Change detection in a rural landscape: A case study of processes and main driving factors along with its response to thermal environment in Farim, Iran

This study aims to investigate the alteration of Land Use/Land Cover (LULC) change and its response to changes in land surface temperature (LST) and heat island phenomena of a rural district known as Farim in the north of Iran from 1990 to 2020 using multi-date Landsat data. The random forest-based algorithm, supported by Google Earth Engine, is used to execute the LULC classification with an overall accuracy of more than 92%. Based on the LULC results, in terms of area changes, the classes of bare land, rice fields, and water bodies encountered an increase, but woods and dry farms decreased. The present study also incorporates the trends of land cover change that are analyzed using regression based on the temporal datasets of the three leading driving factors: temperature, precipitation, and population. The result demonstrates that the main changing factors of the mostly changed class (bare land) are population/precipitation and temperature/population. Additionally, the effect of LULC change on seasonal LST and urban heat island (UHI) is also analyzed in this study. The result witnessed a significant LST rise in the summer and winter seasons of about 12.87 °C and 14.2 °C, respectively over the study period. The Urban Thermal Field Variance Index (UTFVI), characterizing the heat island phenomenon, shows that the strongest UTFVI zone is in the central area and the none UTFVI zone is in the surrounding region. Moreover, both seasons have seen a significant rise in none UTFVI zones compared to decreasing strongest UTFVI zone. The result of the present study will be helpful for urban planners and climate researchers who study future land cover change and its associated driving factors.


Introduction
As the provenance of civilization, landscapes are marked by the signs and impacts of human beliefs and interactions with nature.Conforming to the contemporary demands for a deeper and more accurate analysis of the environment, researchers are constantly exploring the landscape features and traces that relate to the stages of its formation (Mofrad and Razavi 2021).Concerning this, and as icons of the most basic form of cultural heritage, some researchers suggest that rural landscapes provide a rich context for landscape change studies (Antrop 2005;Gholipour Shayan and Razavi 2021).The effects of social, economic, political, legal-managerial, and agro-technical changes of urbanization on rural areas (Movahedi, et al. 2021), besides the biophysical and ecological changes, are essential issues that can play important roles in land use/cover changes and deforestations (Seto et al. 2013).Based on Winoto and Schultink (1996), the urbanization process can lead rural areas to change from agricultural lifestyles and activities to non-agricultural ones and increase rural-urban migrations.Consequently, rural-urban migration or rural abandonment can cause soil erosion and different effects on land use/cover (LULC) patterns (Min 2022;Alibaygi and Karamidehkordi 2009;Gabriel 2002) and the rural ecosystem, which can cause biological misbalance (Pineda 2000).As Stockdale (2004) demonstrates, these migrations are more traceable in developing countries.

Study area
The study area is Farim, a rural district in the Dodangeh region in Mazandaran, Iran.It lies in the northern slopes of the Eastern ranges of the Alborz Mountains, which perform as a barrier between Semnan Province and Mazandaran, retaining the moisture from the Caspian Sea and affecting the distribution of precipitation, temperature, and moisture.
According to geographical coordinates, Farim lies between 36°6′-36°11′ latitudes and 53°6′-53°24′ longitudes in W-E and 36°14′-36°5′ latitudes and 53°18′-53°6′ longitudes in N-S direction (Fig. 1).The Do-Berar summit of Alborz in the south of Dodangeh, with a height of 3000 m, impacts the mountain-plain characteristic of Farim with a range of 500-1900 m from the sea level.The rural, agricultural morphology and the specific flora and fauna of this district follow the area's high altitude.
Farim is categorized as "moderately rainy (no dry season) with hot summers" or "Cfa" based on the twentieth-century investigation of the Köppen-Geiger climate classification of Iran (Raziei 2016).The reported data from the Farim climatology station (Center for Statistics of Islamic Republic of Iran 2013) shows that annual rainfall in Farim surpasses 500 ml, compared with the Caspian littoral annual rainfall average of 100-150 ml (Gale 2007) and an average of 30% rainy days throughout the year.Meanwhile, Farim is presumed to have rich groundwater resources.

Data used
The datal used in this study came from different sources with various configurations.The United States Geological Survey (USGS) Landsat 5 and 8 images for producing land cover mapping each year were obtained in Google Earth Engine (GEE).Images of "USGS Landsat 8 Surface Reflectance Tier 1" were used for the year 2020 analysis that provides corrected surface reflectance atmospherically with a spatial resolution of 30 m and a scale of 0.0001, and images of "USGS Landsat 5 Surface Reflectance Tier 1" was used for the years 1990, 2000, and 2010 analysis.The meteorological information of synoptic stations of Farim from 1985 to 2020 was produced by the Iran Meteorological Organization.The Digital GIS Data and Maps were produced by the Iran National Cartographic Center in 2011.Moreover, the rural guide plans, including land use/cover plans and other physical information about Farim produced by the Islamic Revolution Housing Foundation, and the demography statistics of Farim gathered by the Statistical Centre of Iran indicate the effect of population on this landscape.

Methods
In this study, multi-temporal Landsat data, as mentioned in Table 1, is used to investigate the land use land cover variation, its decadal change, and alternation of one to another land cover classes, seasonal land surface temperature, and urban heat island variation over the Farim land.The overall methodology adopted in this study is shown in Fig. 2.

Image pre-processing
The pre-processing of Landsat images requires some radiometric and atmospheric correction.We have used the geometrically corrected image by the data provider.Radiometric correction is an essential step of satellite image pre-processing that accounts for sensor characteristics and variance in illumination conditions, most importantly sun angle/topography and sensor irregularities distortions.Some portions of the images contained atmospheric noise.It can occur by malfunctioning satellite sensors.Thus, the atmospheric correction was done using the absolute correction method.
At GEE, the raster data are represented as "image" objects that can be constructed by one or more "bands" with their propositions, such as name, data type, and scale.For each year of the process (1990,2000,2010,2020), defining the project zone and the date zone of "Farim" took place.After that, to obtain data with the lowest/without cloud masks, the "cloud_cover" filter was applied to each year's analysis through the following steps.First, the variety of image collections of USGS-Landsat was opened for each year in 1990, 2000, 2010, and 2020.Then all data were specified with some "filters" in the study area or "Farim," "filter Date(April-July)" that is selected in the range of the growing seasons, and the "CLOUD_COVER" of less than 1%.The defined specific "bands" represent the result images as essential based on the availability of bands at each Landsat dataset.With a kernel-defined neighborhood, the computation is done by "ee.Kernel.square"and "entropy" bands.The bands for spectral indices were as follows: NDWI (Normalized Difference Water Index), NDVI (Normalized Difference Vegetation Index), and NDBI (Normalized Difference Built-up Index).The DEM or digital elevation model by "ee.Terrain.Slope" band was calculated in the other phase, and the pixel value statistics in the region were conducted as well.

LULC image classification
The five main land covers over the study area are 0, woods; 1, rice fields; 2, dry farms; 3, water bodies; and 4, bare land.Many machine learning classification algorithms have been developed to obtain more accurate land cover maps.GEE, as a machine, can use these algorithms to predict remotely sensed sample data that make the classification easier, some of which include classification and regression trees (CART) (Breiman et al. 1984), decision tree constructed by Segal (1992), and random forests (R.F.) constructed by Breiman (2001).Accordingly, the confusion matrix has been computed to assess this research's "overall, producer, and consumer accuracy" percentage.

Simultaneous multiple linear regression
After computing the areas of each studied category, the simultaneous multiple linear regression (enter) method was used.Multiple simultaneous regression analysis was used to evaluate the effectiveness of the main factors, and the contribution of these factors was examined by double regression analysis.We used the annual area of different land use categories as a dependent variable and performed regression analysis simultaneously with temperature, precipitation, and population statistic information as independent variables.The multiple regression model is shown in Eq. ( 1): where Y is the dependent variable, X 1 , X 2 , and X 3 are the independent variables, and a is the intercept of the regression model.
The coefficient of determination was chosen to assess the estimation accuracy as shown in Eq. (2) (Van Straten et al. 2010).
The d and p are the actual and predicted values of the output parameters, and d and p are average values. (1)

Estimation of LST
LSTs are theaters, an essential character in understanding the thermal atmosphere and earth's surface of any urbanized area.In this study, the Landsat 5 TM band 6 (1990, 2000, and 2010) and Landsat 8 TIRS band 10 (2020) satellite datasets were used to generate the LST over the study region.
LST from Landsat 5 Conversion of the digital numbers (DN) of a thermal band of Landsat 5 TM sensor into radiance luminance ( R TM6 ) using Eq. ( 3) is the initial step for the LST calculation (Sobrino et al. 2004).
The second step is to convert the radiance luminance into LST in Kelvin using Eq. ( 4).
The final LST was obtained in degrees Celsius using Eq. ( 5).

LST from Landsat 8
The following steps are used for the preparation of LST maps from Landsat 8 thermal bands.
Step 1: Conversion of DNs of ground-based substances to spectral radiance ( L ) using Eq. ( 6).
where L is the top-of-atmosphere (TOA) spectral radi- ance in W/(m 2 .sr.µm), L max is the maximum, and L min is the minimum spectral radiance in W/(m 2 .sr.µm), Qcal max is the maximum, and Qcal min is the minimum quantized calibrated pixel value.
Step 3: Calculation of NDVI (Normalized Difference Vegetation Index) for the brightness temperature correction for surface emissivity.It is estimated using Eq. ( 8) (Estoque et al. 2017).
where Red and NIR are the surface reflectance of red and NIR band.
Step 4: Proportion of vegetation ( P V ) calculation using the maximum and minimum value of NDVI images by applying Eq. ( 9).
Step 5: The land surface emissivity ( ) is calculated using Eq. ( 10) from the proportion of vegetation following the method of Sobrino et al. (2004).
Step 6: Final LST map is prepared by applying the correction of land surface emissivity to brightness temperature using Eq. ( 11) (Estoque and Murayama 2017;Li et al. 2011).

Estimation UTFVI
In this study, we have used the Urban Thermal Field Variance Index (UTFVI) to characterized the UHI phenomenon over the study region.UTFVI maps are investigated using LST data, and the following equation is used in Eq. ( 12) (José Antonio Sobrino and Irakulis 2020; Liu and Zhang 2011).
where LST pixel indicates the LST values of the pixels and LST mean indicates the mean LST of the study region.The UTFVI images obtained can be classified into six levels according to the six distinct ecological zones: none, weak, middle, strong, stronger, and strongest, as shown in Table 2.

Accuracy assessment of classified LULC maps
This study has chosen the random forest (R.F.) algorithm as a classifier.The R.F. consists of many decision trees that identify each entered sample belonging to which class.
Each decision tree has a vote on the class of a sample, and the result class will be the class with the most votes.Varieties of Landsat bands such as PC2, PC1, slope, NDWI, NDVI, NDBI, and texture were merged as input data into the R.F.After the classification, calculating the accuracy of the results was done using the producer's and consumer's accuracy and Cohen's Kappa coefficient statistics methods.
The results have been expressed in Table 3.
The overall accuracy of each year was 95% for 1990, 97% for 2000, 92% for 2010, and 96% for 2020.The Kappa coefficient accuracy statistic is 94%, 97%, 90%, and 95% for each year, respectively, from 1990 to 2020.Based on the statistics, all the accuracies were above 77%, which is an acceptable amount, and woods and water bodies had the most producer's accuracy and consumer's accuracy during the whole period.Dry farms and bare classes have lower accuracy percentages, and the rice fields class is average among all classes.

LULC variation
In this research, change detection of Farim has been accomplished over 30 years in 3-decade intervals: 1990, 2000, 2010, and 2020, respectively.The complete classification and accuracy assessment process has been done on the GEE platform.The classification output using the random forest (R.F.) algorithm was "land cover classification maps" of each year presented in Fig. 3. Based on the resolution of the satellite imageries, each map was classified into five main classes: woods, rice fields, dry farms, water bodies, and bare lands.It should be considered that the bare land category consists of the built area and the barren lands that, due to the resolution, we considered putting in one category.Moreover, with the help of the land cover classification maps, each class's area was calculated.All areas in km 2 and percentages are outlined in Fig. 4. Figure 4 declares that in the area over Farim, the woods class dedicates the most part, and the water bodies class has the lowest amount.The central and western part of Farim primarily consists of dry farms and bare classes due to the distances from water resources and its topography.Rice fields are mainly located around the river, from the northeast to the south and southeast.
The water bodies and bare categories experienced growth during the whole period, while there was a decline in the woods and fluctuations in dry farm and rice field categories.The woods LULC class covered the largest area of the whole study area, between 268.66 and 288.78 km 2 .The rice fields class reduces its area from 1990 (21.28 km 2 ) to 2010 (17.16 km 2 ), but it increases in the last decade of the study to 21.45 km 2 .The dry farms class has an accretion rate of area change from 45.98 km 2 (11.88%) in 1990 to 58.32 km 2 (15.07%) in 2000, while after that, it experienced a reduction until 2020, to 40.23 km 2 (10.39%).The changing pattern of both water bodies and bare classes is ascending.Among all classes, water bodies with the lowest area percentage changed from 0.18 km 2 (0.04%) in 1990 to 1.06 km 2 (0.27%) in 2020.Moreover, the bare class with the third vast area among all categories in 1990 (30.70 km 2 ) increased to 55.53 km 2 in 2020.

Gains and losses of each LULC category
The shift rates of each LULC category can give an overview of each class's gains and loss trends.It should be considered that  although the overall trend of classes is ascending or descending, each class has its losses and gains within its intervals.
Over 3 decades in this study, the main changes in the area of classes were dedicated to the bare class with 24.83 km 2 gain (+ 80.87%), respectively, woods with 20.12 km 2 loss (− 6.96%), dry farms with 5.75 km 2 loss (− 12.5%), rice fields with 0.17 km 2 gain (0.79%), and water bodies with 0.88km 2 gain (488.88%) as shown in Table 4. Therefore, the only loss in the area happened to the woods and dry farms classes in this period.

Decadal LULC variation
Table 5 indicates every 10 years' conversion amount of each LULC class.Moreover, Fig. 5 shows the locations of these LULC changes in Farim.
Between 1990 and 2000, the main changes happened to the woods class into the dry farms, which was 11.92 km 2 , and the lowest change was the conversion of dry farms to water bodies, which was 0. The rice fields class mainly changed to dry farms, 2.15 km 2 , with less conversion to water bodies at 0.21 km 2 .Most parts in the dry farms category changed to the bare class of 9.27 km 2 .During this period, 0.02 km 2 of the water bodies remained unchanged, while most parts changed to the rice fields with 0.06 km 2 .There was 9.28 km 2 conversion into the dry farms in the bare class, which is the most changed area of this class.
From 2000 to 2010, most of the change transformed the dry farms into the bare class of 12.36 km 2 .Moreover, the essential point is that none of the water bodies' areas changed into other classes.The woods mainly changed into the bare and dry farms' classes with 4.93 and 4.62 km 2 of change.The most changing part of the rice fields was its conversion into dry farms, 2.13 km 2 , and that of the dry farms was the bare 12.36 km 2 .
From 2010 to 2020, considerable change was dedicated to converting dry farms into bare class by 4.18 km 2 .No amount of the woods, dry farms, and bare classes changed into water bodies, and 0.07 km 2 of the rice field changed.Most of the transformation of the water bodies was into the bare, and that of the bare category was into the dry farms classes.
Overall, between 1990 and 2020, the most considerable area change was dedicated to the woods into bare class with an 18.49 km 2 change.After that, the dry farms into the bare, with an 11.14 km 2 , change.The wood conversion into the rice fields, and dry farms, was 1.56 and 3.52 km 2 , and the area change in the rice fields was 0.73 km 2 for woods, 0.92 km 2 to dry farms, and 1.26 km 2 for bare classes.In the dry farms class, the conversion into the woods and rice fields was 4.32 km 2 and 0.88 km 2 .The most change in the water bodies category was to rice fields, which was 0.06 km 2 .The bare class changes primarily include the dry farms class, 7.3 km 2 .Furthermore, in the woods, rice fields, dry farms, and bare classes, the lowest conversion happened to the water bodies class with 0.2 km 2 , 0.48 km 2 , 0.03 km 2 , and 0.08 km 2 , respectively.It should be considered that most parts of each class remained unchanged during the whole period.   1 3

Woods changes
Farim is located in an essential Mazandaran region surrounded by protected areas, including Bula, Parvar, Esas, Hezar Jarib, DoDangeh Wildlife Sanctuary, and Kiasar National Park, accounting for the rich fauna and flora of the region.So, the changes in the woods class can cause an essential environmental change in its ecosystem.Based on this study, from 1990 to 2020, most of this class had changed to bare class, which means the destruction of woods to build roads and buildings or abandoning uncultivated zones.Based on Fig. 5, most of the wood conversion was to the bare class, 18.49 km 2 , and happened in the northeast and northwest of Farim and along the cultivation zones.Moreover, the conversion of woods into rice fields and dry farms is around these cultivation zones.

Rice fields changes
The rice fields of Farim are primarily located in the southeast and northeast due to the flatness of the land in terms of topography and proximity to water resources.13.69 km 2 of these fields remained unchanged during the whole period, while 1.26 km 2 changed to bare, 0.92 km 2 to dry farms, 0.73 km 2 to woods, and 0.48 km 2 to water bodies over Farim.Over time, the continued decline in this class would lead to losing a significant portion of the region's food resources.

Dry farms changes
Dry farms of Farim are primarily located on the sloping parts of the topography, in the center of the area, with extensions in the east and west.The main crops grown are plants with low water requirements, including wheat, barley, tobacco, roses, and black root trees.From 1990 to 2020, 20.48 km 2 of this category remained unchanged, but its 11.14 km 2 transformation into the bare class can be a worrying point for the future.Because the cultivated crops of this class are essential in the region's economy, their loss can cause social changes for the people.Most of the conversion of dry farms into bare class happened in the northern parts of Farim and the vicinity of the villages.

Water bodies changes
During the whole period, the water bodies class constituted a small percentage of the area; however, the location of villages and rice fields between them and in the vicinity of rivers shows the proper productivity of villagers from water resources.The flow of Mazandaran, Talar, and Tajan's two main rivers and their branches through Farim nourishes the area with irrigation and alluvial enrichment.However, the water bodies class in this study consists of the areas that gather water, dams, and reservoirs.In 1995, the Rajaei dam was constructed north of the Tajan River, which flows from northeast to southeast of Farim.Specific areas of woods and agricultural lands were either submerged or depleted due to this construction or other ecological factors.Furthermore, the construction of the Maji dam in 1998 in one of the western villages of Farim caused significant landscape degradation in the periphery.Overall, due to the limited area percentage of water bodies in Farim, its conversion to other classes is low, but conversely, the opposite trend is more remarkable.

Bare changes
In this study, due to the resolution of the satellite imageries, the combination of the built area, roads, mines, and barren lands pixels are nominated as the bare class.Based on Table 5, 13.82 km 2 of the bare class had remained unchanged in the whole period, 7.3 km 2 of it changed to dry farms, and its change to the woods, rice fields, and water bodies was 2.35, 1.02, and 0.08 km 2 , respectively.Moreover, based on Fig. 3, the changes in this class mainly occurred in the central parts of Farim and near the cultivation zones.

Affecting factors of land cover change
Based on the area in classified maps that showed the number of changes in each of the five classes each year, investigating the affecting factors on each of them using regression is an essential step in the change detection of Farim.Among the natural processes and human actions, three factors were selected due to their importance and the availability of their official data: the average yearly temperature, total precipitation, and the annual population.Table 6 shows the data used in the regression.In Table 7, R 2 (or coefficient of determination) means how much the contribution is between the constant variables and the changes in each class.The more this amount, the more contribution as a change factor of the class.All the R 2 amounts should be under 1 and, in this research, differs from 0.06 as the lowest to 0.99 as the highest.The equation in each row is the class relationship, and each constant variable can be used as a formula for future research.For example, by having two of the precipitation/temperature/population statistics, the area of that category is predictable.
In the woods class, the lowest R 2 is 0.55 in the temperature/precipitation variable categories, and the temperature/population has the highest amount of R 2 , 0.73, which means that in the woods class, the most impressive changing factors are temperature/population and then population/ precipitation.
In the rice field category, the lowest and highest R 2 are 0.06 and 0.99, respectively.In this class, temperature/ precipitation has the highest impact on changes in the rice fields; after that, temperature/population is the average rate, and population/precipitation is the lowest one.
The most impressive changing factor in the dry farms category is temperature/precipitations, and its R 2 is 0.93.Temperature/population and population/precipitations come with the R 2 of 0.59 and 0.12, respectively.
Two constant variables in water bodies and bare classes are the same in R 2 quantity.These variables are population/ precipitation and temperature/population, with the R 2 of 0.98 for the water bodies class and 0.99 for the bare; they have the same impact on the water bodies' changes.

Seasonal LST variation
The LST variability over Farim is also investigated in this study during the summer and winter seasons.The obtained LST maps over the study area are shown in Figs. 6 (summer) and 7 (winter), respectively, for the years 1990, 2000, 2010, and 2020.The maximum mean LST occurs in the summer, and the minimum LST occurs in winter.Over the study period, there is an increase in mean LST during both seasons.It increased from 20.37 °C in 1990 to 33.24 °C in 2020 during the summer season, while the winter season witnessed an increase from 3.57 °C in 1990 to 17.77 °C in 2020 (Table 8).The maximum LST in the summer season goes up to 51.37 °C (2000) and 29.16 °C (2010) in the winter season.The minimum LST in both seasons shows a value of 11.93 °C (1990) and − 7.18 °C (1990), respectively.The mean LST in the summer season increases by 12.87 °C, while in the winter season, it increases by 14.2 °C during 1990-2020 over the selected study area.The spatial distribution of LST suggests that the mean LST is higher in the central parts of Farim, which are more urbanized than the surroundings.Moreover, the mean LST decreases as the distance from the core increases.This attests to the presence of urban heat islands during both season, whereby the core area shows higher LST.The rise of urbanization is responsible for the increase in urban built-up, which increases the mean LST in the central area compared to the surrounding greenery.
This type of higher urban LST behavior is also seen in recent studies over Cumilla City of Bangladesh (Kafy et   The behaviors of UTFVI during both seasons are more or less similar.None and the strongest UTFVI zones are more dominant than the other four zones during both seasons.During summer, the strongest UTFVI zone characterizes the central urban area, while none UTFVI zone dominates the rural region (Fig. 10).None UTFVI area increased from 173.96 km 2 (55.8%) in 1990 to 207.13 km 2 (66.43%) in 2020, whereas the strongest UTFVI zones decreased from 128.1 km 2 (41.09%) in 1990 to 96.53 km 2 (30.96%) in 2020 (Table 9 and Fig. 10) during the summer season.The spatial distribution of winter UTFVI maps also suggests the urban heat island occurrence over the Farim land, with the dominant presence of the strongest UTFVI zone in the central urban area and no UTFVI in the rural area (Fig. 9).In the winter season, an increase in the strongest UTFVI area from 113.85 km 2 (37.57%) in 1990 to 145.15 km 2 (48.13%) in 2020 and a decrease in none UTFVI area from 186.81 km 2 (61.65%) in 1990 to 146.57km 2 (48.61%) in 2020 (Table 9 and Fig. 10) are seen.However, the strongest and none UTFVI zone follow a similar trend in both the summer/ winter season; the percentage distribution of none UTFVI zone remains higher in the summer season.In contrast, the percentage distribution of the strongest UTFVI zone remains higher in the winter season.
Though remote sensing analysis of LULC change has cost-effective, less time-consuming, and has broader coverage, it still has some limitations.The heterogeneous nature of the earth's surfaces sometimes makes it difficult to accurately classify.An added field survey data is essential to   province, have experienced 20.12 km 2 (− 6.96%) loss; as the results show, temperature/population factors have the most effects on this class.This scale-down can alarm the local and national people and organizations to be more aware of natural resource destruction.Dry farms reduced, with 5.75 km 2 loss (− 12.5%), and rice fields and water bodies increased with 0.17 km 2 gain (0.79%) and 0.88 km 2 gain, respectively, resulting from land covers and land use interactions, such as changing the type of plantings and constructing seals instead of rice fields and dry farms.Results stated that the main changing factors in rice fields and dry farms classes are temperature/precipitation or environmental factors, but human and environmental factors influence the water bodies class.The result of LST variation suggests an increase in mean LST over the study in both seasons over the last 3 decades.The winter season shows a higher increase (14.2 °C) than the summer season (12.87 °C).The spatial distribution of UTFVI witnessed the presence of the strongest UTFVI zone in the central area and none UTFVI zone in the rural area.There is a significant increase in none UTFVI area (33.04% in summer and 27.49% in winter) compared to a decrease in the strongest UTFVI area (25.05% in summer and 21.54% in winter) over Farim.The result of the present study would be helpful for policymakers and urban planners for the future development of Farim under the climate change scenario.

Fig. 1
Fig. 1 The officially defined area of Farim rural district, based on the Islamic Revolution Housing Foundation (2009) in Mazandaran Province of Iran

Fig. 3
Fig. 3 LULC map of the study area for the years 1990, 2000, 2010, and 2020

Fig. 4
Fig. 4 LULC variation over the study region in km 2 (a) and percentage of landscape in class (b) for the years 1990, 2000, 2010, and 2020

Table 1
List of the Landsat scene used in this study(Path/row: 163/035)

Table 2
(Liu and Zhang 2011)s and their corresponding ecological evaluation index(Liu and Zhang 2011)

Table 6
Annual reports and statistics of yearly temperature, total precipitation, and population of Farim from 1990 to 2020

Table 7
The relationship between environmental and social factors with the changes in each class.W-woods, RiF-rice fields, DF-dry farms, WB-water bodies, and B-bare

Table 8
Seasonal LST variation (in °C)over Farim