Landscape fragmentation and habitat loss at multiple scales directly affect species abundance, diversity and their productivity. There is a paucity of information about the effect of the landscape structure and diversity on honeybee colony strength in Africa. Here, we present new insights into the relationship between landscape metrics such as patch size, shape, connectivity, composition and configuration and honeybee (Apis mellifera) colony strength characteristics. Remote sensing-based landscape variables were linked to honeybee colony strength variables in a typical highly fragmented small holder agro-ecological region in Kenya. We examined colonies in six sites with varying degrees of land degradation, during the period from 2017 to 2018. Landscape structure was first mapped using medium resolution bi-temporal Sentinel-1 and Sentinel-2 satellite imagery with an optimized random forest model. The influence of the surrounding landscape matrix was then constrained to two buffer distances i.e., 1 km representing the local foraging scale and 2.5 km representing the wider foraging scale around each investigated apiary and for each of the six sites. The results of zero-inflated negative binomial regression with mixed effects showed that lower complexity of patch geometries represented by fractal dimension and reduced proportions of croplands were most influential at local foraging scales (1 km) from the apiary. In addition, higher proportions of woody vegetation and hedges resulted in higher colony strength at longer distances from the apiary (2.5 km). Honeybees in moderately degraded landscapes demonstrated the most consistently strong colonies throughout the study period. Efforts towards improving beekeeper livelihoods, through higher hive productivity, should target moderately degraded and heterogeneous landscapes, which provide forage from diverse land covers. 

Standard Langstroth hives were used for this study. Each of the hives were labeled randomly from 1 to 10 in each apiary and subsequently the frames in the hives labeled 1 to 10 with each side of the frame labeled A or B. Initially, 60 hives were setup for this study, i.e., 10 hives per site. However, throughout the study period, we only inspected and sampled 30 hives (i.e., colonies) which were the only ones occupied by natural swarms as observed during our first field data collection. Visual estimates from two observers were used, as this method is less disruptive compared to using empirical measurements such as weight of the honeybees (Delaplane et al, 2013a). All the data measurements were conducted during the early hours around 07h00 and 08h00 local time (Greenwich Mean Time: GMT + 3) of the day to control for the foraging activities of honeybees, which might affect the observations, especially of the adult honeybee population. Each of the occupied hives was lightly smoked, opened and frames containing combs were sequentially removed and examined. The estimated percentage coverage of each frame side by the target resource (adult population, sealed and open brood, eggs, honey and pollen) was visually done. Honeybee colony strength was estimated following Delaplane et al. (2013) and Imdorf and Gerig (2001) Liebefeld protocol. 

In order to generate the landscape fragmentation indices, a honeybee habitat map was generated from a fused bi-temporal Sentinel-1 and a single season Sentinel-2 dataset ((European Space Association (ESA), 2017), which had an overall accuracy of 86% (Ochungo et al., 2019), (Figure 1). Fragmentation indices were thereafter generated from the habitat map at various spatial scales ranging from 0.5 km to 3 km.