*4.2. Spatial Distribution of USES*

The spatial distribution of USES of the selected cities was heterogeneous (Figure 5). A visual survey of the RSUSEI maps shows that Chicago and Los Angeles had higher RSUSEI values than Minneapolis, Dallas, Phoenix, and Seattle. Areas with high values of RSUSEI (red color) had a lower quality of USES, which had high heat (LST), imperviousness (ISC) dryness (NDSI), greenness (NDVI), and moisture (Wetness) values, and vice versa. Figure 6 shows the mean value of RSUSEI for Minneapolis, Dallas, Phoenix, Los Angeles, Chicago, and Seattle to be 0.58, 0.54, 0.47, 0.63, 0.50, and 0.44, respectively. The difference in the mean value of RSUSEI between the cities indicated a significant difference in their USES. The best and worst USES were Seattle and Los Angeles, respectively. The SD values of RSUSEI for Minneapolis (0.16), Dallas (0.17), Phoenix (0.19), Los Angeles (0.21), Chicago (0.17), and Seattle (0.19) cities were high. These values indicated the high spatial variability of the USES within each selected city. Overall, the SD values of the different cities were very similar. The highest and lowest spatial variations of USES were observed in Minneapolis and Seattle, respectively.

**Figure 5.** Remotely Sensed Urban Surface Ecological index (RSUSEI), classified RSUSEI and NLCD land cover maps of the selected cities in the U.S.A.

**Figure 6.** Frequency of RSUSEI values of the selected US cities.

The spatial distribution of the RSUSEI classes further revealed the spatial variability of USES across the selected cities (Figure 7). Overall, the majority of the land in the selected cities possessed the USES class from Very Good to Fair. The highest percentage of USES class for Minneapolis, Dallas, Phoenix, Chicago, and Seattle cities was Good and for Los Angeles city was Fair. The Poor class of USES had better spatial coverage of 5% to 14%, compared to that of the Excellent class from 1% to 6%. In addition, the highest percentage of Excellent, Very Good, Good, Fair, and Poor classes of USES was observed in Los Angeles, Seattle, Chicago, Minneapolis, and Los Angeles, respectively (Figure 7). The spatial heterogeneity of surface biophysical characteristics and anthropogenic activities caused differences in USES among the cities and within each city.

**Figure 7.** Area of USES classes of selected cities in the U.S.A. (%).

It is worth noting that the Eigenvalues of the PC1 in RSUSEI modeling for the Minneapolis, Dallas, Phoenix, Los Angeles, Chicago, and Seattle cities were 0.33, 0.34, 0.40, 0.39, 0.29, and 0.37, respectively. For all selected cities, PC1 included more than 93% of the main surface information including greenness, moisture, dryness, heat, and imperviousness. Therefore, using PC1 in RSUSEI can well represent the spatial heterogeneity is the USES.

#### *4.3. Association Degree of Surface Biophysical Parameters on the USES Modeling*

The mean value of RSUSEI varied across different land covers in the selected cities (Table 5). The mean values of RSUSEI in "Developed, Open Space", "Developed, Low Intensity", "Developed, Medium Intensity", and "Developed, High Intensity" lands were 0.35, 0.49, 0.63, and 0.76, respectively (Table 6). In general, "Developed, high intensity" and "Developed, Open Space" lands detected the highest and lowest RSUSEI values in these cities, respectively. This result suggests the effectiveness of RSUSEI to separate USES by land cover.


**Table 5.** The mean RSUSEI of land cover classes of selected cities in the U.S.A.

**Table 6.** Correlation coefficient between RSUSEI and nLST, nNDVI, nNDSI, and nWetness.


Areas with high surface heat (LST), imperviousness (ISC), dryness (NDSI), low surface vegetation density (NDVI), and moisture (Wetness) exhibited poor USES. The locations of these areas corresponded to "Developed, High Intensity" lands (Figure 5). By contrast, "Developed, Open Space" lands possessed the best USES (Table 4), which discovered low values of LST and NDSI and high values of NDVI and Wetness. Mixed pixels in warm and dry cities tended to include built-up lands and lands with low vegetation cover and low moisture content. Due to the high values of RSUSEI for lands with low vegetation density and low moisture content, these cities discovered poor USES. On the other hand, mixed pixels in humid cities were associated with high vegetation density and surface moisture. Since there were low values of RSUSEI for lands with high vegetation density and high surface moisture, these cities experienced poor USES. These findings suggest that RSUSEI holds an excellent ability to differentiate between USES of different land covers (Figure 4 and Table 5).

The association degree of nLST, nNDVI, nNDSI, and nWetness in RSUSEI in the selected cities varied. The mean r between nLST, nNDVI, nNDSI, and nWetness and RSUSEI was 0.47, −0.31, 0.17, and −0.27, respectively (Table 6). The statistical significance of these correlations were confirmed at 95% confidence level. In modeling the USES, the association degree of nLST was found to be higher than nNDVI, nNDSI, and nWetness. For the RSUSEI modeling in Los Angeles, Chicago, and Seattle cities, the association degree of nNDVI appeared higher than nNDSI and nWetness. In contrast, in Minneapolis and Dallas, the association degree of nWetness was higher than nNDVI and nNDSI.

The correlation coefficient between the mean value of RSUSEI and the mean value of NLCD imperviousness percentage was 0.93 for all cities, but it varied within cities. Minneapolis, Dallas, Phoenix, Los Angeles, Chicago, and Seattle yielded an r value of 0.90, 0.99, 0.99, 0.99, 0.98, and 0.98, respectively (Table 6). The statistical significance of these correlations are confirmed at 95% confidence level. These values indicated a positive strong correlation between ISC and RSUSEI. The spatial variation patterns of RSUSEI and NLCD imperviousness were similar to each other (Figure 5). The RSUSEI value increased with increasing the percentage of impervious surface. In the

USES modeling, the association degree of ISC was highest among all the surface parameters used in this study.

#### **5. Discussion**

The SES in urban environments is a function of the surface biophysical, biochemical, and biological properties. Recent studies have used the data of surface greenness, moisture, dryness, and heat for SES modeling [4,6,7]. However, many processes in the urban environments are subject to the impact of surface imperviousness [23,44–46]. ISC has a clear physical meaning in land surface composition, suitable for distinguishing the characteristics of different types of land use and land cover in the urban environments, and is associated with changes in the characteristics of surface greenness, moisture, dryness, and heat [19,20,22,45]. This study shows that the association degree of imperviousness is higher than surface heat, greenness, dryness, and moisture. Therefore, considering surface imperviousness information in USES modeling is very necessary. Other studies have also shown that surface imperviousness affected the SES [4,47,48]. For many cities around the world, surface imperviousness data are available with functional spatial resolution for urban modeling. In this study, five components including surface greenness, moisture, dryness, heat, and imperviousness are considered for RSUSEI development. Results showed that RSUSEI is highly capable in the modeling of the USES spatial heterogeneity in cities with different geographical, climatic, environmental, and biophysical conditions. This index has a high capacity to differentiate between USES of different land covers. Assessment and modeling of USES are crucial in sustainability assessment in support of achieving sustainable development goals such as sustainable cities and communities [8]. Hence, RSUSEI can be used for assessing urban sustainability over space and time.

#### **6. Conclusions**

In this study, an analytical framework is proposed for assessing the SES in urban environments and tested in six selected cities in the U.S.A, i.e., Minneapolis, Dallas, Phoenix, Los Angeles, Chicago, and Seattle. This analytical framework is centered on a new index, Remotely Sensed Urban Surface Ecological index (RSUSEI), which integrated satellite derived information on the greenness, moisture, dryness, heat, and imperviousness in a city. The results showed that the spatial distribution of USES varied with the cities and land cover types. In general, land covers with low vegetation density and moisture, and high heat, imperviousness, and dryness exhibit high RSUSEI values and poor USES, and vice versa. The USES in arid regions, such as Los Angeles, are found to be worse than the USES in humid regions, such as Seattle. The association degree of ISC is higher than nLST, nNDVI, nNDSI, and nWetness in the RSUSEI modeling. An increase in surface imperviousness reduces surface vegetation density and moisture while increasing surface dryness and heat degree, thereby worsening USES. Our results show that RSUSEI has a high capability in revealing the differences in USES within and between cities with different geographical, climatic, environmental, and surface conditions. Due to the functional spatial resolution and continuity of Landsat imagery, the results of this study can be very useful in USES modeling in urban environments with different biophysical, geographical, and climatic conditions. In addition, the availability of NLCD data products in the U.S.A is highly beneficial for USES assessment, monitoring, and modeling. RSUSEI can be used for assessing urban sustainability over space and time. It is suggested that in future studies, the efficiency of disaggregation models in improving the spatial resolution of USES maps should be considered. It is also useful to compare the performance of different spectral indices in surface imperviousness modeling to assess USES. In addition, RSUSEI can be used as a time series to monitor and model the long-term changes in a region and to quantify the impact of anthropogenic activities on USES.

**Author Contributions:** M.K.F. and S.F. conceived and designed the research of, and wrote, the first draft; Q.W. re-designed the research, and revised and edited the paper; M.K. and S.K.A. provided comments. All authors contributed to, and approved, the final manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.
