HR: 14:25h
AN: H53H-04    [Abstracts]
TI: Integrating Multiple Source Data Uncertainty For Spatiotemporally Distributed Hydrological Systems
AU: * Serre, M L
EM: marc_serre@unc.edu
AF: University of North Carolina, Depart of Environmental Sciences and Engineering, Chapel Hill, NC 27599-7431 United States
AU: Lee, S
EM: seungjae@email.unc.edu
AF: University of North Carolina, Depart of Environmental Sciences and Engineering, Chapel Hill, NC 27599-7431 United States
AU: Money, E
EM: emoney@email.unc.edu
AF: University of North Carolina, Depart of Environmental Sciences and Engineering, Chapel Hill, NC 27599-7431 United States
AU: Akita, Y
EM: akita@email.unc.edu
AF: University of North Carolina, Depart of Environmental Sciences and Engineering, Chapel Hill, NC 27599-7431 United States
AB: This presentation will introduce a comprehensive methodological framework to integrate space/time hydrologic data obtained from sources with different uncertainty types. Integrating data from multiple sources is becoming critical for spatiotemporally distributed hydrological systems because of the wealth of data generated by new data sources that may not have been available in the past, but provide critical information about the distribution across space and time of some hydrologic process of interest. These data sources include direct field observations, indirect measurements, secondary variables related to the variable of interest, remote sensing and satellite technology, flow and transport model predictions, and many others. The critical step in the integration of hydrologic data from multiple sources is to identify and properly model the type of uncertainty associated with each data source, and then to merge the data based on their uncertainty models. The goal of this presentation is thus to develop separate models for several of the most common types of uncertainty associated with existing hydrologic data sources, and to produce an estimation of the spatiotemporal distribution of the hydrologic process of interest that is more accurate than that obtained by ignoring the different types of uncertainty associated with each sources. We will consider uncertainty from measurement error, from empirical laws, from location error, and from observation scale. For each of these uncertainty types we model the uncertainty by means of the conditional probability density function (PDF) of the true process given the measured data and its uncertainty type. We then process these so-called soft PDF using a Bayesian Maximum Entropy approach that rigorously accounts for the data uncertainty as well as the space/time variability of the process of interest, and leads to a full stochastic description of the variable of interest at any unsampled points by means of a posterior PDF. We present surface water and groundwater case studies for each of these types of data uncertainty, which demonstrate the usefulness of the approach in practice.
DE: 1819 Geographic Information Systems (GIS)
DE: 1839 Hydrologic scaling
DE: 1871 Surface water quality
DE: 1873 Uncertainty assessment (3275)
DE: 1894 Instruments and techniques: modeling
SC: Hydrology [H]
MN: Fall Meeting 2005