H31C-01
Initial CEOP-based Review of the Prediction Skill of Operational General Circulation Models and Land Surface Models
Using data archived in the Coordinated Enhanced Observing Period (CEOP) project, this study presents an initial evaluation of the prediction skill of five General Circulation Models (GCMs) and three Land Surface Models (LSMs). Comparisons between observations and the GCMs show that all the models are able to produce an afternoon peak in precipitation, but other major features are not well produced, including the total amount of precipitation, onset time of the afternoon peak, the early-evening low (around 1800 LST), and the partition between convective and stratiform rainfall. The ratios of evaporation to precipitation differ among the GCMs. Evaporation in some of the GCMs is even greater than precipitation, perhaps due to the model spin-up effect. In terms of the surface radiation budget, the GCMs generally over-predict downward shortwave radiation and under- predict downward longwave radiation; further investigations of the causes of these trends require cloudiness observations. In terms of the surface energy budget, the GCMs generally over-predict nighttime downward sensible heat fluxes and under-predict diurnal ranges of surface-air temperature difference, as heat transfer resistances are under-predicted. Finally, three offline LSMs driven by identical forcing are evaluated, and we note that the reproduction of surface temperature is not a sufficient condition for a LSM to reproduce surface energy partition.
H31C-02
Use of Factorial Analysis to Determine the Interaction Between Parameters of a Land Surface Model
Land surface models use several parameters to represent biophysical processes. These parameters frequently are unknown, reproducing with uncertainty the characteristics of the ecosystem in study. Model calibration techniques find values for each parameter that reduce uncertainty. However, the calibration process is computationally expensive, since is necessary a lot of model runs to have their parameters adjusted. The more parameters are considered, more difficult the process is, particularly when there are interactions among them, and a modification in a parameter value implies in the change of the optimum value of the other parameters. The use of a factorial experiment allows the identification of possible inert parameters, whose values do not influence the final result of the experiment and, therefore, could be excluded from the calibration process. In this work we used factorial analysis to verify the existence of interaction among 5 parameters of the land surface IBIS model - Beta2 (distribution of fine roots), Vmax (maximum Rubisco enzyme capacity), m (coefficient related to the stomatal conductance), CHS (heat capacity of stems) and CHU (heat capacity of leaves) - evaluated against the output fluxes Rn (net radiation), H (sensible heat flux), LE (latent heat flux) and NEE (net ecosystem CO2 exchange). Data was collected at the Amazon tropical rainforest site known as K83, near Santarem, Brazil. The knowledge of the existing interactions between the parameters can considerably reduce the computational cost of further optimization processes, since each parameter that does not interact with others should be optimized independently.
H31C-03
Retrospective atmospheric forcing datasets in support to the South American Land Data Assimilation System (SALDAS)
Land surface models (LSM) have been extensively used to better represent hydrological and land surface processes, providing lower boundary conditions for numerical climate and weather simulations. However, since the quality of the atmospheric forcing greatly impact LSM simulations the recently, so-called Land Data Assimilation Schemes (LDAS) have been successfully employed to provide improved initial surface fields of soil moisture for use in predictive meteorological models (in near real time) and to address land-surface management issues. However on some regions of the globe, in particular South America, it can pose an important challenge when using LDAS or when assessing the performance of coupled (or uncoupled) land- atmosphere parameterizations due to the scarcity of comprehensive land-surface data at the spatial and temporal resolutions at which the models operate. On that basis a South American Land Data Assimilation System (SALDAS) was created by a joint initiative between NASA/Goddard Space Flight Center and the Centro de Previsao do Tempo e Estudos Climaticos (CPTEC), part of the Brazilian National Institute for Space Research (INPE) and in collaboration with the Department of Hydrology and Water Resources at the University of Arizona. The project has produced retrospective (2000 2004) datasets to support LSM studies over South America, featuring 0.125 degree spatial resolution, 3-hourly temporal output with nine primary forcing fields based on the South American Regional Reanalysis (SARR) produced at CPTEC, supplemented by observed-based shortwave radiation and precipitation from TRMM merged with surface gauges. This study presents an overview of the data production, quality control and validation efforts including comparison against independent automatic surface weather stations which provide high temporal measurements over regions with different climate regimes in South America.
H31C-04
CALIBRATION OF A LAND SURFACE MODEL FOR THE SIMULATION OF SENSIBLE AND LATENT HEAT FLUXES USING THREE DIFFERENT METHODS
In this paper, we compare three different procedures to calibrate land surface models. The procedures are illustrated using the Integrated Biosphere Simulator (IBIS) and a dataset collected at the Flona Tapajos km 83 Amazonian tropical rainforest site, located near Santarem, state of Para, Brazil. The three procedures are: (a) a single criterion procedure (minimization of RMSE H/LE); (b) a two criteria procedure (minimization of RMSEH + RMSELE); (c) a two-step, six-criteria procedure (our proposal). The two-step procedure first filters only the non-biased estimates and then minimizes a four parameter function. The two-step procedure guarantees a consistent non-biased estimation for the evaluated output variables, sensible heat flux (H) and latent heat flux (LE), and seems to be more robust than the other two methods, for the skill functions evaluated.
H31C-05
Investigating the Capability of High Resolution ALSM to Provide Accurate Watershed Delineation and Stream Network Data
The development of geographic information systems (GIS) and digital elevation models (DEMs) has provided an opportunity to describe the pathways of water movement in a watershed. Adequate DEM resolution is of high importance in stream network detection. Local, state, and federal agencies have relied on US Geological Survey 1:24,000 scale topographic maps for information on stream networks for planning, management, and regulatory programs related to streams. DEM creation techniques that avoid map contours as the source of digital heights can improve watershed delineation and stream network data quality. Airborne Laser Swath Mapping (ALSM) technology (also referred to as LIDAR) provides DEMs of fine resolution and high accuracy. However, there are shortcomings in using both low resolution and high resolution DEMs. The focus of this work will be in the unique aspects of using ALSM data in watershed delineation and stream network mapping, in comparison to the other sources of DEM. In particular the reliability of both input data and output results of stream network using different resolutions will be evaluated. In this study, stream location resulting from high-resolution ALSM and low- resolution NED are compared to ground truth locations of the stream in Hogtown Creek Watershed, located in Gainesville, Florida. This study shows that ALSM-derived models are more successful at delineating streams and at locating them in their topographically correct position as compared to lower resolution DEMs. However, high resolution ALSM data produce artifacts that can affect the flow of water as predicted by stream network algorithms. Methods for overcoming the challenges with regard to ALSM data in stream network detection are presented.