A Methodology Simplifying River Habitat Quality Assessment through Regression Analysis: Application in Slovakia

Abstract In the past, river restoration mainly focused on flood protection and somewhat ignored the degradation of river ecosystems, e.g., the habitats of river organisms. In recent decades, however, there has been an increased effort to carefully assess the quality of aquatic habitats as well. The authors have collected data from mountain and sub-mountain streams, including the habitat preferences of fish, topographic surveys, hydrometric measurements, and hydraulic modelling. These data were used to model the quality of stream habitats using the System of Environmental Flow Analysis (SEFA). SEFA, however, may be perceived as impractical for routine usage since the data it requires are extensive and time-consuming to obtain. For simplification, a regression analysis was conducted using only part of SEFA’s input data to balance reliability and data requirements of the computations. The general contribution of the paper is a demonstration of working with small datasets, addressing the challenges of analyzing the quality of river habitat through techniques such as boosting and regularization in regression analyses. The study confirmed a satisfactory agreement between the SEFA model’s results and the proposed regression methods, especially when using the boosting machine learning algorithm for the regression analysis (with a correlation of 0.9). The regression method significantly reduced the input data necessary to evaluate the quality of a habitat compared to the SEFA model. This permits an assessment of the ecological state of streams not only in a scientific context, but also in standard engineering practice.