Uncovering the past: tree cover mapping through historical imagery

Accurate and comprehensive tree cover mapping plays a vital role in forest management and ecosystem assessment. The lack of spectral information and the low quality of historical black and white (B&W) images makes it difficult to retroactively predict. The proposed B&WTreeNet based on semantic segmentation makes full use of the limited labeled training data to obtain tree cover information from historical datasets.  

The model is capable of cross-temporal semantic segmentation, including various data augmentation methods, to address inconsistent tree cover features caused by variations in image quality. The designed luminance enhancer and other modules in B&WTreeNet can extract the characteristics of tree cover effectively, successfully compensating for the limited spectral information in B&W image datasets. The countrywide historical tree cover map in Switzerland generated for the year of 1946 and 1980s can be generated using a limited training dataset by making full use of the vegetation height model (VHM) from current years.