HR: 1330h
AN: H42A-1060 [PDF]
TI: Calibration Strategies for Global Land Surface Modeling
AU: Goteti, G
EM: ggoteti@princeton.edu
AF: Princeton University, Department of Civil and Environmental Engineering, Princeton, NJ 08544 United States
AU: * Sheffield, J
EM: justin@princeton.edu
AF: Princeton University, Department of Civil and Environmental Engineering, Princeton, NJ 08544 United States
AU: Wood, E F
EM: efwood@runoff.princeton.edu
AF: Princeton University, Department of Civil and Environmental Engineering, Princeton, NJ 08544 United States
AU: Cherkauer, K
EM: cherkaue@u.washington.edu
AF: Purdue University, Department of Agricultural & Biological Engineering
225 South University Street, West Lafayette, IN 47907 United States
AB:
Simulated datasets of global land surface fluxes and states are a major element of strategies to evaluate the effect of
global warming on the land surface hydrologic cycle. Generating these datasets requires some level of calibration of model
parameters to ensure that the resultant fields are as accurate as possible. This can be a difficult task given the lack of
large scale validation datasets and the computational limitations of employing recursive calibration strategies. Using the
Variable Infiltration Capacity (VIC) land surface model and a 50-yr forcing dataset we show an efficient strategy for
calibration over large spatial scales. This strategy consists of multi-seasonal and multi-objective calibration, the use of
representative sampling techniques and geostatistical interpolation based on climate zoning and soil and vegetation
distributions. In most past studies, observed streamflow has been the sole calibration index for small and large basin scale
hydrology. This study also uses evaporation as a calibration index as it is generally a larger component of the water balance
than runoff and can provide information about the dynamics within the basin. The evaporation dataset used for calibration is
derived from a 20-yr global and a 50-yr US simulated dataset. Calibration is only carried out on representative sets of grid
cells to reduce the computational burden of running multiple simulations. These cells are chosen to be collectively
representative of the larger area in terms of their hydrologic response and as such reflect the variability in the climate
and land surface of the region. This fact is used to spatially interpolate the resulting calibrated parameters over the whole
modeling domain using geostatistical techniques. The calibrated model is being used to generate global fields of land
surface fluxes and states for climate studies.
DE: 1818 Evapotranspiration
DE: 1833 Hydroclimatology
DE: 1836 Hydrologic budget (1655)
DE: 1860 Runoff and streamflow
SC: Hydrology [H]
MN: 2003 Fall Meeting