Abstract: Sustaining or restoring riparian quality is essential to achieve and maintain good stream health, as well
as to guarantee the ecological functions that natural riparian areas provide. Therefore, quantifying riparian
quality is a fundamental step to identify river reaches for conservation and/or restoration purposes.
Most of the existing methods assessing riparian quality concentrate on field surveys of a few hundreds
of metres, which become very laborious when trying to evaluate whole catchments or long river corridors.
Riparian quality assessment obtains higher scores when riparian vegetation consists of forested
areas, while land-uses lacking woody vegetation typically represent physical and functional discontinuities
along river corridors that undermine riparian quality. Thus, this study aimed to analyse the
ability of riparian land-cover data for modelling riparian quality over large areas. Multiple linear regression
and Random Forest techniques were performed using land-use datasets at three different spatial
scales: 1:5000 (Cantabrian Riparian Cover map), 1:25,000 (Spanish Land Cover Information System) and
1:100,000 (Corine Land Cover). Riparian quality field data was obtained using the Riparian Quality Index.
Hydromorphological pressures affecting riparian vegetation were also included in the analysis to determine
their relative weight in controlling riparian quality. Linear regression showed better predictive
ability than Random Forest, although this may be due to our relatively small dataset (approx. 150 cases).
Forest coverage highly determined riparian quality, while hydromorphological pressures and land-use
coverage related to human activities played a smaller role in the models. While acceptable results were
obtained when using high-resolution datasets, the use of Corine Land Cover led to a poor predictive
ability.
