In-season wheat sown area mapping for Afghanistan using high resolution optical and RADAR images in cloud platform

Abstract

Afghanistan has only 11% of arable land while wheat is the major crop with 80% of total cereal planted area. The production of wheat is therefore highly critical to the food security of the country with population of 35 million among which 30% are food insecure. The lack of timely availability of data on crop sown area and production hinders decision on regular grain import policies as well as log term planning for self-sustainability. The objective of this study is to develop an operational in-season wheat area mapping system to support the Ministry of Agriculture, Irrigation and Livestock (MAIL) for annual food security planning. In this study, we used 10m resolution sentinel - 2 optical images in combination with sentinel - 1 SAR data to classify wheat area. The available provincial crop calendar and field data collected by MAIL was used for classification and validation. Since the internet and computing infrastructure in Afghanistan is very limited thus cloud computing platform of Google Earth Engine (GEE) is used to accomplish this work. During the assessment it is observed that the smaller size of wheat plots and mixing of wheat with other crops makes it difficult to achieve expected accuracy of wheat area particularly in rain fed areas. The cloud cover during the wheat growing season limits the availability of valid optical satellite data. In the first phase of assessment important learnings points were captured. In an extremely challenging security situation field data collection require use of innovative approaches for stratification of sampling sites as well as use of robust mobile app with adequate training of field staff. Currently, GEE assets only contain Sentinel-2 Level 1C product which limits the classification accuracy. In representative areas, where Level 2A product was developed and applied a significant improvement in accuracy is observed. Development of high resolution agro-climatic zones map, will enable extrapolating crop growth calendars, collected from representative areas, across entire study area. While the present study shows a great potential for operational wheat area monitoring, a systematic approach for sample data collection and better understanding of cropping calendar will improve the results significantly.

Varun Tiwari

PhD Candidate

I am a Ph.D. student with the Center for Geospatial Analytics at North Carolina State University.