Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Is Cambodia the World's Largest Cashew Producer?

Veasna Chaya, Ate Poortinga, Keo Nimol, Se Sokleap, Mon Sophorn, Phy Chhin, Andrea McMahon, Andrea Puzzi Nicolau, Karis Tenneson, David Saah

TL;DR

This study addresses the lack of high-resolution, space-based data on Cambodia's expanding cashew plantations and their environmental implications. It introduces a multisource deep learning pipeline based on an attention-enhanced U-Net to fuse PlanetScope, Sentinel-1/2, and Landsat imagery, trained on a large ground-truth polygon dataset to produce probabilistic cashew maps, which are integrated with a GLAD forest-loss dataset to assess historical forest change. The results place Cambodia among the top three global cashew producers by area and yield, with an estimated 580,117 ha of cashew and production around 903,598 tons, concentrated in Kampong Thom, Kratie, and Ratanak Kiri, while highlighting the challenge of detecting young orchards from space. The work underscores the need to balance rapid agricultural expansion with forest conservation and local value addition, and demonstrates how advanced mapping can inform policy and sustainable land-use planning. Open-source code and clear data-access pathways are provided to enable ongoing monitoring and verification.

Abstract

Cambodia's agricultural landscape is rapidly transforming, particularly in the cashew sector. Despite the country's rapid emergence and ambition to become the largest cashew producer, comprehensive data on plantation areas and the environmental impacts of this expansion are lacking. This study addresses the gap in detailed land use data for cashew plantations in Cambodia and assesses the implications of agricultural advancements. We collected over 80,000 training polygons across Cambodia to train a convolutional neural network using high-resolution optical and SAR satellite data for precise cashew plantation mapping. Our findings indicate that Cambodia ranks among the top five in terms of cultivated area and the top three in global cashew production, driven by high yields. Significant cultivated areas are located in Kampong Thom, Kratie, and Ratanak Kiri provinces. Balancing rapid agricultural expansion with environmental stewardship, particularly forest conservation, is crucial. Cambodia's cashew production is poised for further growth, driven by high-yielding trees and premium nuts. However, sustainable expansion requires integrating agricultural practices with economic and environmental strategies to enhance local value and protect forested areas. Advanced mapping technologies offer comprehensive tools to support these objectives and ensure the sustainable development of Cambodia's cashew industry.

Is Cambodia the World's Largest Cashew Producer?

TL;DR

This study addresses the lack of high-resolution, space-based data on Cambodia's expanding cashew plantations and their environmental implications. It introduces a multisource deep learning pipeline based on an attention-enhanced U-Net to fuse PlanetScope, Sentinel-1/2, and Landsat imagery, trained on a large ground-truth polygon dataset to produce probabilistic cashew maps, which are integrated with a GLAD forest-loss dataset to assess historical forest change. The results place Cambodia among the top three global cashew producers by area and yield, with an estimated 580,117 ha of cashew and production around 903,598 tons, concentrated in Kampong Thom, Kratie, and Ratanak Kiri, while highlighting the challenge of detecting young orchards from space. The work underscores the need to balance rapid agricultural expansion with forest conservation and local value addition, and demonstrates how advanced mapping can inform policy and sustainable land-use planning. Open-source code and clear data-access pathways are provided to enable ongoing monitoring and verification.

Abstract

Cambodia's agricultural landscape is rapidly transforming, particularly in the cashew sector. Despite the country's rapid emergence and ambition to become the largest cashew producer, comprehensive data on plantation areas and the environmental impacts of this expansion are lacking. This study addresses the gap in detailed land use data for cashew plantations in Cambodia and assesses the implications of agricultural advancements. We collected over 80,000 training polygons across Cambodia to train a convolutional neural network using high-resolution optical and SAR satellite data for precise cashew plantation mapping. Our findings indicate that Cambodia ranks among the top five in terms of cultivated area and the top three in global cashew production, driven by high yields. Significant cultivated areas are located in Kampong Thom, Kratie, and Ratanak Kiri provinces. Balancing rapid agricultural expansion with environmental stewardship, particularly forest conservation, is crucial. Cambodia's cashew production is poised for further growth, driven by high-yielding trees and premium nuts. However, sustainable expansion requires integrating agricultural practices with economic and environmental strategies to enhance local value and protect forested areas. Advanced mapping technologies offer comprehensive tools to support these objectives and ensure the sustainable development of Cambodia's cashew industry.
Paper Structure (13 sections, 5 figures, 4 tables)

This paper contains 13 sections, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Training data
  4. Neural Network model
  5. Model Training
  6. Inference
  7. Validation and area estimation
  8. Forest loss dataset
  9. Results
  10. Discussion
  11. Conclusion
  12. Acknowledgements
  13. Data and Model Availability

Figures (5)

  • Figure 1: Spatial distribution of the all the different categories of training data.
  • Figure 2: The U-net attention architecture utilizing data from PlanetScope, Sentinel-2, Sentinel-1, and Landsat. PlanetScope data is integrated at the highest level of the network, while Sentinel-2 and Sentinel-1 bands are incorporated post-maxpooling operations.
  • Figure 3: The 2023 cashew probability map for Cambodia.
  • Figure 4: Forest loss per cashew probability category annually, normalised to 100%; the blue line (top axis) shows cashew-related forest loss as a percentage of total yearly loss from 2000 to 2022.
  • Figure 5: Cashew age distribution in main cashew producing areas.