Digital map processing promises computational methods for the extraction of geographic features from scanned historical maps. Such workflows are error prone, with potential spatial uncertainty arising from the initial map production, the processing of the feature extraction, and the eventual application and use of the extracted features. This paper investigates several types of uncertainty emerging the extraction of hydrological features from historical topographic maps for the monitoring of change in ecological indicators describing river ecosystems, such as shoreline length, river sinuosity or number of river nodes and islands. Computational procedures have been developed to simulate various typical, expected sources of error. In a series of experiments investigating three different typical river types, the errors were systematically varied and increased using Monte Carlo simulation whilst studying the errors' impacts on the derived ecological indicators. The results suggest that production-oriented uncertainties emerging the initial map generalization and simplification process have bigger impacts than processing-oriented uncertainties, such as errors from manual digitizing. The results further indicate that the derivation of ecological indicators from braided rivers is more error prone than from straight or meandering rivers, and that topological indicators such as river sinuosity are more robust than indicators derived from the features' geometry.