It has been demonstrated in past studies that drainage area (Ada) and mean annual precipitation depth (Pmean) are significant independent variables for predicting bankfull channel geometry. Bankfull width (Wbf), a particularly useful measure of bankfull channel geometry, is used in hydrology, fluvial geomorphology and stream ecology applications. Since precipitation depth does not generally have a normal distribution over a basin it is reasonable to question whether or not Pmean is the best measure of precipitation across a basin. That is, perhaps the median precipitation depth (P50) or the mean annual precipitation depth at the basin outlet (Poutlet) is more representative of precipitation in a basin. The objective of the study is to see if Pmean, P50, or Poutlet is a significantly better measure of precipitation in a watershed. Along with Ada, each of these precipitation measures was used to develop relationships for predicting Wbf for basins located across the continental U.S. Using several goodness-of-fit statistics i.e., coefficient of determination(R2), Standard error of estimates(SEE) and Akaike Information Criterion(∆AIC), the relationships were compared to determine which one had the greatest predictive strength. Estimates of Pmean, P50, and Poutlet were derived from a precipitation model developed by PRISM. Results from this study tentatively indicate that all three measures of precipitation depth yield models for predicting Wbf that have comparable goodness-of-fit statistics.

This study is concerned with the effect that mean annual precipitation (P) has on the relationship between bankfull channel width (Wbf) and drainage area (Ada). Several other studies have been conducted in which relationships were developed for predicting Wbf as a function of Ada and P. In most cases, however, the relationships were developed for specific regions, e.g., physiographic regions. This study is unusual in that it evaluates the relationship between Wbf, Ada, and P over a broad area (i.e., across a range of geologic, terrestrial, and climatic environments). In one study, where a broad area was considered, the relationship between Wbf, Ada, and P was found to be linear. The dataset for this study was compiled from data in U.S. Geological Survey flood-flow-frequency reports, regional curve studies (i.e., studies in which Wbf vs. Ada relationships are developed) and other sources. A total of 435 sites that span across 12 states of the continental U.S. are represented in the dataset. Streams represented in the dataset are alluvial and have widths from 1 to 110 m, drainage areas from 0.50 to 22,000 km2, and mean annual precipitation depths ranging from 22 to 277 cm/yr. Data from the U.S. Environmental Protection Agency's Wadeable Streams Assessment study were employed in validating the results of this study. An analysis of covariance (ANCOVA) model was developed and it was determined that the intercept coefficient for the relationship between Wbf and Ada varies as follows: for P < 50 cm/yr the intercept coefficient (α) is constant; for 50 cm/yr ¡Ü P ¡Ü 100 cm/yr, α increases with P, and for P ¡Ý100 cm/yr, α is again constant. Across all values of P, the slope coefficient is constant (90% Confidence level). Changes in the relationship between Wbfand Ada are attributed to vegetation by noting that biome types changes from shrubland to forest as P increases from 50 to 100 cm/yr. These findings can be incorporated in regional curve studies and landscape evolution models (i.e., models which aim to integrate hydrology, land use history, geomorphology and climate change with models of vegetation succession).