Use of earth observation satellite data to guide the implementation of integrated pest and pollinator management (IPPM) technologies in an avocado production system

Abstract

Insect pollinators provide an important ecosystem service by improving agricultural productivity. However, their populations have been declining in recent years due to excessive use of synthetic pesticides, climate and land use/land cover (LULC) changes. Climate and LULC changes have resulted in land fragmentation and consequently pollinator habitat loss. To conserve pollinators, there is a need for sustainable agricultural practices such as integrated pest and pollinator management (IPPM), which is a holistic landscape management approach that minimizes pesticides use while conserving pollinator abundance and diversity. This study aimed to use earth observation (EO) data to characterize landscape dynamics in terms of vegetation productivity to guide the implementation of IPPM interventions in an avocado production system in Murang'a (Kenya). Specifically, we utilized Sentinel-2 (S-2)-derived normalized difference vegetation index (NDVI) as a proxy for vegetation productivity to assess IPPM implementation sites. The NDVI was calculated using multi-date S-2 data acquired during the dry and wet seasons and categorized into three vegetation productivity classes - low, medium, and high - using a K-means unsupervised clustering method. We also collected socio-economic baseline data from 410 farmers with a focus on their perception to implement one of four avocado pest management and pollination options: (1) IPPM, (2) integrated pest management (IPM), (3) pollinator supplementation (P), and (4) no intervention (control). The three landscape vegetation productivity classes were then linked with the four farmer preferences with regards to the implementation options. Criteria based on the distances among the sites for implementing the different four options were set for farmer selection and the experiment was replicated three times in each vegetation productivity class (i.e. in total 12 farmers in each class). The results showed that the K-means method was successful in characterizing the landscape vegetation productivity with an overall accuracy of 86.2%. One the other hand, we successfully selected the 36 (12 in each of the 3 vegetation productivity classes) out of 410 farmers who met our distance-based criteria and participated in the implementation of one of the four technology options (i.e. IPPM, IPM, P, and control). In conclusion, NDVI proved to be a vital proxy for assessing the landscape dynamics as it provided a robust view of vegetation productivity patterns, which enabled a well-representative distribution of the four pest management and pollination options across the landscape. Overall, the study shows the utility of integrating EO and socio-economic data in selecting sites for implementing agro-technologies at a landscape scale