Assessing natural and human‐induced river corridor changes using object‐based image analysis (Orljava River, Croatia)

Abstract

This study employs object‐based image analysis to investigate land cover dynamics and channel changes in the managed corridor of the Orljava River following anthropogenic vegetation removal and a flood event. By classifying RGB and near‐infrared (NIR) images from the decade 2011–2021, five land cover classes within the river corridor were mapped: water, bare soil, sparse vegetation, dense vegetation, and shadows. A digital surface model generated from the images was used to differentiate between bare river channel units (river sediments) and bare soil in the floodplain, as well as to identify high vegetation, while agricultural land was classified manually. The study identified two main phases of river corridor changes, driven by a significant flood event in 2014. In the period 2011–2014, which includes the flood event, a substantial increase in the areas covered by water and river sediments was observed. The most notable decrease was in the bare soil class (90%), which primarily covered riverbanks and adjacent areas in 2011 after vegetation removal. In addition, the formation of large in‐channel bars due to sediment accumulation and significant channel migration was recorded. In the second phase (2014–2021), lower discharges facilitated gradual channel recovery, characterized by channel narrowing and an increase in dense and high vegetation. This study demonstrates the high reliability of object‐based classification in mapping river corridor units using optical photogrammetric data. Furthermore, it highlights the deficiencies in inadequate river management practices that led to considerable bank erosion and a loss of agricultural land.