The Human Influence Experiment (Part 2): Guidelines for Improved Mapping of Local Climate Zones Using a Supervised Classification

Abstract

Since 2012, Local Climate Zones (LCZ) have been used for numerous studies related to urban environment. In 2015, this use amplified because a method to map urban areas in LCZs was introduced by the World Urban Database and Access Portal Tools (WUDAPT). However in 2017, the first HUMan INfluence EXperiment showed that these maps often have poor or low quality. Since the maps are used in different applications such as urban modelling and land use/land cover change studies, it is of the utmost importance to improve mapping accuracies and a second experiment was launched. In HUMINEX 2.0, the focus lies on providing guidelines on the use of the mapping protocol based on the results of both HUMINEX 1.0 and 2.0. The results showed that: (1) it is important to follow the mapping protocol as strictly as possible, (2) a reasonable amount of time should be spent on the mapping procedure, (3) all users should perform a driving test, and (4) training area sets should be stored in the WUDAPT database for other users.

Revealing Kunming’s (China) Historical Urban Planning Policies Through Local Climate Zones

“…See Table A1 for more details on all accuracies. This is in line with previous results such as from "The Human Influence Experiment" (HUMINEX) [56,57] and is a limitation of the input feature space that is currently not able to detect height variations (see, e.g., discussion on page 12 of Reference [39]). In Table 3, the area of every LCZ is shown for the different years.…”

“…following [34,[55][56][57]. In order to assess the robustness of the different classification methods, standard deviations (SD) on the OA and F1 are provided as well (see Table A1).…”

Combining expert and crowd-sourced training data to map urban form and functions for the continental US

“…The match percentage was 100% (n [number of matched polygons] = 8) for the full match approach, 87% (n = 69) for the match by centroid, and 65% (n = 141) for the match by intersection (Supplementary Table S1 ). While differences occur, the degree of consistency is actually higher compared to the results of HUMINEX (HUMan INfluence EXperiment 58 , 59 ), that indicated large discrepancies between training area sets from multiple ‘experts’, nevertheless leading to strong improvements in overall accuracy when used all together. Combining expert and crowd-sourced data are therefore a reasonable approach to diversify training data for developing LCZ classification models.…”

“…TA data are generally created by urban experts 31 , a time-demanding procedure, both because of the intrinsic nature of the task (i.e., the extent and heterogeneity of urban areas) and the ability of the urban expert to identify and digitize TAs consistently 58 , 59 . Here, expert TAs are used from nine U.S. cities: Phoenix and Las Vegas 60 , Salt Lake City 61 , Chicago and New York 62 , 63 , Houston, Washington D.C., Philadelphia and Los Angeles.…”

Urban heat island behaviors in dryland regions

“…To analyze neighborhood-scale effects, we used Google Earth Engine to visually identify representative samples exceeding one square kilometer in size of six different land cover types (see figure 1). We elected to use this domain expert approach to ensure near 100% accuracy, rather than employing a machine learning classification algorithm such as the World Urban Database Access Portal Tools (Verdonck et al 2019) to quickly classify such areas with moderate accuracy (70%-85%). Our approach was similar to that used by the HERCULES model for classifying urban patches (Cadenasso et al 2007).…”