High-resolution crop-type mapping in northern Ghana
Reliable crop information is crucial for monitoring food security and agricultural growth, especially when providing extension and advisory services to smallholder farmers in Africa. Timely and accurate data on the extent of croplands and the types of cultivated crops are essential to addressing the...
| Autores principales: | , , , , |
|---|---|
| Formato: | Informe técnico |
| Lenguaje: | Inglés |
| Publicado: |
International Institute of Tropical Agriculture
2024
|
| Materias: | |
| Acceso en línea: | https://hdl.handle.net/10568/159840 |
| _version_ | 1855532807441350656 |
|---|---|
| author | Alabi, Tunrayo Muthoni, Francis Uponi, John Alabi, William Oluwaleye, Josiah |
| author_browse | Alabi, Tunrayo Alabi, William Muthoni, Francis Oluwaleye, Josiah Uponi, John |
| author_facet | Alabi, Tunrayo Muthoni, Francis Uponi, John Alabi, William Oluwaleye, Josiah |
| author_sort | Alabi, Tunrayo |
| collection | Repository of Agricultural Research Outputs (CGSpace) |
| description | Reliable crop information is crucial for monitoring food security and agricultural growth, especially when providing extension and advisory services to smallholder farmers in Africa. Timely and accurate data on the extent of croplands and the types of cultivated crops are essential to addressing the food insecurity challenges that many African countries face. This information improves crop productivity and helps mitigate production constraints such as pests, climate variability, and other biophysical factors. The precise and timely identification of crop types is necessary for effective food security monitoring and planning. This study used Unmanned aerial vehicle (UAV) and GoPro camera-derived crop labels, along with Synthetic Aperture Radar (SAR) data and PlanetScope data, to classify crops in Northern Ghana during the 2023 growing season. The classification used random forest (RF) and convolutional neural networks (CNN) algorithms. The study area covered approximately 290,000 hectares in the Northern and Upper East region of Ghana. An analysis of croplands in the Northern region revealed four primary crops: maize, soybeans, groundnut, and rice. In contrast, the Upper East region primarily cultivates millet, sorghum, maize, groundnut, and rice. The croplands in both regions are complex, mixed, heterogeneous, and patchy. Due to population pressures and the land tenure system, farming systems in the Upper East tend to be smaller and nearer to homesteads. Both algorithms demonstrated satisfactory performance in identifying the predominant crops in each region. The RF and CNN algorithms achieved an overall classification accuracy ranging from 88% to 95% and a crop-specific prediction accuracy ranging from 55% to 80% for soybean, maize, groundnut, and rice in both regions. Integrating SAR datasets significantly improved the precision of the classifications, but more importantly, it provided the capability to map crop types in all weather conditions. The results demonstrate the potential for generating annual crop maps within the study area, even though cloud cover posed challenges throughout much of the growing season. Consequently, this research provides valuable insights into the status of food production, which, in turn, facilitates food security monitoring and strategic planning. The crop type maps produced from this study will aid in targeting crop and site-specific agro-advisories in Northern Ghana. |
| format | Informe técnico |
| id | CGSpace159840 |
| institution | CGIAR Consortium |
| language | Inglés |
| publishDate | 2024 |
| publishDateRange | 2024 |
| publishDateSort | 2024 |
| publisher | International Institute of Tropical Agriculture |
| publisherStr | International Institute of Tropical Agriculture |
| record_format | dspace |
| spelling | CGSpace1598402024-11-16T02:08:24Z High-resolution crop-type mapping in northern Ghana Alabi, Tunrayo Muthoni, Francis Uponi, John Alabi, William Oluwaleye, Josiah soybean machine learning land use classification crop statistics sustainable intensification Reliable crop information is crucial for monitoring food security and agricultural growth, especially when providing extension and advisory services to smallholder farmers in Africa. Timely and accurate data on the extent of croplands and the types of cultivated crops are essential to addressing the food insecurity challenges that many African countries face. This information improves crop productivity and helps mitigate production constraints such as pests, climate variability, and other biophysical factors. The precise and timely identification of crop types is necessary for effective food security monitoring and planning. This study used Unmanned aerial vehicle (UAV) and GoPro camera-derived crop labels, along with Synthetic Aperture Radar (SAR) data and PlanetScope data, to classify crops in Northern Ghana during the 2023 growing season. The classification used random forest (RF) and convolutional neural networks (CNN) algorithms. The study area covered approximately 290,000 hectares in the Northern and Upper East region of Ghana. An analysis of croplands in the Northern region revealed four primary crops: maize, soybeans, groundnut, and rice. In contrast, the Upper East region primarily cultivates millet, sorghum, maize, groundnut, and rice. The croplands in both regions are complex, mixed, heterogeneous, and patchy. Due to population pressures and the land tenure system, farming systems in the Upper East tend to be smaller and nearer to homesteads. Both algorithms demonstrated satisfactory performance in identifying the predominant crops in each region. The RF and CNN algorithms achieved an overall classification accuracy ranging from 88% to 95% and a crop-specific prediction accuracy ranging from 55% to 80% for soybean, maize, groundnut, and rice in both regions. Integrating SAR datasets significantly improved the precision of the classifications, but more importantly, it provided the capability to map crop types in all weather conditions. The results demonstrate the potential for generating annual crop maps within the study area, even though cloud cover posed challenges throughout much of the growing season. Consequently, this research provides valuable insights into the status of food production, which, in turn, facilitates food security monitoring and strategic planning. The crop type maps produced from this study will aid in targeting crop and site-specific agro-advisories in Northern Ghana. 2024-11 2024-11-15T16:44:21Z 2024-11-15T16:44:21Z Report https://hdl.handle.net/10568/159840 en Open Access application/pdf International Institute of Tropical Agriculture Alabi et al. 2024. High-resolution crop-type mapping in northern Ghana. International Institute of Tropical Agriculture, Ibadan, Nigeria. |
| spellingShingle | soybean machine learning land use classification crop statistics sustainable intensification Alabi, Tunrayo Muthoni, Francis Uponi, John Alabi, William Oluwaleye, Josiah High-resolution crop-type mapping in northern Ghana |
| title | High-resolution crop-type mapping in northern Ghana |
| title_full | High-resolution crop-type mapping in northern Ghana |
| title_fullStr | High-resolution crop-type mapping in northern Ghana |
| title_full_unstemmed | High-resolution crop-type mapping in northern Ghana |
| title_short | High-resolution crop-type mapping in northern Ghana |
| title_sort | high resolution crop type mapping in northern ghana |
| topic | soybean machine learning land use classification crop statistics sustainable intensification |
| url | https://hdl.handle.net/10568/159840 |
| work_keys_str_mv | AT alabitunrayo highresolutioncroptypemappinginnorthernghana AT muthonifrancis highresolutioncroptypemappinginnorthernghana AT uponijohn highresolutioncroptypemappinginnorthernghana AT alabiwilliam highresolutioncroptypemappinginnorthernghana AT oluwaleyejosiah highresolutioncroptypemappinginnorthernghana |