Regression Tree Analysis for Stream Biological Indicators Considering Spatial Autocorrelation.

Multiple studies have been conducted to identify the complex and diverse relationships between stream ecosystems and land cover. However, these studies did not consider spatial dependency inherent from the systemic structure of streams. Therefore, the present study aimed to analyze the relationship between green/urban areas and topographical variables with biological indicators using regression tree analysis, which considered spatial autocorrelation at two different scales. The results of the principal components analysis suggested that the topographical variables exhibited the highest weights among all components, including biological indicators. Moran's I values verified spatial autocorrelation of biological indicators; additionally, trophic diatom index, benthic macroinvertebrate index, and fish assessment index values were greater than 0.7. The results of spatial autocorrelation analysis suggested that a significant spatial dependency existed between environmental and biological indicators. Regression tree analysis was conducted for each indicator to compensate for the occurrence of autocorrelation; subsequently, the slope in riparian areas was the first criterion of differentiation for biological condition datasets in all regression trees. These findings suggest that considering spatial autocorrelation for statistical analyses of stream ecosystems, riparian proximity, and topographical characteristics for land use planning around the streams is essential to maintain the healthy biological conditions of streams.