HR: 1340h
AN: H53B-1233    [Abstracts]
TI: Classification of Watersheds for Bioassessment Based on Hydrological Variables
AU: * Chinnayakanahalli, K J
EM: kiran@cc.usu.edu
AF: Civil and Environmental Engineering, Utah State University 4110 Old Main Hill, Logan, UT 84322-4110,
AU: Tarboton, D G
EM: dtarb@cc.usu.edu
AF: Civil and Environmental Engineering, Utah State University 4110 Old Main Hill, Logan, UT 84322-4110,
AU: Hawkins, C P
EM: hawkins@cc.usu.edu
AF: Department of Aquatic, Watershed, and Earth Resources College of Natural Resources, Utah State University, Logan, UT 84322-5210,
AB: A procedure for the classification of watersheds for bioassessment based on their streamflow regime and prediction of hydrologic class from watershed attributes is presented. We first identified a set of stream flow regime variables relevant to biota for the purposes of characterizing the invertebrate population in a stream, that can be abstracted from long term streamflow data measured at gauged sites. The selection of these variables was based on the past literature and discussions with stream ecologists. The following variables were selected: 1) base flow index (BFI) 2) daily coefficient of variation (DAYCV) 3) average daily flow (QMEAN), 4) Number of zero flow days (ZERODAY) 5) bank full flow (Q1.67) 6) Colwell's index 7) seven day minimum (7Qmin) 8) seven day maximum (7Qmax) 9) number of flow reversals (NOR) and 10) flood frequency. These variables were computed at 543 minimally impacted stream gage stations in the thirteen states of Western US. Principal Component Analysis (PCA) and K-means clustering analysis was then used to classify the watersheds into hydrologically different groups. Linear Discriminant Analysis (LDA), Classification and Regression Trees (CART) and Random Forests (RF) models were then developed to predict the class of an ungauged watershed from watershed attributes (climate, geomorphic, geology and soil attributes). We developed a series of classifications (with K equal to 4 to 8 in K-means clustering) that showed a strong geographical structure. The classification is sensitive to the quantity of water present in the stream and it also identified streams that appear similar at monthly time scale but are significantly different at the daily time scale. These differences are important to identify the variation in the biota. For the prediction of watershed class from watershed attributes we found that the RF model was slightly better than the other modeling approaches evaluated (LDA, CART). The class characterized by high BFI was difficult to predict in all models due to the lack of good watershed attribute, among those we considered, that is a reasonable surrogate for subsurface flow. The new hydrologic classification of watersheds provides opportunities to further examine the relationships between hydrology and stream ecology.
DE: 1813 Eco-hydrology
DE: 1860 Streamflow
DE: 1874 Ungaged basins
DE: 1879 Watershed
SC: Hydrology [H]
MN: 2007 Fall Meeting