Feature Importance Analysis for Local Climate Zone Classification Using a Residual Convolutional Neural Network with Multi-Source Datasets

Improving the local climate zone classification with building height, imperviousness, and machine learning for urban models

The Local Climate Zone (LCZ) classification is already widely used in urban heat island and other climate studies. The current classification method does not incorporate crucial urban auxiliary GIS data on building height and imperviousness that could significantly improve urban-type LCZ classification utility as well as accuracy. This study utilized a hybrid GIS- and remote sensing imagery-based framework to systematically compare and evaluate different machine and deep learning methods. The Convolution Neural Network (CNN) classifier outperforms in terms of accuracy, but it requires multi-pixel input, which reduces the output’s spatial resolution and creates a tradeoff between accuracy and spatial resolution. The Random Forest (RF) classifier performs best among the single-pixel classifiers. This study also shows that incorporating building height dataset improves the accuracy of the high- and mid-rise classes in the RF classifiers, whereas an imperviousness dataset improves the low-rise classes. The single-pass forward permutation test reveals that both auxiliary datasets dominate the classification accuracy in the RF classifier, while near-infrared and thermal infrared are the dominating features in the CNN classifier. These findings show that the conventional LCZ classification framework used in the World Urban Database and Access Portal Tools (WUDAPT) can be improved by adopting building height and imperviousness information. This framework can be easily applied to different cities to generate LCZ maps for urban models.

Mapping Local Climate Zones in the Urban Environment: The Optimal Combination of Data Source and Classifier

The novel concept of local climate zones (LCZs) provides a consistent classification framework for studies of the urban thermal environment. However, the development of urban climate science is severely hampered by the lack of high-resolution data to map LCZs. Using Gaofen-6 and Sentinel-1/2 as data sources, this study designed four schemes using convolutional neural network (CNN) and random forest (RF) classifiers, respectively, to demonstrate the potential of high-resolution images in LCZ mapping and evaluate the optimal combination of different data sources and classifiers. The results showed that the combination of GF-6 and CNN (S3) was considered the best LCZ classification scheme for urban areas, with OA and kappa coefficients of 85.9% and 0.842, respectively. The accuracy of urban building categories is above 80%, and the F1 score for each category is the highest, except for LCZ1 and LCZ5, where there is a small amount of confusion. The Sentinel-1/2-based RF classifier (S2) was second only to S3 and superior to the combination of GF-6 and random forest (S1), with OA and kappa coefficients of 64.4% and 0.612, respectively. The Sentinel-1/2 and CNN (S4) combination has the worst classification result, with an OA of only 39.9%. The LCZ classification map based on S3 shows that the urban building categories in Xi'an are mainly distributed within the second ring, while heavy industrial buildings have started to appear in the third ring. The urban periphery is mainly vegetated and bare land. In conclusion, CNN has the best application effect in the LCZ mapping task of high-resolution remote sensing images. In contrast, the random forest algorithm has better robustness in the band-abundant Sentinel data.

Comparing Object-Based and Pixel-Based Methods for Local Climate Zones Mapping with Multi-Source Data

The local climate zones (LCZs) system, a standard framework characterizing urban form and environment, effectively promotes urban remote sensing research, especially urban heat island (UHI) research. However, whether mapping with objects is more advantageous than with pixels in LCZ mapping remains uncertain. This study aims to compare object-based and pixel-based LCZ mapping with multi-source data in detail. By comparing the object-based method with the pixel-based method in 50 and 100 m, respectively, we found that the object-based method performed better with overall accuracy (OA) higher at approximately 2% and 5%, respectively. In per-class analysis, the object-based method showed a clear advantage in the land cover types and competitive performance in built types while LCZ2, LCZ5, and LCZ6 performed better with the pixel-based method in 50 m. We further employed correlation-based feature selection (CFS) to evaluate feature importance in the object-based paradigm, finding that building height (BH), sky view factor (SVF), building surface fraction (BSF), permeable surface fraction (PSF), and land use exhibited high selection frequency while image bands were scarcely selected. In summary, we concluded that the object-based method is capable of LCZ mapping and performs better than the pixel-based method under the same training condition unless in under-segmentation cases.

Advances of Local Climate Zone Mapping and Its Practice Using Object-Based Image Analysis

In the context of climate change and urban heat islands, the concept of local climate zones (LCZ) aims for consistent and comparable mapping of urban surface structure and cover across cities. This study provides a timely survey of remote sensing-based applications of LCZ mapping considering the recent increase in publications. We analyze and evaluate several aspects that affect the performance of LCZ mapping, including mapping units/scale, transferability, sample dataset, low accuracy, and classification schemes. Since current LCZ analysis and mapping are based on per-pixel approaches, this study implements an object-based image analysis (OBIA) method and tests it for two cities in Germany using Sentinel 2 data. A comparison with a per-pixel method yields promising results. This study shall serve as a blueprint for future object-based remotely sensed LCZ mapping approaches.

Combination of Sentinel-2 and PALSAR-2 for Local Climate Zone Classification: A Case Study of Nanchang, China

Local climate zone (LCZ) maps have been used widely to study urban structures and urban heat islands. Because remote sensing data enable automated LCZ mapping on a large scale, there is a need to evaluate how well remote sensing resources can produce fine LCZ maps to assess urban thermal environments. In this study, we combined Sentinel-2 multispectral imagery and dual-polarized (HH + HV) PALSAR-2 data to generate LCZ maps of Nanchang, China using a random forest classifier and a grid-cell-based method. We then used the classifier to evaluate the importance scores of different input features (Sentinel-2 bands, PALSAR-2 channels, and textural features) for the classification model and their contribution to each LCZ class. Finally, we investigated the relationship between LCZs and land surface temperatures (LSTs) derived from summer nighttime ASTER thermal imagery by spatial statistical analysis. The highest classification accuracy was 89.96% when all features were used, which highlighted the potential of Sentinel-2 and dual-polarized PALSAR-2 data. The most important input feature was the short-wave infrared-2 band of Sentinel-2. The spectral reflectance was more important than polarimetric and textural features in LCZ classification. PALSAR-2 data were beneficial for several land cover LCZ types when Sentinel-2 and PALSAR-2 were combined. Summer nighttime LSTs in most LCZs differed significantly from each other. Results also demonstrated that grid-cell processing provided more homogeneous LCZ maps than the usual resampling methods. This study provided a promising reference to further improve LCZ classification and quantitative analysis of local climate.

Mapping Local Climate Zones and Their Applications in European Urban Environments: A Systematic Literature Review and Future Development Trends

In the light of climate change and burgeoning urbanization, heat loads in urban areas have emerged as serious issues, affecting the well-being of the population and the environment. In response to a pressing need for more standardised and communicable research into urban climate, the concept of local climate zones (LCZs) has been created. This concept aims to define the morphological types of (urban) surface with respect to the formation of local climatic conditions, largely thermal. This systematic review paper analyses studies that have applied the concept of LCZs to European urban areas. The methodology utilized pre-determined keywords and five steps of literature selection. A total of 91 studies were found eligible for analysis. The results show that the concept of LCZs has been increasingly employed and become well established in European urban climate research. Dozens of measurements, satellite observations, and modelling outcomes have demonstrated the characteristic thermal responses of LCZs in European cities. However, a substantial number of the studies have concentrated on the methodological development of the classification process, generating a degree of inconsistency in the delineation of LCZs. Recent trends indicate an increasing prevalence of the accessible remote-sensing based approach over accurate GIS-based methods in the delineation of LCZs. In this context, applications of the concept in fine-scale modelling appear limited. Nevertheless, the concept of the LCZ has proven appropriate and valuable to the provision of metadata for urban stations, (surface) urban heat island analysis, and the assessment of outdoor thermal comfort and heat risk. Any further development of LCZ mapping appears to require a standardised objective approach that may be globally applicable.

Improving Local Climate Zone Classification Using Incomplete Building Data and Sentinel 2 Images Based on Convolutional Neural Networks

Recent studies have enhanced the mapping performance of the local climate zone (LCZ), a standard framework for evaluating urban form and function for urban heat island research, through remote sensing (RS) images and deep learning classifiers such as convolutional neural networks (CNNs). The accuracy in the urban-type LCZ (LCZ1-10), however, remains relatively low because RS data cannot provide vertical or horizontal building components in detail. Geographic information system (GIS)-based building datasets can be used as primary sources in LCZ classification, but there is a limit to using them as input data for CNN due to their incompleteness. This study proposes novel methods to classify LCZ using Sentinel 2 images and incomplete building data based on a CNN classifier. We designed three schemes (S1, S2, and a scheme fusion; SF) for mapping 50 m LCZs in two megacities: Berlin and Seoul. S1 used only RS images, and S2 used RS and building components such as area and height (or the number of stories). SF combined two schemes (S1 and S2) based on three conditions, mainly focusing on the confidence level of the CNN classifier. When compared to S1, the overall accuracies for all LCZ classes (OA) and the urban-type LCZ (OAurb) of SF increased by about 4% and 7–9%, respectively, for the two study areas. This study shows that SF can compensate for the imperfections in the building data, which causes misclassifications in S2. The suggested approach can be excellent guidance to produce a high accuracy LCZ map for cities where building databases can be obtained, even if they are incomplete.

A Multi-Sensor Fusion Framework Based on Coupled Residual Convolutional Neural Networks

Multi-sensor remote sensing image classification has been considerably improved by deep learning feature extraction and classification networks. In this paper, we propose a novel multi-sensor fusion framework for the fusion of diverse remote sensing data sources. The novelty of this paper is grounded in three important design innovations: 1- a unique adaptation of the coupled residual networks to address multi-sensor data classification; 2- a smart auxiliary training via adjusting the loss function to address classifications with limited samples; and 3- a unique design of the residual blocks to reduce the computational complexity while preserving the discriminative characteristics of multi-sensor features. The proposed classification framework is evaluated using three different remote sensing datasets: the urban Houston university datasets (including Houston 2013 and the training portion of Houston 2018) and the rural Trento dataset. The proposed framework achieves high overall accuracies of 93.57%, 81.20%, and 98.81% on Houston 2013, the training portion of Houston 2018, and Trento datasets, respectively. Additionally, the experimental results demonstrate considerable improvements in classification accuracies compared with the existing state-of-the-art methods.

Assessment of Local Climate Zone Classification Maps of Cities in China and Feasible Refinements

Local climate zone (LCZ) maps that describe the urban surface structure and cover with consistency and comparability across cities are gaining applications in studies of urban heat waves, sustainable urbanization and urban energy balance. Following the standard World Urban Database and Access Portal Tools (WUDAPT) method, we generated LCZ maps for over 20 individual cities and 3 major economic regions in China. Based on the confusion matrices constructed by manual comparison between the predicted classes and ground truths, we highlight the following: (1) notable variation in overall accuracies (i.e., 60%-89%) among cities were observed, which was mainly due to class incompleteness and distinct proportions of natural landscapes; (2) building classes in selected cities were poorly classified in general, with a mean accuracy of 48%; (3) the sparsely built class (i.e., LCZ 9), which is rare in the selected Chinese cities, had the lowest classification accuracy (32% on average), and the class of low plants had the widest accuracy range. The findings indicate that the standard WUDAPT method alone is insufficient for generating LCZ products that demonstrate practical value, especially for built-up areas in China, and the misclassification is largely caused by the lack of building height data. This result is confirmed by a refinement test, in which the urban DEM retrieved from Sentinel-1 data with radar interferometry technique was used. The study shows a detailed and comprehensive assessment of applying the WUDAPT method in China and a feasible refinement strategy to improve the classification accuracy, especially for the built-up types of LCZ. The study could serve as a useful reference for generating quality-ensured LCZ maps. This study also examines and explores the relationship between socio-economic status and LCZ products, which is essential for further implementations.

Self-Training Classification Framework with Spatial-Contextual Information for Local Climate Zones

Local climate zones (LCZ) have become a generic criterion for climate analysis among global cities, as they can describe not only the urban climate but also the morphology inside the city. LCZ mapping based on the remote sensing classification method is a fundamental task, and the protocol proposed by the World Urban Database and Access Portal Tools (WUDAPT) project, which consists of random forest classification and filter-based spatial smoothing, is the most common approach. However, the classification and spatial smoothing lack a unified framework, which causes the appearance of small, isolated areas in the LCZ maps. In this paper, a spatial-contextual information-based self-training classification framework (SCSF) is proposed to solve this LCZ classification problem. In SCSF, conditional random field (CRF) is used to integrate the classification and spatial smoothing processing into one model and a self-training method is adopted, considering that the lack of sufficient expert-labeled training samples is always a big issue, especially for the complex LCZ scheme. Moreover, in the unary potentials of CRF modeling, pseudo-label selection using a self-training process is used to train the classifier, which fuses the regional spatial information through segmentation and the local neighborhood information through moving windows to provide a more reliable probabilistic classification map. In the pairwise potential function, SCSF can effectively improve the classification accuracy by integrating the spatial-contextual information through CRF. The experimental results prove that the proposed framework is efficient when compared to the traditional mapping product of WUDAPT in LCZ classification.

Local Climate Zone (LCZ) Map Accuracy Assessments Should Account for Land Cover Physical Characteristics that Affect the Local Thermal Environment

Local climate zone (LCZ) maps are increasingly being used to help understand and model the urban microclimate, but traditional land use/land cover map (LULC) accuracy assessment approaches do not convey the accuracy at which LCZ maps depict the local thermal environment. 17 types of LCZs exist, each having unique physical characteristics that affect the local microclimate. Many studies have focused on generating LCZ maps using remote sensing data, but nearly all have used traditional LULC map accuracy metrics, which penalize all map classification errors equally, to evaluate the accuracy of these maps. Here, we proposed a new accuracy assessment approach that better explains the accuracy of the physical properties (i.e., surface structure, land cover, and anthropogenic heat emissions) depicted in an LCZ map, which allows for a better understanding of the accuracy at which the map portrays the local thermal environment.

Local climate zone-based urban land cover classification from multi-seasonal Sentinel-2 images with a recurrent residual network

The local climate zone (LCZ) scheme was originally proposed to provide an interdisciplinary taxonomy for urban heat island (UHI) studies. In recent years, the scheme has also become a starting point for the development of higher-level products, as the LCZ classes can help provide a generalized understanding of urban structures and land uses. LCZ mapping can therefore theoretically aid in fostering a better understanding of spatio-temporal dynamics of cities on a global scale. However, reliable LCZ maps are not yet available globally. As a first step toward automatic LCZ mapping, this work focuses on LCZ-derived land cover classification, using multi-seasonal Sentinel-2 images. We propose a recurrent residual network (Re-ResNet) architecture that is capable of learning a joint spectral-spatial-temporal feature representation within a unitized framework. To this end, a residual convolutional neural network (ResNet) and a recurrent neural network (RNN) are combined into one end-to-end architecture. The ResNet is able to learn rich spectral-spatial feature representations from single-seasonal imagery, while the RNN can effectively analyze temporal dependencies of multi-seasonal imagery. Cross validations were carried out on a diverse dataset covering seven distinct European cities, and a quantitative analysis of the experimental results revealed that the combined use of the multi-temporal information and Re-ResNet results in an improvement of approximately 7 percent points in overall accuracy. The proposed framework has the potential to produce consistent-quality urban land cover and LCZ maps on a large scale, to support scientific progress in fields such as urban geography and urban climatology.

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

Over the last decade, Kunming has been subject to a strong urbanisation driven by rapid economic growth and socio-economic, topographical and proximity factors. As this urbanisation is expected to continue in the future, it is important to understand its environmental impacts and the role that spatial planning strategies and urbanisation regulations can play herein. This is addressed by (1) quantifying the cities’ expansion and intra-urban restructuring using Local Climate Zones (LCZs) for three periods in time (2005, 2011 and 2017) based on the World Urban Database and Access Portal Tool (WUDAPT) protocol, and (2) cross-referencing observed land-use and land-cover changes with existing planning regulations. The results of the surveys on urban development show that, between 2005 and 2011, the city showed spatial expansion, whereas between 2011 and 2017, densification mainly occurred within the existing urban extent. Between 2005 and 2017, the fraction of open LCZs increased, with the largest increase taking place between 2011 and 2017. The largest decrease was seen for low the plants (LCZ D) and agricultural greenhouse (LCZ H) categories. As the potential of LCZs as, for example, a heat stress assessment tool has been shown elsewhere, understanding the relation between policy strategies and LCZ changes is important to take rational urban planning strategies toward sustainable city development.

Urban Landscape Change Analysis Using Local Climate Zones and Object-Based Classification in the Salt Lake Metro Region, Utah, USA

Urban areas globally are vulnerable to warming climate trends exacerbated by their growing populations and heat island effects. The Local Climate Zone (LCZ) typology has become a popular framework for characterizing urban microclimates in different regions using various classification methods, including a widely adopted pixel-based protocol by the World Urban Database and Access Portal Tools (WUDAPT) Project. However, few studies to date have explored the potential of object-based image analysis (OBIA) to facilitate classification of LCZs given their inherent complexity, and few studies have further used the LCZ framework to analyze land cover changes in urban areas over time. This study classified LCZs in the Salt Lake Metro Region, Utah, USA for 1993 and 2017 using a supervised object-based analysis of Landsat satellite imagery and assessed their change during this time frame. The overall accuracy, measured for the most recent classification period (2017), was equal to 64% across 12 LCZs, with most of the error resulting from similarities among highly developed LCZs and non-developed classes with sparse or low-stature vegetation. The observed 1993–2017 changes in LCZs indicated a regional tendency towards primarily suburban, open low-rise development, and large low-rise and paved classes. However, despite the potential for local cooling with landscape transitions likely to increase vegetation cover and irrigation compared to pre-development conditions, summer averages of Landsat-derived top-of-atmosphere brightness temperatures showed a pronounced warming between 1992–1994 and 2016–2018 across the study region, with a 0.1–2.9 °C increase among individual LCZs. Our results indicate that future applications of LCZs towards urban change analyses should develop a stronger understanding of LCZ microclimate sensitivity to changes in size and configuration of urban neighborhoods and regions. Furthermore, while OBIA is promising for capturing the heterogeneous and multi-scale nature of LCZs, its applications could be strengthened by adopting more generalizable approaches for LCZ-relevant segmentation and validation, and by incorporating active remote sensing data to account for the 3D complexity of urban areas.

Pixel-Based Geometric Assessment of Channel Networks/Orders Derived from Global Spaceborne Digital Elevation Models

Digital Elevation Models (DEMs) contribute to geomorphological and hydrological applications. DEMs can be derived using different remote sensing-based datasets, such as Interferometric Synthetic Aperture Radar (InSAR) (e.g., Advanced Land Observing Satellite (ALOS) Phased Array type L-band SAR (PALSAR) and Shuttle Radar Topography Mission (SRTM) DEMs). In addition, there is also the Digital Surface Model (DSM) derived from optical tri-stereo ALOS Panchromatic Remote-sensing Instrument for Stereo Mapping (PRISM) imagery. In this study, we evaluated satellite-based DEMs, SRTM (Global) GL1 DEM V003 28.5 m, ALOS DSM 28.5 m, and PALSAR DEMs 12.5 m and 28.5 m, and their derived channel networks/orders. We carried out these assessments using Light Detection and Ranging (LiDAR) Digital Surface Models (DSMs) and Digital Terrain Models (DTMs) and their derived channel networks and Strahler orders as reference datasets at comparable spatial resolutions. We introduced a pixel-based method for the quantitative horizontal evaluation of the channel networks and Strahler orders derived from global DEMs utilizing confusion matrices at different flow accumulation area thresholds (ATs) and pixel buffer tolerance values (PBTVs) in both ±X and ±Y directions. A new Python toolbox for ArcGIS was developed to automate the introduced method. A set of evaluation metrics—(i) producer accuracy (PA), (ii) user accuracy (UA), (iii) F-score (F), and (iv) Cohen’s kappa index (KI)—were computed to evaluate the accuracy of the horizontal matching between channel networks/orders extracted from global DEMs and those derived from LiDAR DTMs and DSMs. PALSAR DEM 12.5 m ranked first among the other global DEMs with the lowest root mean square error (RMSE) and mean difference (MD) values of 4.57 m and 0.78 m, respectively, when compared to the LiDAR DTM 12.5 m. The ALOS DSM 28.5 m had the highest vertical accuracy with the lowest recorded RMSE and MD values of 4.01 m and –0.29 m, respectively, when compared to the LiDAR DSM 28.5 m. PALSAR DEM 12.5 m and ALOS DSM 28.5 m-derived channel networks/orders yielded the highest horizontal accuracy when compared to those delineated from LiDAR DTM 12.5 m and LiDAR DSM 28.5 m, respectively. The number of unmatched channels decreased when the PBTV increased from 0 to 3 pixels using different ATs.