Land use/land cover (LULC) change and irrigated area monitoring in Eritrea : insights into horticultural production and sustainability

Abstract

In arid and semi-arid regions, where water is scarce and climatic variability is high, monitoring changes in irrigated land is essential for ensuring food security and building resilience. However, few studies have assessed irrigation dynamics in the Horn of Africa using remote sensing, and empirical data from Eritrea remain limited. This study investigates the spatio-temporal dynamics of irrigated agriculture in two contrasting regions of Eritrea, Dighe and Gala Nefhi, using multi-temporal Sentinel-2 imagery and Supporting climatic and agricultural datasets from 2015 to 2024. It aims to map the spatial distribution of irrigated fields, assess their changes over time, and examine relationships with rainfall variability, horticultural crop production, and market fluctuations by comparing trends throughout the study period. A supervised Random Forest classification approach was implemented in Google Earth Engine, incorporating spectral indices and post-classification comparison to quantify the Land Use/Land Cover (LULC) transitions. The classification was based on dry-season imagery to distinguish irrigated from rainfed areas, with seven LULC classes identified. Overall classification accuracy ranged from 0.82–0.86 in Dighe and 0.87–0.89 in Gala Nefhi, with Kappa coefficients of 0.70–0.81 and 0.85–0.86, respectively. Results show a 115.5% increase in irrigated area in Dighe and 65.6% in Gala Nefhi. While Gala Nefhi showed synchronized growth in irrigation and horticultural crop production, Dighe exhibited inconsistent yields despite expanded irrigation. The study shows that expanding irrigation alone cannot increase production without reliable water sources, favorable climate conditions, and institutional support.