Abstract:
The invasion by Striga in most cereal crop fields in Africa has posed an acute threat to food security and socioeconomic integrity. Consequently, numerous technological and research developments have been made to minimize and even control the Striga impacts on crop production. So far, efforts to control Striga have primarily focused on the manipulation of the genetics of the host crops, as well as understanding the phenological and physiological traits, along with the chemical composition of the weed. These initiatives have immensely contributed to the management of Striga across the continents. However, on-farm Striga control technologies require spatial explicit locational information on farms experiencing Striga occurrence and potential risk. This information affords precise and accurate intervention mechanisms and allows for the prescription of site-specific and befitting control approaches. Unfortunately, the requisite baseline information on Striga occurrence, spatial configuration, infestation extent, and intensity remain rudimentary in sub-Saharan Africa. This study, therefore, aimedto examine Striga occurrence and the potential farming areas at risk within different agroecological regions and varying climatic scenarios in Kenya and Zimbabwe. To achieve this aim, the relatively new generation remotely sensed data coupled with biophysical variables, Striga occurrence, and cropping systems data were used. Specifically, thestudy sought to establish operational spatial methodologies that can help understand and empirically determine the prospective risk posed by two of the most economically detrimental Striga species in Africa (i.e. Striga hermonthica and S. asiatica) in agroecological farming systems. In addition, the likely impacts of climate change on Striga distribution and future spread by integrating climatic and cropland data were also examined. Further, different machine learning algorithms were used for data analysisat different mapping scales. Results from this study demonstrated that Striga'soccurrence within agroecological systems can be characterized at reasonable accuracy,using relatively new generation sensor datasets across various scales of monitoring i.e. plot, field, and landscape. Comparatively, in-situ hyperspectral measurements and Sentinel-2 satellite data coupled with machine learning and subpixel classification approaches surpassed the traditional broadband sensor data in the detection and understanding of the spatial dispersion of the two Striga weed species across different agroecological farming systems. Further, the Striga flowering period was established as the most optimal period for its detection and monitoring.It was also observed that as the climatic conditions continue to change i.e. the atmospheric CO2and the temperature increase, the suitable area for Striga propagation will also increase, making more farming areas to be susceptible to a higher risk of invasion. In particular,the use of the projected climate change scenarios showed that by the year 2050, the Striga suitable area propagation will increase and spread into new areas by approximately 0.73%. Also, it was established that the ecological niche and habitat suitability assessments using multi-source remotely sensed data are fundamental in characterizing and monitoring S. asiatica occurrence and risk areas. Therefore, immediate mitigation and adaptive actions such as awareness and advocacy for the adoption of Striga control methods in the current and the future risk areas is critical to contain and manage the spread and intensity of Striga under changing climatic conditions. Overall, the findings of this study underscore the relevance of using multi-source data and machine learning algorithms for Striga weed detection and monitoring across different agroecological farming systems.