Abstract
Abstract. As any supervised classification procedure, also Local Climate Zone (LCZ) mapping requires reliable reference data. These are usually created manually and inevitably include label noise, caused by the complexity of the LCZ class scheme as well as variations in cultural and physical environmental factors. This study aims at evaluating the impact of the training set configuration, i.e. training sample number and imbalance, on the performance of Canonical Correlation Forests (CCFs) for a classification of the 11 urban LCZ classes. Experiments are carried out based on globally available Sentinel-2 imagery. Besides multi-spectral observations, different index measures extracted from the images as well as the Global Urban Footprint (GUF) and Open Street Map (OSM) layers are fed into the CCFs classifier. The results show that different LCZs favor different configurations in terms of training sample number and balance. Based on the findings, majority voting of different predictions from different configurations is proposed and performed. This way, a significant accuracy improvement can be achieved.
Highlights
Local Climate Zone (LCZ) mapping (Stewart and Oke, 2012), originally developed for meta-data communication of observational Urban Heat Island (UHI) studies, has gained great interest in the field of remote sensing
In order to guarantee a satisfying LCZ mapping accuracy, the training data should be of sufficient size and provide well balanced sample numbers for all 17 LCZ classes.Unlike training data for other land cover/land use classifications, LCZ training samples are not easy to extract from existing databases
Aiming for global LCZ mapping for which training data is costly and resource intensive to collect, our work intends to provide an answer to these questions: How do the training set size and the distribution of training samples across the classes impact the LCZ classification performances? For simplicity, we focus on the first ten LCZ classes, which are referred to as urban LCZ classes in this paper
Summary
Local Climate Zone (LCZ) mapping (Stewart and Oke, 2012), originally developed for meta-data communication of observational Urban Heat Island (UHI) studies, has gained great interest in the field of remote sensing. The supervised strategy requires a training dataset in order to train a classifier, which can later be used to predict the labels of unseen samples. Each sample in the training dataset is defined by a feature vector and its class label. This training dataset is crucial for the classification accuracy as well as the generalization ability of the trained classifier. In order to guarantee a satisfying LCZ mapping accuracy, the training data should be of sufficient size and provide well balanced sample numbers for all 17 LCZ classes.Unlike training data for other land cover/land use classifications, LCZ training samples are not easy to extract from existing databases. Obtaining reference data of high quality is challenging, especially when it comes
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
