Irrigation Monitoring Using Geospatial Techniques in Plastic Greenhouse Landscapes: Case Study – Chtouka Plain, Morocco

Abstract

The practice of under plastic agriculture is a key economic driver in Chtouka plain in Morocco and plays an important role in agriculture systems. The spatio-temporal information of PGH is still not available countrywide neither at regional nor local scales. This information is required for agriculture and water managers as well as environmental agencies, especially since it represents the first pressure on groundwater resources, which knows a deficit of 60 to 80 million m3 per year. This chapter represents an optimized workflow for monitoring irrigation activities in a plastic greenhouses landscape, where machine learning techniques are used to build a model for automatic extraction of plastic greenhouses as well as irrigated areas in open fields using time-series Sentinel-2 imagery. The model was tested and validated firstly on one acquisition (29th July 2019) and then run on the cloudless time-series Sentinel-2 images for the period (August 2019-August 2020) at a monthly basis. The overall accuracy achieved exceeds 97%, and kappa coefficient was around 98%.