Biological Integrity of Constrained Streams by Stream [ds2809]

The landscape model was developed for California using land use data, stream hydrography, and biological assessments. Stream data from the National Hydrography Dataset Plus (NHD-plus) were used to identify stream segments in California for modelling biological integrity. Stream segments designated in the NHD-plus were used as the discrete spatial unit for modelling biological integrity. Hydrography data were combined with landscape metrics available from the StreamCat Dataset to estimate land use at the riparian zone (i.e., a 100-m buffer on each side of the stream segment), the catchment (i.e., nearby landscape flowing directly into the immediate stream segment, excluding upstream segments), and the entire upstream watershed for each segment. The California Stream Condition Index (CSCI) was used as a measure of biological condition in California streams. The CSCI is a predictive index that compares the observed taxa and metrics at a site to those expected under reference conditions. A dataset of 2620 unique CSCI scores was used to calibrate and validate the landscape model.

A quantile random forest model was developed to estimate ranges of CSCI scores associated with land use gradients, such as road density or urban and agricultural land use. Measures of land use development were quantified for riparian, catchment, and watershed areas (as defined above) of each stream segment in California using the StreamCat dataset. Expected CSCI scores were modelled using estimates of canal/ditch density, imperviousness, road density/crossings, and urban and agricultural land use for each stream segment.

We applied the model to stream segments statewide to estimate the extent of streams in one of four different constraint classes: likely unconstrained, possibly unconstrained, possibly constrained, and likely constrained. Here and throughout, constrained is defined as a biological community that is impacted by large-scale, historic alteration of the landscape. Consequently, achieving biological integrity in constrained communities may present management challenges given that many stressors in altered landscapes originate at spatial or temporal scales that are typically beyond the scope of most management applications or where resources for mitigation may be prohibitive.

The stream classifications were based on the comparison of a CSCI threshold representing a management goal and the predicted range or predicted median score at a segment. These two decision points (i.e., the threshold and the size of the predicted range) were critical in defining segment classifications. The dataset includes stream classifications using three thresholds for the CSCI (first, tenth, and thirtieth percentile of reference sites) and a prediction interval ranging from the 10th to the 90th percentiles of the quantile predictions. Stream segments with the range of CSCI score expectations entirely below the threshold were considered likely constrained, whereas those with expectations entirely above were considered likely unconstrained. The remaining sites were classified as possibly unconstrained or possibly constrained, based on whether the median expectation was above or below the threshold (respectively). Streams with insufficient data to predict score expectations were not assigned a classification.