LAUSR

Abstract Invasive plant species in eastern Africa severely impede rangeland and cropland productivity with dire consequences for livelihoods of agro-pastoralist communities. We produced the first occurrence and spread map of invasive plant species (Prosopis: Prosopis juliflora and Parthenium: Parthenium hysterophorus) for western Somaliland (a region of eastern Africa) using vegetation productivity and phenology trends from 250 m MODIS (Moderate Resolution Imaging Spectroradiometer) EVI (Enhanced Vegetation Index) time-series data (2001–2014). Binomial logistic regression models were created to predict the presence or absence of the invasive species from the MODIS EVI phenometrics and vegetation productivity trends. Model training pixels were extracted from a 30 m Landsat-based classification that mapped areas of propagation of the two invasive species between 2001 and 2015. Field observations collected during 2014 and 2015 were used as reference data for the Landsat classification. After optimization of the logistic regression models, a probability of occurrence map was produced and evaluated for each of the two invasive species. The probability maps predicted that the cropland-dominated areas in the southwestern part of Somaliland were considerably infested with Parthenium while Prosopis was most abundant in the peri-urban zones and the central and eastern regions. Vegetation amplitude (the seasonal cycle of vegetation between the vegetation peak and the trough) was most relevant and statistically significant for predicting the spread of Parthenium. This highlights the importance of vegetation seasonality variables for the wide-area mapping of herbaceous life forms in semi-arid biomes. Mann-Kendall trends based on annual summed EVI value and seasonal EVI peak value trends were the most relevant predictors for the occurrence of Prosopis. Phenometric trends show immense potential to map shifts in vegetation patterns in relation to the spread of invasive species as a consequence of global change effects, particularly in African drylands.