Spatiotemporal Features of the Surface Urban Heat Island of Bacău City (Romania) during the Warm Season and Local Trends of LST Imposed by Land Use Changes during the Last 20 Years

Abstract

Using MODIS and Landsat LST images, the present paper advances a series of results on the characteristics of the surface heat island (SUHI) of Bacău City (Romania) during the warm season (April to September) for a period of 20 years (2001–2020). At the same time, given their higher temporal resolution and their availability for both day and night, MODIS LST was used to understand the spatial features of the SUHI in relation to land use. In this way, a total of 946 MODIS Terra and 483 Landsat satellite images were used to outline the main LST characteristics of the days with clear sky in this middle-sized city in northeast Romania. In order to analyze MODIS LST changes in relation to land use changes in the period 2001–2018, we used the standardized CORINE Land Cover datasets. With the help of the Rodionov test, we were able to determine the geometry and intensity of the SUHI. During the day, the spatial extension of the SUHI reaches its maximum level and is delimited by the isotherm of 31.0 °C, which is 1.5–2.0 °C warmer than the neighboring non-urban areas. During the night, the SUHI has a more regulated spatial extension around the central area of the city, delimited by the 15.5 °C isotherm with LST values that are 1.0–1.5 °C warmer than the surrounding non-urban areas. Additionally, from a methodological point of view, we highlight that resampled MODIS and Landsat images at a spatial resolution of 500 m can be used with confidence to understand the detailed spatial features of the SUHI. The results of this study could help the elaboration of future policies meant to mitigate the effects of urbanization on the SUHI in an era of increasing air temperatures during summer.

1. Introduction

The combined effects of rapid urbanization and ongoing global climate change lead to an overexposure of the urban population to the unbearable effects of more intense and long-lasting heat waves [1,2,3,4]. These combined trends also drive the urban environments to the amplification of the existing urban heat island (UHI) effects or to the development of new urban heat spots.

The UHI is generally defined by the thermal difference between the air temperature values measured at a weather station or temperature monitoring point in the quasi-central area of the urban environment and other measurements points located in the neighboring rural or cropland areas that are not significantly influenced by the UHI [5,6,7]. The interest in the UHI topic is generated mainly to the fact that it causes an increase in outdoor bioclimatic discomfort [8,9] in the densest parts of the cities, directly impacting human health and quality of life, but the effects of UHI manifestation are felt also at the level of each building through the increase in the energy necessary for mechanical ventilation and, implicitly, greenhouse gas emissions [1,2]. This thermal discomfort is amplified in urban areas during the summer season of the temperate zone [8,10,11,12], with a tendency to prolongate discomfort intervals toward the end of the spring and the beginning of the autumn.

The UHI assessment is classically performed by measurements in the cross-section through the city, covering both the center and the periphery of the city, carried out 2 m above the ground [3,13,14]. The second method, widely used nowadays and also applied in the present study, aims to quantify the UHI using satellite products that define the surface urban heat island [15,16]. In this case, the input temperature used to define the UHI is represented by the land surface temperature products, and the result is represented by the surface urban heat island (SUHI). The increase in satellite image accuracy and availability, and also the reduction in their costs during past years [17,18,19], enabled their large-scale use in climate studies, especially in the field of urban climatology. The main limitation of the SUHI assessment is given by the fact that LST products are available only for some moments during the day and at different spatial resolutions. Due to this issue, comparing the SUHI features obtained using LST products from different satellites, and even comparing these results with in situ observations, became mandatory in urban climate studies.

LST products used to describe the SUHI can be derived from many satellite platforms. After the launch of the Terra satellite (EOS AM-1) in 2001, a series of climate studies based on MODIS satellite products have been published [20,21], and LST products provided by MODIS sensors were used extensively to study atmospheric conditions in urban environments, including the SUHI [18,22]. In spite of its low spatial resolution, MODIS LST products have the advantage of their high temporal resolution, covering both daytime and nighttime intervals and helping to obtain a robust image of SUHI’s main features.

For Romania, some studies were performed aiming to analyze the UHI for some of the main cities mainly based on MODIS satellite products [16,23,24,25,26,27], but Landsat has also been used [28].

This paper advances a series of results regarding the characteristics of the surface urban heat island (SUHI) in the city of Bacău, a medium-sized city in northeast Romania, during the warm season (April–September) for a period of 20 years (2001–2020), using MODIS satellite products. It is to be noticed that medium-sized cities, defined as cities having a population between 200 and 500 k inhabitants [29], as it is Bacău city with its urban area—nowadays gather a large part of the urban population worldwide [30] and are considered not being properly assessed concerning their UHI, including SUHI-related aspects. The obtained results accurately describe the intensity, geometry, and limits of the SUHI, and also its relationship with the categories of land use in the urban environment. We also quantified the thermal impact determined by land use changes that occurred in these last 20 years. Additionally, for a better characterization of the SUHI, we have compared the MODIS LST daytime results with those obtained from Landsat LST images. In this way, our study proposes a way to harmonize the interpretation of the SUHI features obtained from these two LST products.

2. Data and Methodology

2.1. Study Region and Its Main Thermal Conditions

Regarding its geographical position, Bacău city is located in northeastern part of Romania (Figure 1) and, including the settlements in its near surroundings, it has more than 200k inhabitants. The heart of the city is developed mainly on the lower terrace of the Bistrița River, but in the past decades it exhibited a large spatial extension in the major riverbed of this river. The Bistrița River lowland is divided into two terraces, a lower one, with relative altitudes between 2–4 m, and an upper, higher one, whose terrace has altitudes between 7 and 10 m [31]. The middle terrace with relative altitudes of 2–4 m extends from the city center toward the eastern and lower neighborhoods of the city [31]. These physico-geographical features with extended flat areas enabled in the communist period the development of a large and dense residential area dominated by collective buildings and artificial surfaces, surfaces known by their high absorption/emission capacity. Due to these general geographical aspects, we have chosen to analyze a region including a larger area around the city, bordered by the hills on the west and east of the urban area.

According to data extracted from the ECAD database [32], the monthly mean air temperature (Figure 2) reaches its peak in July (21.8 °C in the interval 2001–2020, 20.4 °C in the interval 1961–2005), while the minimum monthly mean air temperature occurs in January (−2.3 °C in the interval 2001–2020, −3.2 °C in the interval 1961–2005). Between the two time intervals, temperature increases in monthly mean values are observed, probably as an effect of increasing temperature determined by the current climate change [33], but an impact of the change in air temperature due to the expansion of artificial surfaces should not be excluded.

2.2. Data

2.2.1. Land Surface Temperature (LST)

As mentioned previously, in order to analyze the extent, geometry, and intensity of the SUHI of the city of Bacău, we used the dataset obtained by MODIS sensors located on the Terra satellite platform (EOS AM-1) [35,36,37,38], a satellite that aims to provide satellite information for a series of global studies on atmospheric, terrestrial, and oceanic processes [16]. This satellite, launched on 18 December 1999, is functioning since 2001, and is extensively used in urban climate studies, including those conceived for several cities in Romania [16,23].

Among the products offered by the satellite sensors, in the present study we chose to use the LST offered by the ORNL DAAC-Collection-6 MODIS land surface temperature products platform [39,40,41,42]. The LST MOD11A1 product has a pixel spatial resolution of 1 km, and the pixels are included in a 1200 by 1200 m grid [39,43]. The LST value for each pixel is derived from the MOD11_L2 product band [43]. When the platform crosses the study region on a downward/upward trajectory, the local crossing time is around 10:30 a.m./10:00 p.m. [16,39], but this can vary with 2–3 h around this time. Thispass time is actually very similar to Landsat (10:58 a.m.), enabling the comparison between the two LST products.

A major disadvantage of MOD11A1 LST is represented by the qualitatively rather low spatial resolution of approximately 1 km (0.928 km) [38]. Instead, the main advantage of MODIS satellite imagery is that it provides a raw LST that can later be processed and be integrated into a wide range of climate studies with high temporal resolution, providing one image during the day and one during the night.

Due to the low spatial resolution of MODIS LST, we have used in our study Landsat 5, 7, and 8 LST for the same interval (2001–2020). The Landsat LST has a spatial resolution of 30 m, offering the possibility to obtain detailed thermal features at the local level. Nevertheless, its temporal resolution is weak and thus it is not able to obtain a long-term image of the SUHI.

Regardless of the delivering satellite platform, the availability of the LST is controlled by several objective factors, cloudiness being the most important. Clouds could absorb some of the long-wave infrared radiation, diminishing the quality of the LST observations [44]. Consequently, a rigorous selection of images was necessary, and the selection criterion focused on the absence of cloud cover over the study region. The MOD11A1 and Landsat LST images were subjected to a sorting process, after which 90% (MODIS) and 60% (Landsat) of the total number of valid pixels were selected.

Following this sorting process, we selected all the daily images available for the warm season (April–September), with a cloud cover of less than 10%, from April 2001 to September 2020. In this way, 946satellite images were acquired (649 images for day and 297 images for night), based on which the main thermal characteristics of the studied city were outlined. Instead, for the Landsat LST images, due to their higher spatial resolution, we have kept in our analysis all the images with less than 60% cloud cover, and the pixels missing LST data were filled through gap filling. In this way, a total of 483 Landsat LST images were included in the analysis.

2.2.2. Land Cover Data

The relationship between the SUHI and the urban land use categories was investigated using the Copernicus Global Land Service data portal and, more precisely, the standardized dataset CORINE Land Cover (CLC) from 2018 (Figure 3), which is provided by the Copernicus Land Monitoring Service [45] that is part of the European Union’s International Earth Observation System (JRC) [46]. Using CLC 2000, we have identified the changes in urban land use that occurred in this interval. Generally, all the cities in Romania, Bacău included, experienced urban sprawl and a wide conversion of large industrial units in commercial areas in the last decade [47]. These regions have been analyzed after that separately regarding the change occurred in their LST between 2001 and 2020.

2.3. Methodology

2.3.1. Extraction and Treatment of LST Products

Using the R package, we automated the extraction of sorted MODIS satellite images from the database that contained metadata related to the pixel quality assurance indicators of the LST MOD11A1 product. However, some of the pixels did not have valid data, which necessitated gap filling. To fill in these data gaps in the selected images, we used the gap-filling function r.fillnulls from QGIS. The method for filling the gaps was r.resamp.bspline bilinear interpolation.

The computation of Landsat 5, 7, and 8 LST data involves the utilization of a code repository implemented within the Google Earth Engine (GEE) platform [48]. The computation of the LST relies on the utilization of the Statistical Mono-Window (SMW) algorithm [49]. In addition to Landsat data, the LST production code integrates two supplementary datasets available within the GEE platform. Firstly, atmospheric data obtained from the NCEP/NCAR re-analysis are incorporated [50]. Secondly, surface emissivity data from the Advanced Spaceborne Thermal Emission and Reflection Radiometer Global Emissivity Database (ASTER GED) from NASA are utilized [51]. Additionally, in our study, in order to address the LST images characterized by less than 60% cloud cover, a data gap-filling procedure was performed using the Data Interpolating Empirical Orthogonal Functions (DINEOF) method. This method, implemented within the sinkr package in R as in Cheval et al., 2020 [52], helped us to interpolate missing data and reconstruct the complete LST images. By employing DINEOF, the gaps in the LST data resulting from limited cloud cover were effectively filled, enhancing the overall data completeness and reliability.

2.3.2. Resampling of LST Products at Various Spatial Resolutions

To enable the analysis of the LST at the level of land use categories, we enhanced the spatial resolution of the satellite images from 1 km to 500 m by resampling the daily satellite images. Furtherm we improved the spatial resolution of the satellite imagery by regridding them from 500 m to 100 m using the R package [53], which provided more detailed spatial information. For the resample, the function resamplefrom the raster R package was used [54]. Bilinear interpolation is particularly suitable when the source and destination grids are rectilinear and is appropriate for many applications where variables only vary slightly [55]. After processing the satellite images, the monthly and seasonal multi-year averages from 2001 to 2020 were calculated.

To ascertain the correlation between Landsat and MODIS LST products, a spatial resolution upscaling from the 30 m spatial resolution of Landsat was employed, specifically utilizing the bilinear interpolation method (implemented in ArcGis Pro). This methodology enabled the alignment of the Landsat LST to the same resampled spatial resolution as the MODIS LST (100 m, 500 m, and 1 km). By harmonizing the spatial resolutions, a direct and meaningful comparison between the two datasets was facilitated, allowing for the evaluation of statistical correlations through straightforward regression analysis.

2.3.3. Elaboration of the Grid Point MODIS LST Database

To assess the SUHI spatial distribution, a grid of points was created, where each point corresponded to the center of each pixel (Figure 4a). Afterward, the temperature values were extracted from the images for each grid point. For the analysis of the extent and geometry of the SUHI, we designed two cross-sections following the N-S and E-W directions, similar to Cheval et al., (2009a) for Bucharest City [23] or Crețu et al., (2020) for Galați [16]. The purpose of these transects was to assess the distribution of the LST values in the urban environment and its surrounding areas, as well as the identification of change points of the average level obtained by applying the test proposed by Rodionov (2004) [56].

To highlight the boundaries, geometry, and spatial extent of the SUHI, the points of change along the two cross-sections were identified by applying the Rodionov test, also called Regime Shift Detection software [56]. The Rodionov test, with a p-value> 0.1, assumes that each data point is independent of the other measurements, but at the same time, it is based on the sequential application of Student’s t-test that detects significant changes in the mean regime within a dataset [16,24,56]. Thus, the isotherm that highlights the average LST calculated for the central points of each pixel belonging to the urban area to which the standard deviation of the LST is added was considered as the limit of the SUHI [16,23,56]. Thus, following the application of the Rodionov test, it is observed that by summing the temperature average of the LST with a standard deviation of the LST, we obtain the isotherm that defines the SUHI limit [16,23].

To be able to quantify the intensity of the urban heat island, we designed a buffer zone using the multiple ring buffers tool with an extension of up to 6 km and a distance of 0.5 km between the buffer zones, starting from the center of the SUHI toward the surrounding non-urban areas [16,23].

To assess the LST differences between urban, rural agricultural, forest, and aquatic surfaces, based on the grid of points, we extracted the type of land use and the temperature value corresponding to the center of each pixel. Based on the extracted values, we determined the mean LST of all corresponding points for each major usage category within the CLC 2018 dataset.

2.3.4. Detrending of LST Data for 2001–2020

Aiming to assess the impact of land use change on the LST during the study period, we have used the detrending analysis (DA), allowing the identification of cyclical or other patterns [57]. To perform this, first, we fit a linear regression model to the time series data, assuming that the data follow a linear relationship between the time steps and the corresponding values. Afterward, to detrend the data, the predicted values (representing the trend) from the actual values at each time step weresubstracted. This operation removes the linear component or trend from the data, leaving behind the residual component. Additionally, it is worth noting that the effectiveness of linear regression for detrending depends on the assumption that the trend is linear, which is applicable in our case.

Climate studies also use this method to remove trend effects from the time series. In this way, we ensured that the changes in the LST between 2001 and 2020 are not determined by the general increase in air temperature due to the recent climate change, but only due to the changes in land use.

3. Results and Discussion

3.1. General Characteristics of the Warm Season SUHI in Bacău City

For the warm season, during the daytime, the SUHI is well-contoured over the densest area of the city and reaches its maximum of 32 °C. In this area, the mean LST decreases homogeneously to 30 °C at the boundary of the urban built-up area. Thus, the intensity of the SUHI could be assessed at nearly 2 °C (Figure 5). During this interval, the impact of the largest artificial lake on the Bistrița River is very clear, generating apparently a cold pool that overpasses the limits of the water body and interacts directly with the SUHI. The coldest area of this cold pool is outlined by the 28 °C isotherm, limiting the extension of the SUHI toward the northeast, so that it remains elongated from the northwest to the southern part of the city. However, when we compare this cold pool with the humid area from the southeastern part of the urban area, which reaches 26 °C, it shows that, most likely, the water body cold pool is reciprocally mitigated by the SUHI, leading to its higher LST values. This result is, most likely, an output of the low spatial resolution of MODIS, but further studies could investigate not only whether the SUHI is moderated by water bodies located in cities, but also if the SUHI can buffer the cooling power of water bodies. On the eastern part of the city, the LST mean values decrease in direct relation to the altitude, a spatial feature that outlines the daytime SUHI of Bacău cityity during the warm season. The second smaller lake, Bacău I (Figure 1b), instead, has a smaller influence on the LST values due to its smaller surface area, with daytime mean values close to 30 °C.

During the night, there is no effect of warming from the water bodies, and the core of the SUHI is well-outlined by the 15.5 °C isotherm, while the limits of the urban built-up area are contoured by the 14.0 °C isotherm. During this time, radiative loss during the thermal inversion effect, which is common for the region [8], is easily indicated by the increased LST mean values with altitude.

Therefore, we can conclude that the SUHI is well-defined during the warm season in Bacău City both during the day and night. It is very stable and has its warmest core developed over the same densely built-up area in both intervals. The mean day–night LST value reaches 24 °C, while the limit of the SUHI reaches 22 °C. Thus, by comparing these values with those indicated in Figure 2 Table the LST values are at least 5 °C higher than the air temperature at the screen level, which is close to other results in the region [47].

These general details of the warm season are largely more complex at the monthly level, both for the day (Figure 6) and night (Figure 7). The cooling effect of the lake during the day reaches (5 °C), its maximum during July, and is present during all months (Figure 6). Additionally, during the daytime in April and May, the radiative heating of barely arable land leads to higher LST values in these areas compared to the core of the SUHI; thus, the SUHI is very weak and even absent in May. We understand then that the SUHI intensity assessment worldwide should carefully take into account the type of land use in the surrounding area of the city.

During the night, the SUHI is present in all warm season months with a constant intensity of approximately 2 °C, with LST values reaching their maximum in June (23 °C), 2 °C higher than July–August (Figure 7). This is most probably caused by the longest sunshine duration at the time of the summer solstice. The satellite images for the night are captured just 1 h after sunset when the heat storage at the level of urban structures is very high. The thermal inversion effect is perceivable at night during all months, while the warming effect of the water bodies is very weak in all months.

3.2. Comparison between the SUHI Derived from MODIS Terra and Landsat

Through their pass time over the region nearly before noon, MOD11A1 and Landsat LST products offer a comparable image of the LST distribution. This enables us to produce a comparison between these products. This was made based on spatial interpolation of the resampled products at similar spatial scales (1000 m, 500 m, and 100 m). The results indicate that MOD11A1, despite its low spatial resolution, gives an accurate image of the SUHI (Figure 8a), even if the geometry of the SUHI depends strictly on the LST product that is used. The agreement between the original MOD11A1 LST resolution and the resampled Landsat LST at the same scale is very good (R² = 0.88). The correlation between the resampled MOD11A1 and Landsat LST is still high even at 100 m (R2 = 0.62—Figure 8i), which proves the complementarity of these products in building an image of an SUHI that can benefit from the high temporal resolution of MOD11A1 and high spatial resolution of Landsat.

3.3. Limits and Geometry of Bacău City’s SUHI

In defining the boundaries, geometry, and spatial extent of the SUHI, the most significant role is played by the combination of the type of land use and the overall geographical characteristics of the city. After applying the Rodionov test, we obtained a limit of the SUHI by summing the mean LST with the one standard deviation of the LST for the daytime and two deviation standards for the nighttime due to the higher standard deviation during the daytime as applied in other studies [16,23]. The results have been mapped for the entire period of analysis (2001–2020), but also for 2001–2010 and 2001–2020 in order to identify changes in the SUHI geometry during this time.

The geometry and spatial extension of the SUHI are in direct relation with the distribution of the artificial surfaces and their location in relation to the major water body in the area of the city (Lake Şerbănești), which has the capacity to diminish the influence of the SUHI, limiting its expansion toward the northern part of the urban area. During the day (Figure 9a), the limit of the SUHI appears to be regular and extends to the 31–32 °C isotherms, overlapping the neighborhoods and extending to the contact between the lower and middle terraces of the Bistrița River. Additionally, the SUHI did not record large shifts during this time in the last 20 years.

During the night, the limit of the MODIS SUHI tends to be more uniform, extending to the 14.5–15.5 °C isotherms (Figure 7b) and showing a slightly different extension than during the day. During the night, Lake Şerbănești reduces its influence on the thermal regime due to its thermal inertial effect and the more stable stratification of air masses. Likewise, the extension over the peri-urban areas located on the east and west of the city is greatly reduced where individual dwellings predominate, and consequently, the density of built space is lower. Additionally, the agricultural areas have an appreciable extension, a fact that favors the nocturnal radiative loss and the reduction in LST values. For the nighttime, it is easy to remark that the SUHI limits extended significantly during the last 20 years, indicating that the nighttime SUHI is more sensitive to climate change than the daytime SUHI.

As indicated in the methodology, with the help of the Rodionov test, we identified the change points of the mean level along the two cross-sections (Figure 3b,c), and we were able to highlight the limits of the geometry and the spatial extension of the SUHI [16,23,56]. At the daytime level, it can be seen on the average LST cross-sections constructed in the N-S and W-E that the SUHI is delimited by the isotherms at 31–32 °C, while the LST differences between the urban area and the surrounding non-urban areas reach from 1.5 to 2 °C. The moderating effect of Şerbăneşti Lake causes a sudden drop in LST values to 28.5 °C on the N-S cross-section for points ranging from 110 to 150.

3.4. SUHI Intensity

To obtain a more accurate assessment of SUHI intensity, we constructed a sequence of buffer zones extending up to 6 km from the city center, with each buffer zone covering a distance of 0.5 km from the previous one. This method indicates the differences in the mean LST as we move away from the center of the SUHI toward the city’s outskirts and represents a tool for directly assessing the SUHI’s intensity (Figure 10a,b).

During the daytime, the LST values decrease progressively from the center of the SUHI to the distance of 3.5 km in all months of the warm season as a direct consequence of the manifestation of the SUHI. It should be noticed that there is a steep decrease in the LST between the distances of 3.5 and 4.5 km (Figure 10a) as a result of the influence of the colder water surfaces of Lake Şerbănești, especially in the morning at the MODIS Terra pass time, at time at which the lake has the capacity to reduce the LST values from the daytime in its proximity to 31 °C in July, the month with the highest mean LST values. For the first buffer zone (0–0.5 km), the mean diurnal LST value is about 35.5 °C in July, while for the buffer zones between 3.5 and 4.5 km, the daily average values are approximately 31 °C in July (Figure 10a), resulting an SUHI intensity of 4.5 °C. In April, the distribution of the SUHI along the same buffer zones does not outline the manifestation of the SUHI, especially due to the higher LST over the arable land.

During the nighttime (Figure 10b) the SUHI’s intensity declines for the entire season to 2 °C, the difference between the first and last buffer zones, with the same intensity for all the months. The effect of the water bodies disappears, and the LST distribution over the region is controlled only by the SUHI intensity.

Our results indicate an SUHI intensity higher than that assessed for Bacău in a country-scale study [58], in which the SUHI intensity was estimated at 1.2 °C for the daytime and 0.8 °C for the nighttime. These differences are explained mainly by the different methodological approaches in assessing the SUHI.

3.5. Influence of Land Use on the SUHI

In similar studies, thermal differences between different land use types are frequently used to describe the intensity of the SUHI [16], unmasked by the impact of land use on the LST. Generally, LST differences between urban and rural areas [34,59], aquatic [1], or agricultural surfaces [21] are compared and, therefore, we performed a comparative analysis of the LST distribution during the warm season taking to account 16 CLC types of land use (Figure 11).

During the daytime, the urban surfaces show the highest LST ranging from 29.6 °C in green urban areas to 31.9 °C in areas occupied by sports and leisure facilities (Figure 11). The mean daytime LST of the artificial surfaces during the warm season is 30.8 °C. Artificial surfaces have a high absorption capacity of energy fluxes between the active surface and the lower troposphere due to their specific heat and low permeability, different thermal conductivity, and low albedo values. The discontinuously built urban space presents the largest expansion among all artificial surfaces, its thermal role being a very important one in defining the SUHI intensity with a mean LST value of 29.8 °C, while the commercial and industrial units have an appreciable expansion, with a mean LST of 30.8 °C.

During the night, artificial surfaces show rather small LST differences between distinct types of land use. The mean nocturnal LST in the warm season of artificial surfaces is 14.8 °C. The LST varies from 13.5 °C in a small stone quarry area in the region to 16 °C in the areas occupied by sports and leisure facilities (Figure 11). Artificial surfaces exhibit high radiative losses, but these are lower than in the case of agricultural surfaces due to the specific heat and low permeability.

Analyzing the mean LST for major land use units (urban, rural, aquatic, and forest) for the daytime (Figure 12a), one can easily observe the major difference between built areas (urban and rural) and both water bodies and forests. Partially, the low LST of forests is explained by their microclimatic features, with the LST taken at the canopy level, but also by their location on higher grounds with low air temperature. Even so, these results underline the capacity of forested areas to reduce the mean LST by 3–5 °C in the proximity of urban areas from the temperate zone and confirm the role of green areas in Bacău City to reduce the LST [60]. The water bodies have the same impact, with mean monthly LST 1–2 °C lower than the urban areas.

During the nighttime (Figure 12b), the monthly mean LST underlines in its evolution that June is a little bit cooler on artificial surfaces of the urban area than the other types of surfaces. On one hand, this is an effect of the fact that June records the maximum amount of precipitation, and thus, the high water content of vegetation and soil in non-urban areas buffers the radiative loss sustaining higher values of LST. The urban surfaces, on the other hand, have a higher radiative loss and are colder in the first part of the night at the MODIS pass time. This situation fades in July and August when the lack of precipitation imposes the capacity of urban surfaces to record the highest values of the LST, and the largest difference is observed between urban and non-urban areas.

3.6. Land Use Changes and Their Impact on the LST

To quantify the impact of land use changes on the LST distribution, we used the CLC Change data to identify surfaces changing land use from 2000 to 2018. For the CLC 2000 land use type, we assigned the mean LST of the warm season from 2001 to 2010 as being representative, while the CLC 2018 land use type was assigned to 2011– 2020 interval. The land use change that took place between 2000 and 2018 consisted generally on the transformation of some agricultural areas (complex crops, non-irrigated arable land, and pastures) into urban commercial areas (discontinuously built urban space and commercial and industrial units) to reduce real estate pressure from the city level and also offer commercial units more accessibility for the city dwellers.

In this way, around 260 ha (Figure 13a) of recorded changes in land use categories were directed, mainly from agricultural areas to artificial ones, a shift that presumably had a direct impact on the LST. The transformation of agricultural land (complex crops, non-irrigated arable land) into built-up urban areas is the most represented, summing up over 120 ha (Figure 13a).

Due to the recent increase in air temperature induced by the ongoing climate change, it is hard to disentangle this effect from that of land use change when we analyze changes in the LST over a region. Therefore, the original LST values were detrended. At first glance, a general increase in the LST values on surfaces that have undergone land use change (Figure 13b) can be observed during the day, but in the meantime, the LST values decrease during the night. The results obtained following the detrending application show a difference in the mean LST of +0.1 to 0.2 °C (Figure 11b) in the case of agricultural areas transformed into urban areas between 2000 and 2018 for the day, while during the night, the mean LST is lower with −0.3 °C. We can observe that during the day, a drop in the LST of −0.2 °C is recorded on the surfaces that have been transformed from water surfaces to wetlands, but the same area is warmer by +0.4 °C during the night. These differences were tested for statistical significance using a t-test (p < 0.01), and the results indicated that they are statistically significant during the nighttime and not statistically significant during the daytime, except for the shift from water bodies to inland marshes.

The results show that changes from natural to artificial land use tend to increase the LST during the day, but also decrease the LST during the night, which induces a higher range in diurnal LST variation. This indicates that the impact of urbanization and the shift from natural to artificial surfaces in the LST is not always immediate, and the thermal imprint of these kinds of surfaces on the LST needs time to be visible in the long time series. Nevertheless, it is reasonable to consider that these results draw the major lines on how the LST changes under the effect of changing land use, and in further studies, a detailed image could be obtained using LST data with higher spatial resolution.

4. Conclusions

During the daytime of the warm season, the spatial extension of the SUHI is appreciable, delimited by the 31 °C isotherm. Therefore, the densely built-up area of the city is 1.5–2 °C warmer than the surrounding non-urban areas. The SUHI extends in the central and southern areas of the city, but a slight extension of the SUHI toward the peri-urban areas in the east is observed, precisely in that region where real estate pressure has led to an expansion of urban space. Generally, the spatial extension of the SUHI is in direct relation with the artificial surfaces and its location in relation to water bodies.

During the nighttime, the spatial extent of the SUHI is more compact and uniform around the city’s central area, limited by the 15.5 °C isotherm. At this time, the densely built-up area of the city is 1.0–1.5 °C warmer than the surrounding non-urban areas. At night, the artificial surfaces often have a lower mean LST due to their more radiative loss.

During both day and night, the SUHI gradually narrows toward the northern part of the city, where the presence of Şerbănești Lake diminishes the influence of the SUHI and determines the moderation of the LST values by 28 °C.

The rural and cropland areas in the east of the city have generally higher LST values than the areas west of the city. In the rural areas extending over the terraces on the right side of Bistrița River, a lower LST can occur in April and the first part of May due to cold air advection resulting from northern and southern circulation. The LST values progressively decrease from the SUHI center to a distance of 3.5 km due to the SUHI effect. Between 3.5 and 4.5 km from the city center, the effect of the cooler water surfaces of Lake Şerbăneşti can be felt.

The LST impact of land use changes that occurred between 2000 and 2018 indicates that built-up surfaces replacing agricultural ones drive a higher range in the daily LST values with a higher/lower LST during the daytime/nighttime and a prevalence of night cooling.

Our study underlines that the assessment of the SUHI/UHI at local scale represents the best approach of this topic so that this kind of assessment could be useful for local policymakers in their attempt to develop strategies for mitigating the impact of climate change and driving the urbanization process toward the most sustainable path. In this idea, our main interest in further studies will be focused on more detailed studies by using high-resolution LST images or assessing local conditions inside the city with urban climate modeling techniques.