A New Fully Coupled Model for Improving the Representation of Urban Heterogeneous Hygrothermal Processes

Urban complex geometry and heterogenous hygrothermal properties significantly influence the energy-water cycle in cities. Moreover, representation of surface overland flow and subsurface flow (specifically in places where groundwater is shallow) affects the soil moisture distribution and evolution in atmospheric-hydrological modeling. We built a new model, by coupling two existing models (WRF-ParFlow and WRF-PUCM, where PUCM is the Princeton Urban Canopy Model) to improve the physical representation of: (1) the heterogenous hygrothermal properties of urban terrain; and (2) the terrestrial hydrology of urban areas. As a test illustration of model capability, we applied WRF-PUCM-ParFlow and WRF-PUCM to Dead Run watershed in Baltimore, MD, USA. The extent of the domain was 10.8 km in north-south and east-west directions (total area 116.6 km²), with a horizontal resolution of 90 m and 120 grid cells in each direction. The analysis simulation period was 96 hours, starting from July 19, 2008 to July 23, 2008. The main reason for choosing this period was to compare the two models’ performances during the initial dry-down period followed by two rain events toward the end of the simulation. This combination of events enabled comparing the difference between representation of terrestrial hydrology and its effect on energy-water cycle in the two simulations. The two models started from the same initial conditions. Simulated soil moisture (SM), land surface temperature (LST), air temperature, and precipitation were compared. Area-averaged time-series of precipitation were almost identical in the two simulations. Area-averaged time-series of SM showed different responses. The area-averaged SM increase in WRF-PUCM run alone was four times larger than in the WRF-PUCM-ParFlow. Comparing the contour plots of hourly-averaged soil moisture, a significant difference between the SM spatial distribution over the whole domain was observed toward the end of the simulation. The WRF-PUCM model showed that concentrated infiltration of precipitation propagated over the east side of the domain even though impervious surfaces covered more than 50 % of the area, whereas the WRF-PUCM-ParFlow model had less infiltration over the impervious surfaces on the same side of the domain and turned most of the precipitation into runoff. The difference between the spatial distribution of soil moisture caused significant differences between the area-averaged LST, near the end of the simulation. Between hour 84 to hour 90 of the simulation (7 AM to 1 PM locally), the area-averaged LST difference reached ≈ 2 Cº. The results suggest that this difference is caused by the difference between representation of terrestrial hydrology in the two models, and underlines the importance of highly-accurate representation of hydrological processes.