Woodmouse Scenario Construction
Modelling of population effects of non-target species (e.g. woodmouse) due to pesticide exposure requires appropriate scenarios. The core element of scenario construction is to develop land use/cover (LULC) information based on high resolution image data. eCognition offers numerous possibilities to fuse data of different formats and combine different methods for object-based LULC classification.
The classification routine followed a hybrid top-down approach, applying knowledge-based thresholds, machine learning methods as well as sample-trained convolutional neural networks. The preliminary results were then reshaped to decrease object vertices and receive almost GIS-ready classified segments. The results of this LULC analysis were used as input for geospatial scenario modeling.