Map teff in Ethiopia: An approach to integrate time series remotely sensed data and household data at large scale

It is important to know cropland distribution and allocation of crop types at finer resolution in order to monitor and make plans for agricultural resource management especially in developing countries where smallholders are more vulnerable to food insecurity and poverty. Teff is one of the most staple crops in Ethiopia and it is estimated that it made up to 20% of all cultivated area in Ethiopia. This research will focus on developing a new approach to produce accurate and reliable teff distribution estimates in Ethiopia using time series remotely sensed data and household survey data. The production information from household survey and census usually are collected at subnational level and it is hard to know how the crops are distributed within the administrative unit. On the other hand, time series remote sensing data can be applied successfully in land cover classification studies but it is usually hard to identify crop types and distribution at pixel level without high quality ground truthing datasets which is very costly to acquire in at large scale in developing countries. The proposed method investigates the possibility of combining both information from household crop information (e.g. production, area) and NDVI time series in order to improve the crop allocation at finer resolution at large scale. The method developed in this study shows that available household data and satellite imagery can be used in mapping teff area. Combing information from different sources, the teff density map could be produced at higher resolution.