Date of Award
Fall 2016
Document Type
Thesis
Degree Name
Master of Science (MS)
First Advisor
Corliss, George F.
Second Advisor
Kiware, Samson S.
Third Advisor
Povinelli, Richard J.
Fourth Advisor
Merril, Stephen J.
Abstract
We explore the use of near infrared spectrometry to classifying the age of a wild malaria transmitting mosquito. In Chapter Two, using a different set of lab-reared mosquitoes, we replicate the Mayagaya et al. study of the accuracy of near-infrared spectrometry (NIRS) to estimate the age of lab-reared mosquitoes, reproducing the published accuracy. Our results strengthen the Mayagaya et. al study and increase confidence in using NIRS to estimate age classes of mosquitoes. In the field, we wish to classify the ages of wild, not lab-reared mosquitoes, but the necessary training data from wild mosquitoes is difficult to find. Applying a model trained on spectra from lab-reared mosquitoes to estimate the age of wild mosquitoes is appropriate only if spectra collected from lab-reared mosquitoes are equivalent to those collected from wild mosquitoes. In Chapter Three, we apply k means cluster analysis to a mixture of spectra collected from lab-reared and wild Anopheles arabiensis mosquitoes to determine if there is significant difference between these spectra. We find no significant difference (P = 0.245) in distributions between the wild and lab-reared mosquitoes in the two formed clusters. The two formed clusters have average silhouette coefficient values (cluster quality measure) of 0.51 and 0.77, respectively, which shows that the clusters were reasonable and strong, respectively. Basing on results from Chapter Three, we estimate the age class of wild Anopheles arabiensis mosquitoes using a classification model trained on lab-reared Anopheles arabiensis. We validate the accuracy of the model by comparing its estimates with ovary dissection estimates. While our model estimated 86% and 14% of wild Anopheles arabiensis to be < 7 and 7 days old, respectively, ovary dissection estimated 72% as young and 28% as old. Studies show that wild mosquito populations generally consist of more young than old mosquitoes. Therefore, our model estimates age of wild mosquitoes in consistency with ovary dissection and other studies conducted to determine age structure of wild mosquitoes.