Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Embedding -based Crop Type Classification in the Groundnut Basin of Senegal

Madeline C. Lisaius, Srinivasan Keshav, Andrew Blake, Clement Atzberger

TL;DR

The paper addresses the challenge of producing accurate, wall-to-wall crop-type maps in smallholder landscapes of West Africa, where traditional satellite-based methods struggle due to small field sizes and intercropping. It evaluates embedding-based geospatial foundation models, specifically TESSERA and AlphaEarth, against baselines across four criteria: Performance, Plausibility, Transferability, and Accessibility, using data from Senegal's groundnut basin. The results show that TESSERA embeddings consistently deliver high accuracy and stable, plausible maps with strong cross-year transferability for crop-type classification, while AlphaEarth offers competitive land-cover performance but variable results for crop-type tasks and higher computational requirements. The work demonstrates that pre-generated embeddings, particularly from TESSERA, can enable practical, low-cost crop mapping in smallholder contexts, contingent on high-quality labeled data and careful task alignment, with potential for broader application in similar regions.

Abstract

Crop type maps from satellite remote sensing are important tools for food security, local livelihood support and climate change mitigation in smallholder regions of the world, but most satellite-based methods are not well suited to smallholder conditions. To address this gap, we establish a four-part criteria for a useful embedding-based approach consisting of 1) performance, 2) plausibility, 3) transferability and 4) accessibility and evaluate geospatial foundation model (FM) embeddings -based approaches using TESSERA and AlphaEarth against current baseline methods for a region in the groundnut basin of Senegal. We find that the TESSERA -based approach to land cover and crop type mapping fulfills the selection criteria best, and in one temporal transfer example shows 28% higher accuracy compared to the next best method. These results indicate that TESSERA embeddings are an effective approach for crop type classification and mapping tasks in Senegal.

Embedding -based Crop Type Classification in the Groundnut Basin of Senegal

TL;DR

The paper addresses the challenge of producing accurate, wall-to-wall crop-type maps in smallholder landscapes of West Africa, where traditional satellite-based methods struggle due to small field sizes and intercropping. It evaluates embedding-based geospatial foundation models, specifically TESSERA and AlphaEarth, against baselines across four criteria: Performance, Plausibility, Transferability, and Accessibility, using data from Senegal's groundnut basin. The results show that TESSERA embeddings consistently deliver high accuracy and stable, plausible maps with strong cross-year transferability for crop-type classification, while AlphaEarth offers competitive land-cover performance but variable results for crop-type tasks and higher computational requirements. The work demonstrates that pre-generated embeddings, particularly from TESSERA, can enable practical, low-cost crop mapping in smallholder contexts, contingent on high-quality labeled data and careful task alignment, with potential for broader application in similar regions.

Abstract

Crop type maps from satellite remote sensing are important tools for food security, local livelihood support and climate change mitigation in smallholder regions of the world, but most satellite-based methods are not well suited to smallholder conditions. To address this gap, we establish a four-part criteria for a useful embedding-based approach consisting of 1) performance, 2) plausibility, 3) transferability and 4) accessibility and evaluate geospatial foundation model (FM) embeddings -based approaches using TESSERA and AlphaEarth against current baseline methods for a region in the groundnut basin of Senegal. We find that the TESSERA -based approach to land cover and crop type mapping fulfills the selection criteria best, and in one temporal transfer example shows 28% higher accuracy compared to the next best method. These results indicate that TESSERA embeddings are an effective approach for crop type classification and mapping tasks in Senegal.
Paper Structure (20 sections, 7 figures, 13 tables)

This paper contains 20 sections, 7 figures, 13 tables.

Table of Contents

  1. Introduction
  2. Recent work
  3. Methods
  4. Criteria
  5. Labeled data preparation
  6. Remote sensing data
  7. Model training and evaluation
  8. Results
  9. Ensemble land cover classification
  10. Wall-to-wall land cover maps
  11. Temporal transfer learning for land cover classification
  12. CPU use
  13. Crop land change across years
  14. Ensemble crop cover classification
  15. Wall-to-wall crop maps
  16. ...and 5 more sections

Figures (7)

  • Figure 1: Visualized are a) the location of the region of interest within Senegal and b) Sentinel-2 satellite imagery of the region of interest. This region is dominated by dry shrub land that has been converted into agricultural land and features a river delta in the south.
  • Figure 2: Pictured are three example label polygons, not at the same scale, showing a) bare soil, b) shrub land, and c) build-up surface. There is meaningful overlap in characteristics of ground cover amongst the three, particularly with build-up surface.
  • Figure 3: Land cover maps
  • Figure 5: Crop cover maps
  • Figure 7: Diarère crop type classification map
  • ...and 2 more figures