**5. Discussion**

In our study, nighttime light data were corrected to characterize the changes in urbanization intensity by the DN value and various functional regions of the city were extracted to express the various stages of urbanization. Combined with the vegetation index data, we quantitatively analyzed the contribution of urbanization and vegetation to variations in albedo. The contribution analysis method used in this study is a partial derivative method, which is widely used in studies on the effects of climate on hydrological dynamics [77,78] and studies on climate response [74,79]. Based on the results of this quantitative analysis, we concluded that the significant increase in the urbanization contribution and the decrease in the vegetation contribution after 2005 were the main reasons for the significant decreasing growth rate of albedo after 2005. In terms of mechanisms, this was consistent with previous conclusions on urban albedo; that is, urbanization could cause a decrease in albedo, which is generally correlated to the surface roughness. For example, albedo observations based on model experiments from Aida [16] showed that multiple reflections of solar radiation in urban canyons increased the absorption of solar radiation in cities, resulting in a reduction in urban albedo. Kondo [80] used the Monte Carlo ray tracing method to show that building height affects albedo, and low-rise buildings have a high albedo. The impact of urban areas, which are one of the most densely populated areas, on urban albedo is multi-fold [81–84]. On one hand, urbanization is accompanied by changes in land cover. In general, the process of transition from a village to a city involves replacing natural surfaces (e.g., farmlands and forests) with impervious surfaces (e.g., cement and asphalt). Due to the changes in the thermal conductivity of the Earth's surface, albedo changes, and the water and heat exchange between the Earth's surface and the atmosphere also changes. The 3D solid surface formed during urbanization has resulted in an increase in surface roughness [16,76,80,85,86] and solar radiation absorption [77], which share the same mechanism with soil roughness and soil albedo. Inner spaces enable the multiple reflection of lights, which increases the absorption of radiation. For this reason, urban areas usually have low albedo [8]. The multiyear average of reflectivity calculated by MODIS albedo products also showed this rule; that is, that albedo in urban areas is generally low (Figure 2). On the other hand, the ability of cities to attract people is also obvious. Urban areas account for approximately 0.5% of the total land area in the world, but they accommodate more than half of the world's population [87]. Due to the complexity and uncertainty of the human activities during the urbanization process, it is very difficult to identify the urbanization effects. The DN value of the night-time lights data is used as the index of urbanization intensity, which could show comprehensive

impacts of human activities. This enables us to simplify the impact of urbanization on albedo changes and helps us to quantify the contribution of urbanization and vegetation to the changes in albedo from a macro perspective.

As the parameter indicating the surface's ability to reflect solar radiation, albedo plays a key role in the energy balance at the surface, and the effects of albedo on climate change have raised substantial attention from many scholars. Akbari et al. [88] simulated the long-term effects of urban albedo growth using a mesoscale complex global climate model (UVic Earth System Climate Model), and it was believed that an increase in surface albedo by 0.01 over a square meter could reduce the long-term global temperature by 3 × 10−<sup>5</sup> K, which was equivalent to reducing CO2 emissions by 7 kg [89]. Sailor [8] analyzed the surface albedo in Los Angeles based on a three-dimensional meteorological model and found that albedo increased by 0.14 in urban regions and 0.08 in basin areas, which could reduce the maximum heat by 1.5 ◦C in summer. Based on a mesoscale atmospheric model, Humdi [40] analyzed the intensity of UHIs and found that the increase in albedo over three types of urban surfaces (walls, roofs, and roads) could reduce the UHI both during the day and at night. Wang et al. [89] analyzed the impact of land use change in urban regions on extreme heat events with the Weather Research and Forecasting model (WRF) in Jing-Jin-Ji and found that an increase of the albedo on urban roofs from 0.12 to 0.85 could reduce the urban mean temperature by 0.51 ◦C, which was equivalent to 80% of the heat caused by urban expansion in the last 20 years. Menon [7] increased the albedo of roofs and roads using the GEOS-5 basin surface pattern, and the result showed that increasing the albedo of roofs and roads by 0.25 and 0.15, respectively, could result in approximately 57 Gt of CO2 from global urban regions. Compared with the aforementioned study, it is clear that the decrease of albedo (approximately 0.05) in the Jing-Jin-Ji region, caused by the increasing urbanization contribution and the decreasing contribution of vegetation over 2001–2011, is in a relatively reasonable numerical range, and the effects of this variation in albedo on urban temperature are not negligible. Urbanization could change both the urban morphology and the urban environment. Energy-budget parameters are also varied during this process. The heat-trapping morphology of the 3-D surface and the reduced areas of vegetation and water bodies both contribute to albedo variation and could result in urban climate change [82,90]. The decrease of the albedo (~0.05) in our result showed that there might be a high possibility of temperature changes due to urbanization, and it could also affect the UHI in this area. In this way, the difference between the vegetation and urbanization contribution to albedo might be useful in urban planning to mitigate the intensity of the UHI, and helps to offer better comfort conditions to residents [91–93].

Because of the significant influence of albedo on temperature in urban areas, increasing the albedo in urban areas to mitigate the urban heat island intensity has become an important aspect of urban energy conservation research. Taha [94] found that the changes in urban albedo via whitewashing could save 35% of the cooling peak power and 62% of the cooling energy. The research of Akbari [95] also showed that urban trees and high albedo could potentially reduce air conditioning energy by 20%, which would save about \$10 billion a year in energy costs and help improve urban air conditions. Therefore, it is necessary and meaningful to understand the reason why urban albedo changes during the process of urbanization, which would be helpful for future studies on urban climate change and for the development of urban energy conservation strategies.

There have been many detailed studies on the factors that may influence albedo, such as the solar zenith angle [16], the underlying surface regime [21], soil moisture [96,97], and meteorological conditions [98–100]. Based on a mathematical statistics method, our study quantitatively calculated the contributions of multiple factors to urban albedo in each pixel and identified the main contribution factors, which is one of the highlights of our study. However, our study is still insufficient. Because the basis of the calculation method in our study was a comprehensive differential equation, we cannot evaluate the uncertainty of the results. Instead, we can only compare the results with other studies or use other auxiliary data to validate the reliability of our conclusions. Second, one of our study conclusions was that the contribution of albedo in the Expanded Area that also includes the new urban areas and in the suburb-located Fringe Area was mainly affected by urbanization. Since our study simplified the urbanization process, more specific reasoning, such as why urbanization is the dominant factor, still requires more detailed scientific research in the future. Third, we used the DN value of the DMSP/OLS nighttime light data to represent the urbanization intensity, but whether or not this index can assess urban development levels accurately has not been evaluated. In addition, there are many types of nighttime light correction methods, but the methods used in Cao [63] were applied in this study due to the adequate correction effect and more applicable study area (China). However, whether or not this method can be applied at a global scale remains to be explored. Finally, as we have only taken the fastest population growth period (~10 years) into consideration, the length of the time series is relatively short. Therefore, the impact of data length still needs to be evaluated.
