SHORT-TERM NDVI FORECASTING USING MACHINE LEARNING: A CASE STUDY OF OLIVE-GROWING REGIONS IN MEKNES, MOROCCO

Olive farming is the backbone of Moroccan agriculture and is closely related to the Meknes region,
with productivity and sustainability closely associated with vegetation. NDVI is regarded as a good
indicator of crop vigor, but there are challenges in short-term forecasting and management owing
to changing climatic conditions and variance. This study examined the feasibility of predicting the
mean NDVI of olive orchards in Meknes 15 days in advance using meteorological data from 2012
to 2022, coupled with NDVI derived from satellite observation (2012–2022), through regression
based machine learning models: Linear, Polynomial, and Ridge regression, as well as exploratory Kmeans clustering to characterize vegetation–climate regimes. Ridge regression proved to be the best,
outperforming linear and polynomial models. Cluster analysis revealed distinct NDVI states resulting
from two seasons: growth and summer stress conditions. These results demonstrate the ability of
simple machine learning methods to produce reliable short-term NDVI forecasts, which could be
useful decision-support tools for irrigation management, early stress detection, and sustainable olive
cultivation in semi-arid Mediterranean environments.