Predicting channel stability in Colorado mountain streams using hydrobiogeomorphic and land use data: A cost-sensitive machine learning approach to modeling rapid assessment protocols.

摘要

Natural resource data are typically non-linear and complex, yet modeling methods often utilize statistical analysis techniques, such as regression, that are insufficient for use on such data. This research proposes an innovative modeling method based on pattern recognition techniques borrowed from the field of machine learning. These techniques make no data distribution assumptions, can fit non-linear data, can be effective on a small data set, and can be weighted to include relative costs of different predictive errors.; Rapid Assessment Protocols (RAPs) are commonly used to collect, analyze, and interpret stream data to assist diverse management decisions. A modeling method was developed to predict the outcome of a RAP in an effort to improve accurate prediction, weighted for cost-effectiveness and safety, while prioritizing investigations and improving monitoring. This method was developed using channel stability data collected from 58 high-elevation streams in the Upper Colorado River Basin. The purpose of the research was to understand the relationships of channel stability to several hydrobiogeomorphic features, easily derived from paper or electronic maps, in an effort to predict channel stability. Given that the RAP used was developed to evaluate channel stability, the research determined: (1) relationships between channel stability and major land-use and hydrobiogeomorphic features, and (2) if a predictive model could be developed to aid in identifying unstable channel reaches while minimizing costs, for the purpose of land management.; This research used Pearson's and chi-squared correlations to determine associative relationships between channel stability and major land-use and hydrobiogeomorphic features. The results of the Pearson's correlations were used to build and test classification models using randomly selected training and test sets. The modeling techniques assessed were regression, single decision trees, and bagged (bootstrap aggregated) decision trees. A cost analysis/prediction (CAP) model was developed to incorporate cost-effectiveness and safety into the models. The models were compared based on their (1) performance and (2) operational advantages and disadvantages. A reliable predictive model was developed by integrating a CAP model, receiving operator characteristic curves, and bagged decision trees. This system can be used in conjunction with a GIS to produce maps to guide field investigations.