work to field level analysis to develop productive indicators that inform soil health benefits in the agroecological framework.
    3. The goal is to expand it to the rest of the state.
  4. crop rotational diversity
    1. temporal patterns of crops derived from DWR crop maps to develop methods to identify rotational complexity linked with remote sensing data.

 

  • Timeline of activities and proposed deliverables

F25 - S26: Pilot Phase in 3 counties in Central Coast 

  1. Collect field-level satellite/drone data for cover crops and foliar nitrogen across San Benito, Santa Cruz, and Monterey
  2. Process 5 years of historical imagery to generate: Cover crop presence/absence maps, Nitrogen content maps (e.g., NDRE, CI-RedEdge)
  3. Hypothesis and test Continuous Living Cover (CLC) dataset for field level analysis for productivity indicators meaningful for soil health understanding.
  4. Solicit ground truth data from growers (rolling basis)
  5. Begin compiling socio-economic factors (e.g. acreage, proximity to cities)
  6. Initiate exploratory hypothesis development based on observed adoption patterns 

Deliverables

  1. Baseline geospatial datasets: cover crop and nitrogen maps for 3 counties (5-year archive)
  2. Preliminary data-driven categories based on socio-economic indicators
  3. Theories for field level CLC compilation.
  4. Present work at European Geosciences Union (EGU) 

F26 - S27: Central Valley Expansion & Hydrological Integration

Activity

  1. Extend field-level mapping (cover crops + nitrogen) to additional Central Valley counties
  2. Acquire and integrate ECOSTRESS or other thermal/SWIR data to estimate: Water retention potential, Evapotranspiration (ET) as a proxy for soil moisture
  3. Map groundwater recharge indicators (e.g. ET surplus)
  4. Analyze opportunity for high resolution data to support identification of intercropping and crop rotation practices and how they compare with biochemical contents such as nitrogen and water 

Deliverables

  1. Expanded geospatial coverage: nitrogen and cover crop maps for >10 counties
  2. Analytical report: linking field practices to hydrological function
  3. 1 Peer-reviewed publications
  4. Presentation at European Geosciences Union (AGU) 

F27 – S28: Full-State Mapping and Synthesis

Activities

  1. Complete mapping for remaining California counties
  2. Refine machine learning models using entire state dataset
  3. Integrate multi-year socio-economic trends and policies
  4. Conduct mixed-methods analysis: combine geospatial and qualitative data

Deliverables

  1. Statewide, high-resolution maps of cover crops and nitrogen content
  2. Classification schema of agroecological transitions by landscape
  3. Policy-ready synthesis report connecting practices to carbon, nitrogen, water outcomes

** Ground truth data from California growers will be solicited on a rolling basis, and collaborators are welcome to contribute data or establish connections at any point during the project timeline.

 

  • Tensions and challenges

Sources of information for ground truth data for mapping cover crops and nitrogen content is essential but challenging to access. We are interested in grower connections to develop a high quality and representative ground truth dataset for model training.

The COAR project has ambitious goals and is essential for bringing together researchers across California to work on a project for the first time together. It is a very exciting time for agroecology in California! Tim feels a tension between devising and proposing work that could meaningfully contribute to the ambitious goals of the project and what is possible for the level of resources the COAR grant directly brings, which for this project is four semesters of GSR support and a little summer salary. 

 

  • How will the project support addressing historical injustices & current structural inequities?

There are at least two plausible possibilities. First is identifying where winter cover cropping is or isn’t being used and increased in order to protect groundwater quality, the pollution of which disproportionately affects rural people of color. This in turn could help identify the effectiveness of water quality regulations. Second is understanding how the adoption or lack of adoption of diversification practices relates to structural factors like land values. For example, in prior mapping work we found that cover crops and hedgerows were more likely to be adopted in hillier areas of Santa Cruz Co. but not the higher value flatlands. This would be well-paired with qualitative research.