Modeling temperature and humidity profiles within forest canopies

Abstract Physically-based models are a powerful tool to help understand interactions of vegetation, atmospheric dynamics, and hydrology, and to test hypotheses regarding the effects of land cover, management, hydrometeorology, and climate variability on ecosystem processes. The purpose of this paper is to evaluate recent modifications and further refinements to a multi-layer plant canopy model for simulating temperature and water vapor within three diverse forest canopies: a 4.5-m tall aspen thicket, a 15-m tall aspen canopy, and a 60-m tall Douglas fir canopy. Performance of the model was strongly related to source strength and profile stability within the canopy. Root mean square deviation (RMSD) between simulated and observed values tended to be higher for the summer periods when there was much more heat and vapor added to the canopy space due to solar warming and transpiration. Conversely, RMSD for vapor pressure was lowest for the winter periods when vapor additions within the canopy space were minimal. RMSD for temperature ranged from 0.1 °C for the top of the 15-m aspen canopy during the winter to 1.6 °C for the bottom of the 4.5-m aspen thicket during the summer period. RMSD for vapor pressure ranged from 0.002 kPa for the top of the 15-m aspen canopy during winter to 0.141 kPa for the bottom of the 4.5-m aspen thicket during the summer. Unstable profile conditions were simulated better by the model than stable conditions for all sites. RMSD for temperature at the bottom of the 4.5-m aspen, 15-m aspen and 60-m Douglas fir were 0.89, 0.77, and 0.85 °C, respectively, for unstable conditions compared to 1.44, 0.89 and 1.16 °C for stable conditions. Stable profiles are more challenging to accurately simulate because dispersion within a stable profile is lower thereby creating larger gradients. Temperature differences between the bottom and above canopy sensors were within 3 °C for unstable conditions for all sites, but were as much as −10 °C under stable conditions. The model exhibited the greatest discrepancies relative to measurements in the 4.5-m aspen thicket under stable conditions, likely due to horizontal ejections from this relatively small patch of vegetation that could not be addressed by the one-dimensional model. At each site, the model performed best near the top of canopy where the air was well mixed and gradients between it the meteorological conditions above the canopy used to force the model were minimal.