Meteorological and Geographic Relationships to West Nile Virus in the Southern Plains

Methods
	To aid in analyzing our meteorological variables and West Nile Virus (WNV) outbreaks together, we aggregated the WNV data which was reported daily by county, to the same 2.5-degree, weekly grid that our meteorological variables were reported on. This picture represents our grids. We analyzed our data in python by creating time-series arrays of each variable and for encephalitis and non-neuroinvasive cases separately and together. We also used the least squares method to determine simple correlations between the incidence rate for each year and the winter temperature anomalies of the preceding winter, as well as the summer average temperature.
Time Series: Figure 1:West Nile Virus Cases for Grid M (including Dallas and Southern Oklahoma) Figure 2:West Nile Virus incidence rates vary widely between 2011 and 2012, which happen to be the lowest and highest years for Incidence Rate respectively. These plots suggest that WNV is likely effected by different processes in our Western Grids than Eastern grids. Upon initially examining our data, we created time series plots for each grid. Although too much information for the poster presentation, they're included below because they show many of the conclusions listed on our paper and the reasoning that led us to these conclusions. From this time series, you can see some of the trends found throughout most of our study area. West Nile Virus infections peak between June and August, are nearly 0 for most of the year, and are sensitive to temperature. More specifically, winter temperature anomalies and, to a lesser extent average summer temperature (Figure 8) were found to exert the greatest influence over incidence rate. Figure 3:Comparing yearly Incidence Rate trends across our region suggests a few trends: (1) that the infections in the northern areas of the study region begin much later in the year than in the southern areas (especially in the northeast, where meteorological effects are most important), (2) that in general Incidence rates are higher on the High Plains than in the Eastern regions, as has already been discussed, and (3) that higher years in the Eastern area tend to correspond to extremely high years in the west.
Geography Figure 4:Incidence rates over our entire time period and geographic range revealed that Western portions of our area were most prone to West Nile Virus infections Figure 5:(top-left):Land Cover according to the NLDC Project. (top-right):Agriculture Type from NASS CDL program. (bottom):10 year total Incidence rates overlaid with 33 counties that grow primarily cotton crop. After examining time series plots, we began examining the geography of our region using ArcGIS software, at which time we were surprised to see that the highest incidence rates were in the driest areas of our range, as shown if Figure 3. Upon further examining the plots however, we realized that these areas correlated positively agricultural land-use in western areas of our plots, but not the centrally located agricultural land Figure 4. Further examining USDA agriculture-type plots (Figure 4) reveals that these western areas of the study region correlate to mainly to corn and cotton, which are heavily irrigated. More specifically, there are 33 counties in western Texas that primarily grow cotton crop. When the total incidence rates of these three counties are averaged together, the mean incidence rate is 0.64765. This places the cotton producing counties in the 88th percentile for highest incidence rates of the 436 county study region.
NCEP/NCAR Reanalysis Anomalies Winter Temperature Anomalies: Figure 6:When comparing Average MSLP winter temperature anomalies, it's easy to see that low winter temperature anomalies (left) correlate to low yearly Incidence Rates. The inverse is true, with high winter MSLP winter temperature anomalies correlating to the 3 highest years in our study area. Summer Temperature Anomalies: Figure 7:Summer Average MSLP temperature anomalies tell a much less compelling story. Parts of Kansas are in the same range for our three highest years (right) and our three lowest years (left). However, further study should be done, as the graphs differ much more at lower latitudes. It's likely that threshold temperatures, rather than anomalies, play a much larger role than anomalies.
Begining Statistical Work Figure 8:Across much of our region, Winter Temperature Anomalies (top) have a strong positive correlation with Yearly Incidence Rates. As discussed earlier, summer temperature anomalies do not have such a strong relationship. Summer Average temperatures (bottom) also do not have a strong correlation, especially when not considering 2012 (the solid line in the picture) which was an anomalously high year for West Nile Virus in many locations.