Regression Tree Analysis for Stream Biological Indicators Considering Spatial Autocorrelation

Area of Science:

• Environmental Science
• Ecology
• Hydrology

Background:

• Land cover significantly influences stream ecosystems, but spatial dependencies are often overlooked.
• Existing research lacks consideration of the systemic structure of streams in land cover analyses.

Purpose of the Study:

• To analyze the relationship between land cover (green/urban areas), topographical variables, and biological indicators in streams.
• To incorporate spatial autocorrelation into stream ecosystem analysis at multiple scales.

Main Methods:

• Regression tree analysis was employed to model relationships, accounting for spatial autocorrelation.
• Principal components analysis identified key variables, with topographical factors showing high importance.
• Moran's I statistic confirmed significant spatial autocorrelation in biological indicators (Trophic Diatom Index, Benthic Macroinvertebrate Index, Fish Assessment Index).

Main Results:

• Topographical variables, particularly riparian slope, were critical in differentiating biological conditions.
• Significant spatial dependency was confirmed between environmental and biological stream indicators.
• Biological indices consistently demonstrated strong spatial autocorrelation.

Conclusions:

• Spatial autocorrelation must be considered in stream ecosystem studies for accurate analysis.
• Riparian proximity and topography are vital factors for land use planning to maintain stream health.
• Integrating spatial analysis enhances understanding and management of stream ecosystems.