As part of the Baltimore Ecosystem Study (BES), a Long Term Ecological Research site, we made continuous measurements of air temperature and humidity at the 1.5-m height at five sites near the city of Baltimore, MD from 2003 to the present. These sites are in an untreed pasture, a nearby woodlot, a residential area with heavy tree cover but few buildings, a lawn area with nearby trees between two large apartment complexes, and a residential area with some trees and large lawn areas. Data are also available from two National Weather Service Automated Surface Observing System (ASOS) sites—in downtown Baltimore (call sign DMH) and at the Baltimore/Washington International Airport (BWI). An objective of our measurements is to model and map the influence of tree cover and changes in tree cover on below-canopy air temperature within the urban landscape. Urban air temperature is important for its effects on space-conditioning energy use, carbon cycling, soil and stream temperatures, ozone formation, human thermal comfort, and human health.

For one summer of hourly data, we carried out regression analysis of the temperature difference, DT, between DMH as a reference and each of the other sites. We derived predictor variables from differences in upwind tree, impervious, and water cover from the 30-m resolution National Land Cover Data (NLCD) 2001 using wedge-shaped focal analysis functions in ESRI ArcGIS. Additional predictor variables for DT were atmospheric stability (Turner Class, derived from BWI cloud and wind speed data), vapor pressure deficit, antecedent precipitation, and descriptors of topography. Interactions between stability and the other predictor variables were included to account for the predictor influences varying with stability. The DMH site was generally warmer than the other sites, with DT being as large as 11 ºC. Significant predictors of DT were land cover, topographic relationships that included elevation differences between the sites and indicators of potential for cold air drainage, a combination of water cover with Chesapeake Bay water temperature, and antecedent precipitation; but these variables strongly interacted with Turner Class. The most significant prediction equations were found with stable atmospheric conditions, which occur at night with clear skies and low wind speed.

We used the regression results to prepare GIS maps from predicted DT for every cell in the NLCD and DEM images of the Baltimore region. GIS tools were also used to create modified tree cover scenarios to simulate and map examples of the regional influences on air temperature from changes in tree cover in developed land-use classes. The largest temperature difference occurred for stable atmospheric conditions when all trees were removed (leaving grass cover) from all developed land uses. Doubling tree cover in developed land uses caused much smaller simulated reductions in air temperature than the increases in temperature by removing all trees. Daytime temperature differences among different tree cover simulations were small.