The study area is located in Shenyang, the capital of Liaoning Province, in northeastern China (122°25′-123°48′E, 41°12′-43°2′N). The terrain is relatively flat and mainly consists of plains. The eastern boundary extends from the hills of eastern Liaoning, and the western region comprises the alluvial plain formed by the Liao and Hun rivers. Shenyang has four distinct seasons and a temperate semi-humid continental climate with an annual average temperature of 6.2–9.7°C. Precipitation occurs mostly in summer, often in the form of torrential rains in July and August. The study area is within the Fourth Ring Road of Shenyang (Figure 1), which is the central urban area of socio-economic development in Shenyang. To balance the processes of urban development and the construction of human settlements, it is essential to study the thermal environment.

The data includes land use, Landsat-8 images, MODIS images, building vectors, and other auxiliary data. The detailed attributes and sources are listed in Table 1. A land use and cover change data including secondary classification is used. The land use data in this study have been recorded every 5 years. We used one of its pictures in 2015. To ascertain the actual land use in the study area, the land use data of 2015 were partially updated using the multi-spectral image data of Landsat-8 in 2018. Using manual visual interpretation, the green space contour of some large parks in Shenyang was extracted. Finally, potential features classified by LCZ are reclassified into seven types in the figure. The updated land use data are shown in Figure 1.

We used remote sensing data, land use data and building vector data to calculate LST and divide LCZs. The concept of first-level grid was introduced to calculate the area proportion of LCZ. The correlation was calculated for area proportion of LCZ and LST. The stepwise regression model was used to analyze the warming and cooling effects of LCZ under the combined influence. The frame figure is shown in Figure 2.

LCZ classification results of the study area are shown in Figure 4A. The study area contained 12 of the total 13 LCZ types, excluding the category of LCZF (bare soil or sand). Overall, for the construction area: the overall proportion of LCZ1–LCZ6 was 19.3%, and the overall proportion of LCZA–LCZG was 80.7%. The proportion of all LCZs, from high to low, was as follows: low plants LCZD (44.4%) > bare rock or paver LCZE (30.0%) > open middle-rise building LCZ5 (5.8%) > open low-rise building LCZ6 (5.0%) > compact low-rise building LCZ3 (3.5%) > compact middle-rise building LCZ2 (3.2%) > water LCZG (2.5%) > dense trees LCZA (2.4%) > open high-rise building LCZ4 (1.2%) > scattered trees LCZB (1.0%) > compact high-rise building LCZ1 (0.6%) > bush and scrub LCZC (0.4%).

The calculated LSTs of the study area are shown in Figure 4B. In the summer of 2018, the LST in the urban area within the Fourth Ring Road of Shenyang was between 29.08 and 45.65°C, and the average LST was 36.88°C. Approximately 95% of the pixels were at temperatures within the range of 31.54–41.2°C, with few extreme temperature pixels.

Figure 4C shows the temperature characteristics of various LCZs. In the LCZ category of the built-up area, the average surface temperature (TsMean) was higher than 38°C. The LCZ type with the highest TsMean was LCZ2 compact middle-rise building (39.82°C) and that with the lowest TsMean was LCZ4 open high-rise building (38.51°C). The highest statistical range of the inter-class mean surface temperature in the LCZ of the built-up zone was 13.5°C, corresponding to LCZ6. In the natural area LCZs, the highest TsMean type was LCZE (38.51°C) and the lowest was LCZG water body (34.21°C), which also had the lowest average surface temperature among all LCZ types.

In the case of building class LCZ, the low- and middle-rise building classes (LCZ2, LCZ3, LCZ5, and LCZ6) had maximum and average temperatures higher than that of the high-rise buildings (LCZ1 and LCZ4) because tall buildings tend to cast shadows, which reduces the surface temperature; this is consistent with the findings of other studies (56). However, the minimum temperature of the high-rise buildings was higher than that of the middle- and low-rise buildings, under the corresponding BD, which may be due to the relatively high warming effect of LCZ types with high average BHs. The LCZ type with a high BD contributes more to the LST, and the warming effect is more pronounced.

To establish the relationship between LCZ and the thermal environment, the average temperatures within LCZs are usually used to estimate the thermal characteristics (32). We also applied this method to analyze the thermal environment characteristics of Shenyang in summer, as described in section 3.1. Previous studies show that many factors within the LCZ can be quantified, such as NDVI, BH, the proportion of impervious water surface, and the average tree height using this method (26, 41). These studies elucidated the influence of internal characteristics of LCZ on LST. However, they estimated the impact of each type of LCZ on LST in an independent state. Therefore, when the correlation between LCZ1 and LST is analyzed, the area and location of other LCZs, such as LCZ2 and LCZB, do not have any impact on the correlation results. But in fact, LCZs do not exist independently in the city; they are adjacent to, or intersect with, other LCZs. Therefore, an important but unanswered question was whether LCZs had combined or competitive influence on LST. To answer this question, we proposed to use the area proportion to describe LCZs. In this method, we demarcated a series of first-level grids and calculated the area proportion of LCZs.

As the area proportion is competitive, all LCZs are no longer independent LST-related variables, but their changes are closely related to other LCZs. They are placed in an environment with a total of “1,” competing for influence on the LST. The results of the calculated area proportion are described in section 3.2. However, the relevant analysis method is independent and does not reflect the competition and combined influence of LCZs. Therefore, the ranking of thermal characteristics measured by average temperature is consistent with the correlation. As all LCZs were included in the regression model, they required a simultaneous impact on LST, and the results changed. We found that the warming effect of LCZE and the cooling effect of LCZG might have been underestimated due to the neglect of the combined and competitive effects of LCZs in the past. It should be noted that these results apply only to the current LCZ configuration in Shenyang. However, the method of using area proportions to analyze the combined effect of LCZs on LST can be applied to other regions to explore the factors that have been neglected in the independent analysis.

We can find that the influence of different LCZ categories on the LST, calculated by a stepwise regression (section 3.3), is not completely consistent with their thermal characteristics (i.e., the corresponding average LST of each LCZ category; section 3.1), or the degree of correlation (section 3.2). The average temperature ranking of LCZ was a description from a regional perspective not involving the interaction between LCZ categories. The combined effect emphasized that LCZ impacts LST at the same time, considering the interaction between different categories. The differences reflected from the two perspectives suggest the presence of combined effect. This may be associated with the change in heat storage capacity caused by the arrangement of LCZ. Although LCZ2 has the highest mean surface temperature, the regression results show that LCZE has a higher warming effect on LST than LCZ2 across the four scales from 270 to 810 m. Previous studies focused on the heat generated by building coverage LCZs to the urban thermal environment. The strong warming effect produced by LCZE has not been discussed. Meanwhile, despite the low surface temperature of LCZG, it failed to produce a sufficient cooling effect in the stepwise regression model. A different perspective on the LST contribution can explain this contradiction. The average temperature of an LCZ shows its thermal state, which cannot be used to estimate the actual contribution to urban surface temperature because its area and distribution are not considered. The standardized coefficient obtained by stepwise regression modeling represents a relative state and is used to calculate the degree of LST change caused by LCZ. Combining the two factors can better analyze the urban thermal environment. The cooling of the city alone cannot explain the local minimization; instead, it is essential to consider the combined influence of various factors to achieve the optimal effect.

To achieve the goal of urban cooling, many studies have been conducted on land use adjustment, planning, and construction (15, 57). For example, in China's Beijing-Tianjin-Hebei, Pearl River Delta, and Yangtze River Delta urban agglomerations, the LST of the compact medium- and high-rise buildings is high, which is not conducive to urban cooling (40). However, green spaces and water bodies can be added to compact high-rise buildings to meet the cooling requirements (58). Optimizing ventilation corridors and increasing the area of green space and water bodies can cool the city (59–61).

This study provides some new cooling ideas based on the contribution of LCZ to LST. The thermal characteristics (i.e., average LST) and their ranking based on temperature-increasing ability (i.e., standardized coefficient) of different LCZs were determined. For LCZ types with high average temperatures, such as building-type LCZs and LCZE, measures should be taken to reduce their LST. Increasing the greening between buildings and optimizing roof materials or ground paving materials may play a cooling role. To alleviate the influence of high-temperature, reduce the LST, and achieve the purpose of cooling the city, we should explore the cooling potential of trees in the city (62, 63). For LCZs with a strong warming effect, such as LCZ2, LCZ3, and LCZE, the excessive increase should be avoided in the process of urban planning. Simultaneously, a database on the variation of the warming effect of different LCZs in cities should be established. The thermal characteristics and warming effect of LCZs should be recorded and analyzed regularly to provide a reference for the thermal environment background of future urban construction.

For the study area, the main central high temperature area is continuous. This is similar to the distribution trend of LCZE. Therefore, in the future urban planning of Shenyang, we proposed building greenways, increasing LCZD and reducing the combined warming effect caused by impervious surface. On a smaller scale, especially in the built-up areas of the city center, including Shenhe, Heping, and Huanggu Districts, we propose the construction of pocket parks to improve the cooling capacity.

To facilitate the calculation of LCZ area proportion, we preferentially choose the 30 m LCZ division, which may cause the spatial discontinuity of LCZs. With respect to the classification of LCZs in this study, there is still room for the improvement in classification accuracy owing to the timeliness and availability of data (64). Results can be improved by optimizing the LCZ classification and area proportion calculation methods.

Using only the area proportion as a measure of LCZ cannot reflect the spatial location attributes of LCZ and the intra-type differences of LCZ. Our study could not reveal whether the spatial distribution of LCZs affects LST, which requires some new models or perspectives to be considered, such as the geographic weighted regression model and landscape pattern index (65, 66).

Finally, although we could quantify the relative contribution, the causes have not been identified. It may still be necessary to consider the mutual relationship and competitive impact between LCZs, which will be addressed in future research.