Journal
ECOLOGICAL MODELLING
Volume 188, Issue 2-4, Pages 174-216Publisher
ELSEVIER
DOI: 10.1016/j.ecolmodel.2005.01.057
Keywords
multi-layer models; latent heat; model sensitivity; SVAT schemes; grassland; forest ecosystems
Categories
Ask authors/readers for more resources
Latent heat, which plays a major role in the energy balance of ecosystems, is usually calculated in different types of models, e.g., water balance approaches, aggregated big-leaf models, multi-layer ecosystem process models, that can be generally catalogued as Soil-Vegetation-Atmosphere-Transport models (SVATs) in a broader sense, etc. In this study, four (multi-) layered models (PLant-ATmosphere INteraction model (PLATIN), NiixFor-SVAT, SVAT-CN, PnET-N-DNDC) and an atmospheric boundary layer model including vegetation (HIgh Resolution Vegetation Atmosphere Coupler (HIRVAQ) were used to predict latent heat (LE) and sensible heat (H) exchange in the target area of VERTIKO, a research project on vertical transports of energy and trace gases and their spatial/temporal extrapolation under complex natural conditions. Investigated vegetation types were a Norway spruce forest, and two grassland sites in Brandenburg and Saxony (Germany) during field campaigns in 2001 and 2002. Four models had half-hourly to hourly time steps, but PnET-N-DNDC had a daily time step. All models used meteorological conditions above the canopy as input, and computed LE and H independently, or H as the residual of the energy balance. Intercomparisons and discussions of model parameters focussed on vegetation type (or species), and LAI. Model predictions were compared to half-hourly and daily eddy covariance measurements of ecosystem LE and H during two periods (cold and wet fall 2001, warm and dry summer 2002) to include contrasting weather conditions in the model evaluation. The agreement between modeled and measured fluxes was reasonable for most models. Yet, as the models were principally based on energy balance, the lack of energy balance closure as found in the data made it difficult to evaluate models against data, indicating a need for more attention to measurement accuracy in future studies. Systematic differences between the models were obvious. In general, HIRVAC model predictions of LE were routinely the lowest, and PnET-N-DNDC the highest. The tendency of under- or overestimating LE fluxes was comparable in PLATIN, SVAT-CN and MixFor-SVAT. SVAT-CN especially underestimated LE and overestimated H of the forest site. Under- or overestimation by the models could not be attributed to vegetation type. Both, MixFor-SVAT and SVAT-CN underestimated LE of one of the grassland sites, and overestimated the other. When averaging all model outputs, daily fall fluxes reached between 75 and 115% of the data for the three sites (76-83% for the summer period), and between 65 and 90% for half-hourly fluxes (73-97% for summer). Sensitivity analysis for an important model parameter, LAI, revealed only small effects on modeled LE in most models, probably due to compensation in the relative amounts of evapotranspiration and soil evaporation, at least for the limited range of LAI we analyzed (mean +/- 1m(2) m(-2)). The sensitivity to LAI was smaller for forest simulations than for grasslands. For a given site PLATIN and MixFor-SVAT showed lowest, SVAT-CN intermediate, and HIRVAC largest sensitivity to changes in LAI. For latent heat simulations of forests, the sensitivity of LE to changes in LAI was smaller than the differences in the LE predicted by the five models. Further, it is much more important to determine vegetation type and accurate leaf physiology than the LAI of the system. The models differed widely in their approach for canopy radiation, leaf physiology, energy balance, atmospheric transport or soil water modeling, yet the treatment of evaporative losses from soil and interception pools seemed to cause the larger differences between the models after rain events. (c) 2005 Elsevier B.V. All rights reserved.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available