University of Cagliari
The challenge of fully-predictive hydrologic models supported by observations: recent experiences and prospects in semiarid systems and Feasibility of high-resolution distributed hydrologic modeling using high performance computing
05 July 2011
Dipartimento di Ingegneria del Territorio (DIT)
Il prof. Enrique Vivoni dell’Arizona State University terrà i seguenti due seminari:
Martedì 5 Luglio ore 9.00 - The challenge of fully-predictive hydrologic models supported by observations: recent experiences and prospects in semiarid systems. Il seminario si terrà nell’Aula A della Sezione d’Idraulica della Facoltà d’Ingegneria
Giovedì 14 Luglio ore 9.00 - Feasibility of high-resolution distributed hydrologic modeling using high performance computing. lI seminario si terrà nell’Aula A della Sezione d’Idraulica della Facoltà d’Ingegneria
The challenge of fully-predictive hydrologic models supported by observations: recent experiences and prospects in semiarid systems.
Abstract:
After several decades of development and applications, distributed watershed models are now common tools in hydrologic research and increasingly used in practice. Unfortunately, the sophistication of these tools has not been accompanied by comparable levels of observational data collection. Ultimately, predictions from distributed watershed models need to be verified to build confidence in their ability to simulate the past and subsequently the future, under changing conditions. This talk will describe our efforts to develop, apply and test a distributed watershed model for semiarid regions in the southwestern United States and northwestern Mexico. We will provide examples from forested mountain, semiarid desert and subtropical shrubland systems which all have marked seasonality due to the North American monsoon. Our previous efforts include comparisons to eddy covariance data, distributed soil moisture patterns and runoff observations, among others. After building model confidence, we explore the underlying patterns, processes and mechanisms in the distributed simulations to identify emergent behavior. The intent of this synthesis stage is to infer generalizable features of the hydrologic system that may help explain the observed data and build predictive capacity for other settings. We also discuss prospects for improved joint use of distributed watershed models and distributed observations from ground, aerial and satellite platforms.
Feasibility of high-resolution distributed hydrologic modeling using high performance computing.
Abstract:
A primary obstacle towards advances in large-scale watershed simulations has been the limited computational capacity available to most hydrologic models. In this study, we present the developments in the parallelization of a fully-distributed hydrologic model based on a triangular irregular network (TIN). We describe the model implementation using parallel computing and its application to operational watersheds in Oklahoma. Our approach utilizes domain decomposition based on sub-basin partitioning of the watershed using a stream reach graph. Individual sub-basins also share subsurface hydrologic fluxes across adjacent boundaries. We conduct experiments for a range of basin sizes and partitioning methods to analyze the impact of TIN resolution and number of processors on model performance.
We also discuss the opportunities afforded by parallel processing for hydrologic simulations that preserve landscape properties at high-resolution.
____________________________________________________________________
Prof. Roberto Deidda
Dipartimento di Ingegneria del Territorio (DIT)
UNIVERSITA’ DEGLI STUDI DI CAGLIARI
Phone: +39 070 6755324
Fax: +39 070 6755310
e-mail: rdeidda@unica.it
____________________________________________________________________