Spatiotemporal Variability in Potential Evapotranspiration across an Urban Monitoring Network

Evapotranspiration in urban and peri-urban environments is difficult to measure and predict. Barriers to accurate assessment include: the wide range of microclimates caused by urban canyons, heat islands, and park cooling; limited instrument fetch; and the patchwork of native soils, engineered soils, and hardscape. These issues combine to make an accurate assessment of the urban water balance difficult, as evapotranspiration calculations require accurate meteorological data. This study examines nearly three years of data collected by a network of 18 weather stations in Dallas, Texas, designed to measure potential evapotranspiration (ETo) in support of the WaterMyYard conservation program (http://WaterMyYard.org). Variability amongst stations peaked during the summer irrigation months, with a maximum standard deviation of 0.3 mm/hr and 4 mm/d. However, we found a significant degree of information overlap in the network. Most stations had a high correlation (>0.75) with at least one other station in the network, and many had a high correlation with at least 10 others. Correlation strength between station ETo measurements did not necessarily decrease with Euclidean distance, as expected, but was more closely related to differences in station elevation and longitude. Stations that had low correlations with others in the network typically had siting and fetch issues. ETo showed a strong temporal persistence; average station autocorrelation was 0.79 at a 1-hour lag and 0.70 at a 24-hour lag. To supplement the larger-scale network data, we deployed a mobile, vehicle-mounted weather station to quantify deviations present in the atmospheric drivers of evapotranspiration: temperature, humidity, wind, and solar radiation. Data were collected at mid-day during the irrigation season. We found differences in mobile and station ETo predictions up to 0.2 mm/hr, primarily driven by wind speed variations. These results suggest that ETo variation at the neighborhood to municipality level may be as important as variations at the scale of the Metroplex for determining watering requirements. We are currently exploring the use of the Landsat8 thermal infrared sensor (TIRS) to characterize variations in land surface temperature and NDVI, which may help us better predict ETo at this important intermediate scale.