A GIS framework to demarcate suitable lands for combine harvesters using satellite DEM and physical properties of soil

Authors: RAHMAN, MIZANUR - Bangladesh Rice Research Institute; KAMRUZZAMAN, MOHAMMAD - Bangladesh Rice Research Institute; SHAHID, SHAMSUDDIN - University Technology Malaysia; Thorp, Kelly; RAHMAN, HAFIJUR - Bangladesh Rice Research Institute; SHAHRIYAR, MAHIR - Bangladesh Rice Research Institute; ISLAM, SAIFUL - Bangladesh Rice Research Institute; HUDA, DURRUL - Bangladesh Rice Research Institute

Submitted to: Journal of Geovisualization and Spatial Analysis
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/24/2023
Publication Date: 9/11/2023
Citation: Rahman, M., Kamruzzaman, M., Shahid, S., Thorp, K.R., Rahman, H., Shahriyar, M., Islam, S., Huda, D. 2023. A GIS framework to demarcate suitable lands for combine harvesters using satellite DEM and physical properties of soil. Journal of Geovisualization and Spatial Analysis. 7(27):1-18. https://doi.org/10.1007/s41651-023-00156-y.
DOI: https://doi.org/10.1007/s41651-023-00156-y

Interpretive Summary: Mechanization of rice harvesting can reduce food insecurity in Bangladesh. However, some land areas are more suitable for mechanized rice harvesting than others. This study used data from soil samples and satellite sensors to develop a classification map of land suitability for combine harvesters in Bangladesh. The results will guide Bangladeshi farmers and policymakers to improve rice production efficiency in the country.

Technical Abstract: Rice harvesting in Bangladesh is impacted by the absence of advanced harvesting technologies, high labor costs, and natural calamities, which frequently interrupt the harvesting schedule. Mechanized harvesting methods, such as combine harvesters, could enable large-scale, efficient harvesting with reduced labor dependency. However, the use of such machinery is complicated by the varying and limited size of rice fields across the country. This research aimed to develop a suitable land classification map for combine harvester operations using satellite-derived digital elevation models (DEM) and soil physical property datasets at Kalikoir, Gazipur, Bangladesh, which will identify the most suitable rice fields for quick and efficient harvesting. The study considered eight thematic layers for developing the model, including sand, silt, clay, soil bulk density, soil moisture, dry density of soil, water holding capacity, and slope. The relative weight of selected layers was determined using the Extra Tree Classifier machine learning algorithm within the Jupiter Environment. The land classification map was subsequently generated using a weighted overlay analysis technique within the ArcGIS environment. The resulting map revealed that 1.88 km² (19.84%) was highly suitable for combine harvester use, 4.16 km² (43.53%) was moderately suitable, 2.67 km² (27.43%) had limited suitability, and 0.86 km² (9.19%) had very limited suitability. The classification map was validated using a ground truth dataset with several performance metrics, including overall accuracy, precision, recall, F1 score, and threat score. The model demonstrated robust performance with an overall accuracy of 71%, precision of 85%, recall of 79%, F1 score of 81%, and threat score of 69%. A further assessment of accuracy using Area Under Curve (AUC) measures indicated a 60% success rate. The results provide valuable and precise insights that can benefit commercial combine harvester users, farmers, and policymakers, helping to identify optimal rice-harvesting locations in Bangladesh. This in turn can support more effective resource allocation, reduce costs, and ultimately enhance rice production yield in the country.