Application of Geospatial Methodologies to Elucidate Honey bee (Apis mellifera) Colony Loss

Abstract

Geospatial methods can be used in a wide variety of applications to elucidate interactions among variables across space and time and at various spatial scales. The identification of spatial trends is a foundational element of geospatial methods and can be determined through the use of spatial analysis. Honey bees (Apis mellifera) are a unique species for exploring geospatial methods as they are pervasive across the United States, have seen increasing losses in the past decade, and have complex interactions with their environment which are difficult to measure and understand across space and time. This thesis will explore how geospatial methods and spatial analysis assists in elucidating factors that may be leading to honey bee colony loss at various scales. Chapter one of this thesis provides an overview of the main geographical, apicultural, and statistical concepts needed to understand the methods developed for this research. Specifically, spatial autocorrelation as measured using spatial statistics including the global and local Moran’s I, honey bee colony loss as driven by Varroa destructor, and non-invasive colony inspection completed through thermal image capture. Chapter two of this thesis aims to understand whether the abundance of Varroa destructor, a parasitic mite, shows clustering patterns at the national and regional scales using spatial statistics. Varroa destructor has been found to cluster spatially in New Zealand and Argentina, however spatial trends of V. destructor have not been investigated in the United States. Based on case studies in other countries, it was hypothesized that V. destructor spatial clustering will also be present in the United States. Spatial autocorrelation of V. destructor abundance was calculated using the global Moran’s I and local Moran’s I statistics, from both manual and automated tasks in GIS. The results showed that V. destructor was spatial clustered in the United States and that the automated combination of the global and local Moran’s I artificially inflates the detection of spatial clusters, and thus the manual calculation of the global and local Moran’s I may be more appropriate. Chapter three investigates the use of a thermal sensor on a drone to capture the average colony temperature and compare those temperatures with colony health metrics (number of adult bees, amount of brood, and amount of honey). I hypothesized that the sensor would capture the most accurate reading of average colony health with both lids removed and that the amount of brood and honey would significantly contribute to the average colony temperature. Data were collected in November 2020 and health metrics included the number of frames of adult bees, brood, honey, and the proportion of adults, brood, and honey in the top brood box. Thermal images were taken above 15 colonies with the outer and inner lid in place, the inner lid in place, and no lids. Backwards stepwise model building was used to identify variables with statistical significance as indicated by the p-value. Results indicate that the removal of the outer lid is most appropriate as it allows for the estimation of more honey bee health metrics and that a drone equipped with a thermal sensor can be used to conduct non-invasive colony health inspections.