Expert-based methods are essentially based on authors’ local expertise and knowledge of specific urban areas, or on general urban surface calculations that are not described in detail in the publication (Lehnert et al., 2021). This method adapts the guidelines set out by (Stewart & Oke, 2012), which includes collecting site metadata, defining the thermal source area and selecting the local climate zone. Consequently, in some cases, the identification of LCZs utilises Google Earth and Bing Maps.

As found in the literature review, an expert-based method was employed by (Thomas et al., 2023) to identify land use within the LCZ scheme in the city of Kochi, India. In addition, (Villadiego & Vellay-Dabat, 2014) mapped LCZs in the city of Barranquilla, Colombia, using Google Earth and Bing Maps, combining in-depth knowledge of urban morphology with field studies to observe and verify the information. (Morris et al., 2015) took the same approach, identifying LCZs based on expert knowledge with the aid of Google Earth in Putrajaya, Malaysia. In contrast, (The ef-fects of urban morphology in a city with a tropical continental climate in the context of heat islands, 2025) deployed 91 observation stations to identify LCZs in the city of Cuiabá, Brazil. Overall, it appears that this method is a traditional one; effective in smaller scale studies, but impractical when applied on a large urban scale.

In principle, LCZ mapping using GIS-based methods utilises urban canopy parameters (UCPs), as defined in Table 1. This approach was introduced by (Lelovics et al., 2014), who defined seven of the 10 UCPs, namely the sky view factor (SVF), building surface fraction (BSF), pervious surface fraction (PSF), impervious surface fraction (ISF), height of roughness elements (HRE), terrain roughness class (TRC) and surface albedo (SAL).

These parameters are obtained from various sources, such as satellite and aerial images, 3D building databases, CORINE land cover databases, road databases and maps (Gál et al., 2015). The method relies heavily upon GIS data, which is not always available in various urban areas, and follows six key steps (Quan & Bansal, 2021), as presented in Figure 8.

In the GIS method, vector-based GIS data collection, such as building footprint data and road cover/path data, is used to generate urban canopy parameters (UCPs), including the sky view factor (SVF), aspect ratio (AR) and roughness element height (HRE) (Quan & Bansal, 2021). The resulting UCP values are then translated into basic spatial units (BSUs) that refer to a specific spatial scale. This procedure is followed by the quantification of each UCP into BSUs based on the LCZ scheme. Post-processing is aimed at collecting units for simplification and size adjustment (Han et al., 2024).

(Quan & Bansal, 2021) grouped LCZ classifiers in GIS-based methods as 1) standard rule-based classifiers; 2) modified standard rule-based classifiers; 3) fuzzy rule-based classifiers; and 4) other classifiers. The standard rule-based classifier is based on the UCP value range proposed by (Stewart & Oke, 2012). The method was employed by (Perera & Emmanuel, 2018), who mapped the LCZs using fish-eye lens photographs, field surveys, detailed maps and Google Earth to identify UCPs in the city of Colombo, Sri Lanka. The LCZ map they produced was also utilised by (Emmanuel et al., 2023).

The method was also used by (Ekanayaka et al., 2025), who identified LCZs using building data, NDWI and NDVI to extract UCP also in the city of Colombo. A modified standard rule-based classifier adjusts the range of UCP values if there is insufficient data to apply the original UCP range. A fuzzy rule-based classifier is based on fuzzy rules designed to address the limitations of standard rule-based classifiers, while, ‘other classifiers’ also aim to address the limitations of standard rule-based classifiers. This other classifiers was employed by (Kotharkar & Bagade, 2018), who identified LCZs using Bhuvan land use and land cover maps from the NRSC, Google Maps, Landsat 7 ETM+ and a primary survey, by applying a new rule-based classifier that adopted standardised methods aligned with qualitative and descriptive parameters to extract UCPs in the city of Nagpur, India. Subsequently, the method was used as the LCZ base map by (Kotharkar & Bagade, 2018), (Kotharkar et al., 2019), (Kotharkar et al., 2019), (Kotharkar et al., 2020), (Kotharkar et al., 2021), (Kotharkar et al., 2022), (Vaidya et al., 2024).

In terms of data sources, GIS streams mostly utilise vector-based GIS data combined with satellite imagery to construct UCP properties. In the literature review, SVF was interpreted through fish-eye lens photographs ((Quan & Bansal, 2021); (Perera & Emmanuel, 2018))), while others UCPs were interpreted using Autodesk Auto-CAD format generated from vector data combined with Landsat 7 ETM+ satellite imagery ((Perera & Emmanuel, 2018); (Kotharkar & Bagade, 2018)). For accuracy assessment, Google Earth and field surveys are used to validate LCZ results.

Overall, this method performs relatively well compared to expert knowledge-based methods. However, it has limitations in terms of the availability of vector-based GIS data, which varies from region to region. In other words, the method is promising for regions with comprehensive vector data.

The Remote sensing-based method is widely popular in LCZ classification due to its applicability, simplicity and availability of free data (Zhang et al., 2024). In our literature review, remote sensing-based methods are categorised into pixel-based and object-based methods (Table 3).

In the LCZ mapping repository, the pixel-based method with supervised classification has gained momentum since (Bechtel et al., 2015) proposed the World Urban Database and Access Portal Tools (WUDAPT) framework to address the limitations of information on the form and function of cities at various spatial resolutions. WUDAPT is a collection of urban canopy modelling and information infrastructure databases produced by the international community to facilitate research applications focused on urban climate, weather, air quality and energy use (Ching et al., 2018). Figure 7 presents the WUDAPT standard LCZ mapping workflow. (Mills et al., 2015) categorised WUDAPT based on data collection into level 0 data, level 1 data and level 2 data. All three types use data from a two-dimensional medium scale to a three-dimensional detailed scale. In this literature review, the majority of studies used the WUDAPT level 0 framework.

The WUDAPT procedure starts with raster processing involving the cropping of the area of interest (ROI) and resampling of the image data into a uniform grid of Landsat 8 OLI-TIRS satellite imagery as raster data. SAGA–GIS software is recommended for the satellite image processing. In addition, (Demuzere et al., 2021) developed the LCZ Generator (Website link) to run the overall WUDAPT framework process.

The next step is to digitise homogeneous surface training data, with a 50% training area and 50% validation area (Bechtel et al., 2018) and a minimum polygon area of 1 km² (Verdonck et al., 2017) using the Google Earth platform. The quality of the resulting LCZ map is highly dependent on the quality of the training area created (minimum average accuracy > 50%), therefore local knowledge is essential in this process. The overall process sequence within the WUDAPT framework is followed by an accuracy assessment. (Bechtel et al., 2018) divided the three approaches to accuracy assessment in WUDAPT into cross-validation, manual review and cross-comparison with other data. In the literature review, the combination of cross-validation and manual review was seen to be the most commonly used approach in LCZ mapping. Cross-validation is calculated using a confusion matrix, which will produce an overall accuracy (OA). The literature review identified that accuracy was assessed using a confusion matrix, and all articles reported LCZ mapping with an overall accuracy of >60%, meeting expectations by (Bechtel et al., 2018).

The final step is the application of random forest classification. The random forest algorithm introduced by (Breiman, 2001) is the best-performing machine learning method for WUDAPT-based LCZ mapping ((Bechtel & Daneke, 2012); (Bechtel et al., 2018)). Although the series of steps outlined above may appear complicated, they are in line with the project’s ambition to provide a global LCZ database quickly and cost-effectively; the availability of the LCZ generator enables local urban experts without experience in RS and GIS to perform LCZ classification within their urban environments. 

In the literature review, the standard WUDAPT procedure developed by (Bechtel et al., 2015)was the most widely used classification system and was employed in the majority of articles. In summary, this procedure is characterised by: 1) pixel-based classification; 2) the use of SAGA GIS software, the LCZ Generator and the Google Earth Engine platform as processing tools; 3) the use of freely available Landsat datasets; and 4) the use of random forest (RF) classifiers. The overall standard procedure results in moderate accuracy, with an overall accuracy (OA) ranging from 50–60% ((Bechtel et al., 2018)), the accuracy of the built-up type being lower than that of land cover.

Several notes are detailed below based on the previous description.

a. Basics of classification. As previously mentioned, the WUDAPT procedure is a pixel-based LCZ classification system. Consequently, most of the documents in this study utilise the standard WUDAPT procedure with a pixel-based classification method to map LCZs. Pixel-based LCZ mapping is constrained by the need to resample to a grid size of 50–200 m (typically 100 m), which results in smaller pixel sizes for identifying LCZ types (Ma et al., 2021). For example, in a Landsat dataset with a pixel size of 30 m, a single pixel is too small to represent a single type of LCZ. Consequently, resampling was conducted to a 100 m grid to accommodate the definition of an LCZ as ‘several hundred metres or kilometres horizontally’, which generally resulted in a fragmented LCZ classification. Therefore, (Brousse et al., 2019) tested the mapping of LCZs using object-based image analysis (OBIA) in cities across sub-Saharan Africa. This approach was also followed by (Simanjuntak et al., 2019), who used the OBIA method to map the LCZ in the city of Bandung, Indonesia. (Ma et al., 2021) and (Yan et al., 2022) propose that OBIA could be a promising method for future LCZ mapping practices. In addition, in the studies of (Fardani et al., 2024), who mapped LCZs in the city of Bekasi, Indonesia; (Nurwanda & Honjo, 2018), who mapped LCZs in Bogor, Indonesia; (Daramola & Balogun, 2019), who mapped LCZs in the city of Akure, Nigeria; and (Wibowo et al., 2023), who mapped LCZs in the city of Bogor, Indonesia none of the four followed the WUDAPT procedure for LCZ classification.

b. Datasets and software. Following standard WUDAPT procedures, most of the documents in the literature review utilised the Landsat dataset series (Landsat 5 TM, Landsat 7 ETM+, Landsat 8 OLI/TIRS, Landsat 9 OLI/TIRS) for the identification of single LCZs and the detection of LCZ changes. Pleiades and SPOT-6 were selected for the OBIA method. Meanwhile, SAGA GIS software, the LCZ Generator platform, and the Google Earth Engine platform are the most popular software options, in line with standard WUDAPT procedures. Therefore, most studies following standard WUDAPT procedures used this software, with the exception of (Choudhury et al., 2021), who mapped the LCZ in the Asansol-Durgapur Development Area (ADDA), India; (Khamchiangta & Dhakal, 2019), who mapped the LCZ in Bangkok, Thailand; and (Kotharkar & Dongarsane, 2024), who used ArcGIS 10.X software to map the LCZ in the city of Nagpur, India. Meanwhile, the classification of LCZ using the OBIA method utilises R software and eCognition software, whilst other LCZ classifications use Quantum GIS and ArcGIS Pro software. 

c. Classifier.(Bechtel & Daneke, 2012) tested several classifiers within pixel-based methods and due to its robustness recommended the random forest (RF) method for use as the classifier in the standard WUDAPT procedure. Consequently, the RF classifier has become popular, and was commonly found in this literature review, alongside the traditional maximum likelihood classification (MLC) used by (Khamchiangta & Dhakal, 2019), (Kotharkar & Dongarsane, 2024) and (Gupta et al., 2024). The RF classifier is also used in the OBIA method. Other documents that did not follow the standard WUDAPT procedure did not specify which LCZ classifier was used. Several strategies have been employed to improve the accuracy of WUDAPT standard mapping, including those of (Vaidya et al., 2024), who utilised deep learning (3DNN and 1D-CNN) and a subset ensemble, which proved capable of improving the overall accuracy (OA) of LCZ mapping compared to a RF classifier.

Overall, the remote sensing-based method in the WUDAPT classification system performs better than the GIS-based method, particularly in spatial and temporal resolution and in its ability to rapidly generate LCZ maps on a global scale. In addition, the OBIA method, which is gaining popularity, can provide new insights into LCZ mapping. Despite its advantages, the RS-Based method has limitations, particularly regarding transferability, spatial and temporal resolution, the quality of cloud-free data (especially in tropical regions), and sample quality; it also requires specialist expertise in remote sensing to operate.

UHI studies dominate many literature reviews, which is expected, given that the LCZ scheme was originally designed as a standardised classification system for UHI research (Xue et al., 2020). UHIs are commonly measured by UHI magnitude/UHI intensity, as discussed by (Stewart & Oke, 2012). In our literature review, this method was applied to study the appearance of UHIs based on the distribution of air temperatures in each LCZ class.

Our review included several UHI studies conducted by various researchers. (Morris et al., 2015)  studied the UHI in Putrajaya (Malaysia) by integrating several LCZ observation points into a Weather Research and Forecasting (WRF) simulation and found the UHI intensity to range from 1.9 to 3.1 °C. On the other hand, (Anjos & Lopes, 2017) mapped the LCZs and linked it to UHI intensity, combined with park cool island (PCI) intensity, to study the dynamics of both in Aracaju (Brazil). They found that the UHI intensity was in the range of 1.5–2°C. A UHI study was also conducted by (Kotharkar & Bagade, 2018) in Nagpur (India), which examined the relationship between different types of LCZs and UHIs, finding that UHI intensity during the winter season ranged from 1.76 to 4.09 °C. (Roth et al., 2022) measured the UHI magnitude for each type of LCZ in Singapore and found that the average range of night-time UHI intensity across seven built-up LCZ types ranged between 1.8 and 3.5°C, with the highest temperatures observed in LCZ 3 (compact lowrise) and LCZ 8 (large lowrise). (Sanchez et al., 2023) used the diagnostic equation proposed in (Theeuwes et al., 2017) to detect the maximum canopy-layer urban heat island (CL-UHI) in Singapore and found that the equation was effective for UHI detection. (Gupta et al., 2024) applied Moran’s I clustering method to study the dynamics of the UHI in Kolkata (India) and found that, between 2003 and 2014, there was an increase in UHI intensity of up to 4°C. In addition, (The ef-fects of urban morphology in a city with a tropical continental climate in the context of heat islands, 2025) conducted a study in Cuiabá (Brazil) and identified an average UHI intensity across all seasons ranging from 0.013°C to 0.53°C. Together with the combination of methods, UHI intensity has also been studied to provide recommendations; for example by (Lefevre et al., 2025) in relation to urban planning on Réunion Island (France), with a difference in UHI intensity of up to 4.14 °C found between urban and natural areas.

Besides detecting the UHI effect based on LCZ type, some researchers have also studied UHI thermal characteristics in relation to the ‘urban’ and ‘rural’ dichotomy in order to recommend the most effective mitigation measures to slow down the effects of climate change. A study conducted by (Kotharkar et al., 2019) using regression analysis with the lmtest package in the R statistical software in Nagpur (India) found that there had been a reduction in UHI intensity, influenced by distance dynamics from the CBD (DI), surface albedo (AL), aspect ratio (AR) and vegetation density ratio (VDR). (Kotharkar et al., 2020) also evaluated strategies involving the installation of cool roofs and pavements using the ENVI-Met model to mitigate the urban heat island effect in Nagpur.

Overall, the use of LCZs for UHI studies in macro-tropical climates is well established. By integrating temperature data from satellite image analysis, field measurements, and mobile measurements, researchers can obtain more precise data, whilst gaining comprehensive understanding of the UHI effect in tropical macroclimate regions. UHI intensity appears to vary across cities. This observation requires more detailed analysis to provide a comprehensive description of the nature of such intensity in the tropical macroclimate regions, thereby aiding decision-making regarding future climate change adaptation.

Understanding the spatial and temporal dynamics of temperature in the context of LCZs is crucial in formulating urban thermal quality improvement strategies (Feng & Liu, 2022). Therefore, this theme has attracted much research, in addition to that on UHIs. Surface temperature monitoring generally concerns the relationship between LCZ and land surface temperature (LST) based on remote sensing methods utilising Landsat datasets (Landsat 5 TM, Landsat 7 ETM+ and Landsat 8 OLI-TIRS), which have thermal bands for surface temperature identification. The utilisation of multi-temporal satellite imagery has become a cornerstone of researchers' efforts to understand the linkages between LCZ changes and their impact on LST changes. In this literature review, urban expansion through such a linkage was conducted by Nurwanda and Honjo (2018) in Bogor (Indonesia) during the period 1990-2017 and by (Khamchiangta & Dhakal, 2021) in Bangkok (Thailand) over the period 1991-2016, and estimated until 2026. In addition, (Simanjuntak et al., 2019) monitored changes in LST due to the development of the city of Bandung (Indonesia) during the period 2013 - 2016, while (Fardani et al., 2024) conducted a similar study in relation to Bekasi (Indonesia) for the period 1993 - 2023, as did (Vaidya et al., 2024) for Nagpur (India) over the summer and winter periods in 2019-2020. A variation to such studies was conducted by (Khamchiangta & Dhakal, 2019), who evaluated the physical and non-physical factors of urban Bangkok (Thailand) that triggered LST changes during the period March to April, 2016. Moreover, (Ochola et al., 2020) analysed Nairobi (Kenya) urban hotspot management through LCZ differentiation at various LSTs during the period 2013 - 2018. Likewise, (Ekanayaka et al., 2025) studied the impact of urban ventilation corridors on LST and wind in Colombo (Sri Lanka) in 2017 and 2020, while (Kabano et al., 2021) studied the relationship between LST, vegetation phenology and LCZ during the period 2013–2015 in Kampala (Uganda). Finally, (Choudhury et al., 2021) investigated thermal behaviour patterns (TBPs) in the industrial Asansol-Durgapur development area in India.

Overall, the majority of studies conclude that there is a strong link between LCZ and LST. Furthermore, changes in LCZ have led to an increase in LST over time. This scientific evidence can assist in future urban planning.

In addition to UHI studies and LST monitoring, researchers have begun to consider other topics related to the LCZ relationship and the spatial and temporal dynamics of temperature, such as heat risk, heat stress and heat disasters. All three conditions can provide early warning of the dangers of increasing urban temperatures. (Emmanuel et al., 2023) defined heat risk as the occurrence of high/extreme temperatures (hazard), mediated by urban morphology including buildings, shade, vegetation, and various types of infrastructure (exposure), applying the concept in Colombo (Sri Lanka). Heat stress is defined as the discomfort and physiological burden of exposure to extreme heat (Wills, 2016) and has been studied using several methods. (Kotharkar et al., 2021) and (Kotharkar et al., 2024) in Nagpur (India) and (Walle et al., 2022) in Kampala (Uganda) assessed heat stress through heat stress by heat index, while (Obe et al., 2024) in Lagos (Nigeria) ran a WRF-urban scheme simulation to assess such stress. In addition to these topics, (Wibowo et al., 2023) investigated the urban heat hazard, considered as the occurrence of extreme temperatures (exceeding 30°C) in the urban area of Bogor that can endanger humans.

Overall, several studies addressing the topics of heat risk, heat stress and heat hazards have demonstrated a relationship between LCZ and threats of heat, that can affect anyone. It is therefore important to take these various hazards into account in all decision-making, particularly with regard to the sustainability of the urban environment.

Thermal comfort is defined as a subjective assessment of human comfort in relation to the surrounding thermal environment (Feng & Liu, 2022) and is achieved by several methods. The parameters generally vary spatially in urban areas as a result of the dynamics of surface energy balance and radiation in different LCZs (Xue et al., 2020). Thermal comfort has been shown to indirectly affect the quality of life in urban areas (Han et al., 2024).

In our literature review, thermal comfort was assessed using common methods such as comparing objective thermal conditions (temperature measurement) and subjective thermal conditions (comfort questionnaires); use of a thermal comfort index; and simulation using SOlar and Long Wave Environmental Irradiance Geometry (SOLWEIG) and ENVI-met. Thermal comfort analysis of tropical regions was pioneered by (Villadiego & Vellay-Dabat, 2014), who assessed outdoor thermal comfort in Barranquilla (Colombia), while (Kotharkar et al., 2019) and (Kotharkar & Dongarsane, 2024) investigated outdoor thermal comfort in Nagpur (India) using different methods. In addition, (Das et al., 2020) assessed outdoor thermal comfort in the Sriniketan-Santiniketan Planning Area (India); (Anjos et al., 2020) analysed sea breeze front and outdoor thermal comfort in Aracaju (Brazil); and (Pereira et al., 2021) examined socio-spatial inequality and its correlation to thermal (dis)comfort in Santos (Brazil).

In short, although thermal comfort studies generally cover a limited area, the representation of each location within different LCZ types can provide an overview of thermal sensations at each study site, whilst also aiding the design of future cities that are climate adaptive.

Globally, nearly 75% of energy supplies are consumed in urban areas, which are home to 55% of the world's population (Kotharkar et al., 2022). Since urban morphology and energy use are related to the interaction between urban form and microclimate, LCZ can be an alternative typology for analysing both (Wang et al., 2024). In addition to energy consumption, knowledge of surface energy distribution, especially in urban areas, is also of interest to researchers, due to its important role in regulating the exchange between land surface and atmospheric energy. Therefore, our review addresses both topics. (Kotharkar et al., 2022) studied the energy consumption of the building sector with the LCZ approach using building models for energy simulation in Nagpur (India). Likewise, (Pokhrel et al., 2019) built a scenario of building cooling load reductions for sustainable energy with building effect parameterisation and building energy model (BEP-BEM) simulation schemes in the San Juan Metropolitan Area (Puerto Rico). Similarly, (Daramola & Balogun, 2019) and (Obe et al., 2024) focused on investigating surface energy flux by simulating the surface energy balance algorithm for land (SEBAL) and surface urban energy and water balance scheme (SUEWS) models in the cities of Akure and Lagos (Nigeria).

Overall, various studies conducted in different locations indicate that LCZs are effective in monitoring energy consumption (particularly built-up types) with a view to creating cities that are resilient to climate change.

In addition to field observations and satellite monitoring, the integration of regional climate models (RCMs) and urban canopy parameters (UCPs) in specific regions is an important method for urban environmental research (Wang et al., 2024). Urban climate modelling can help describe current and future urban climate conditions while urban climate simulation has been partially discussed in relation to topics such as UHI, thermal comfort and energy. In the literature review, in general urban climate modelling is performed to understand overall urban climate conditions. (Mughal et al., 2019) utilised the Weather Research and Forecast (WRF) model to understand the urban climate conditions of Singapore during the hot period of April 2016; (Mancheno et al., 2024) used the same model to simulate the meteorological conditions of Quito (Ecuador); while (Patel et al., 2020) also employed the model to predict very heavy rainfall events in Mumbai (India). In addition, (Brousse et al., 2020) utilised the COSMO-CLM model to assess and understand the impact of recent urbanisation on the (surface) urban heat island ((S)UHI) under clear-sky conditions in Kampala (Uganda). Similarly, (Walle et al., 2021) used the model to illustrate the influence of urban parameters on urban climate formation in Kampala (Uganda).

In general, the simulation models developed by the researchers mentioned above indicate that LCZs can serve as tools for predicting future urban climate dynamics. This is important in order to provide input to decision-makers so that they do not overlook climate-related factors that could contribute to multi-hazard risks.

The various climate-related issues in the tropical macroclimate regions described above would not pose a serious problem if climate-inclusive urban planning were implemented from the outset. Recognising this, researchers need to consider climate-resilient urban planning solutions to ensure the sustainability of cities in the future. Therefore, local characteristics (natural and cultural landscapes) must be taken into account when planning a city. In this literature review, a study of the prospects of LCZs in urban planning was conducted by (Perera & Emmanuel, 2018) in Colombo, Sri Lanka, who demonstrate how the LCZ framework can be utilised to improve urban planning. Although only one relevant urban planning study was identified, the co-occurrence of the keyword “urban planning” more than three times suggests that the various themes addressed in the different studies will contribute to climate-resilient urban planning.