The present study constructed a spatial-temporal statistical model to identify the risk and protective factors for haemorrhagic disease (HD) in white-tailed deer in the five states of Alabama, Georgia, South Carolina, North Carolina and Tennessee. The response variable was binary, indicating the presence or absence of HD in an individual county, measured annually from 1983 to 2000. Predictor variables included climatic factors of temperature, rainfall, wind speed and dew point, remotely sensed data of normalised difference vegetation index (NDVI) and land surface temperature derived from archived remotely sensed advanced very-high-resolution radiometer (AVHRR) satellite data, elevation, a spatial autocorrelation (SA) term and a temporal autocorrelation term. This study first applied principal component factor analysis to reduce the volume of climatic data and remotely sensed data. Then, a generalised linear mixed model framework (GLMM) was used to develop a spatial-temporal statistical model. The results showed that the area under receiver operating characteristic curve (ROC) was 0.728, indicating a good overall fit of the model. The total prediction accuracy over the 18 year period with optimal cut-off probability was 67 per cent. The prediction accuracy for individual years ranged from 48 to 75 per cent.
Statistics from Altmetric.com
Provenance not commissioned; externally peer reviewed