Assessing Impact of Land Use/Land Cover Dynamic on Urban Climate Change in a Semi-Arid Region – Case Study of Agadir City (Morocco)

This research sought to assess historically the urban expansion of Agadir city in Morocco within the 35-year timespan (1984–2019), and the influence of those changes on the regulating services of Agadir. It was achieved by applying support vector machine supervised (SVM) algorithm on each Landsat products to derive land use/ land cover (LULC) maps. High accuracy assessment values were obtained for all Landsat classified maps. Spectral radiance model was exploited successfully to highlight the spatiotemporal changes of thermal behavior of city surfaces. Terrestrial carbon dynamics of Agadir LULC was evaluated by applying a process-based carbon model. The outcomes of this paper revealed an important urban expansion within the 35-year timespan with an important loss of vegetation and bare land. This urban evolution impacts the land surface temperature (LST) and caused carbon storage loss that contributes to local climate change. These findings could assist policy-makers to characterize a sustainable evolution of urban area, especially, to interpret how and where LULC changes might alter the dynamics of climate regulation and ecosystem services.


INTRODUCTION
As stated in the Global Urbanization Forecasts presented by the United Nations in 2018, the worldwide demographic trends in cities will reach 68% by the end of 2050 with an accentuated growth in Asia and Africa [Chen 2024].The rapid growth of populations and their activities increased the greenhouse gas concentration in the Earth's atmosphere [Qi et al. 2022].In general, the unsustainable evolution of the human activities has a direct consequence on the ambient atmospheric conditions that affects the ecosystem services.Thus, scrutinizing the reciprocal impact of land use/land cover (LULC) transformations on micro-climate play a crucial role in accelerating transformations toward renewable development and mitigating Earth's atmosphere balance.Hence, two known parameters of the earth climate system are widely used to assess and monitor the effect of LULC changes namely, Land Surface Temperature (LST) and terrestrial carbon storage [Xiao et al. 2019].
LST is defined as the thermal behavior of a terrestrial surface and it is affected the anthropogenic activities that convert natural landscapes into impermeable surfaces [Adeyeri et al. 2024].As listed by [Patel, Indraganti, and Jawarneh 2024], a variety of methodologies have been established and employed for extracting LST from remote sensing datasets, namely Physiologically

ECOLOGICAL ENGINEERING & ENVIRONMENTAL TECHNOLOGY
Equivalent Temperature (PET, universal thermal climate index (UTCI), and Surface Urban Heat Island Intensity (SUHII).
In fact, many scholarly inquiries have explored how changes in LULC patterns affect LST.[Orhan and Yakara 2016] investigated the correlation between alterations in vegetation and LST rates.Their study emphasized that agriculture activities in Konya, Turkey, contributed to about 5°C temperature increase between 1984 and 2014.[Nguemhe Fils et al. 2018] used multi-temporal Landsat scenes to conclude a continuous increase in linear link between LST and impervious areas in Douala Metropolis, Cameroon, from 1986 to 2016.[Sheik Mujabar 2019] revealed a significant difference in recorded temperature over an industrial city in Saudi Arabia, utilizing thermal infrared remote sensor (TIRS) data.He reported that the temperature in residential area ranges between 40°C and 50°C for summer, while in suburban areas it is moderate and in industrial area goes from 50°C to 55°C during the entire study period.In addition, he highlighted that temperatures more of than 60°C are recorded in the industrial facilities producing iron and steel.
Furthermore, the Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) algorithm deemed to be the simplest and most reliable tools for estimating the quantity of stored carbon [Verma et al. 2024].The primary purpose of this model is to determine ecosystem carbon storage.It achieves this by multiplying mean density of carbon in reserves, namely living biomass (above and below ground), soil, litter and dead organic matter, for each LULC class with their corresponding areas, following the methodology proposed by [Mandle et al. 2016].
The research literature devoted to examining how urban growth affect ecosystem carbon reserves is extensive [W.Jiang et al. 2017;Verma et al. 2024].Yet, only handful made use of the In-VEST tool.The model was utilized by [Zhang et al. 2017] to monitor and assess the effects of shifts in land occupation on land-based carbon reservoirs in Uganda between 2006 and 2010.Outcomes indicated a reduction of approximately 29.1% in the sum total of carbon stocks across Uganda, averaging an annual decrease of about 7.3%.An integrated approach involving the Cellular Automata-Markov and InVEST tool was used by [Zhao et al. 2019] to evaluate the contribution of eco-friendly engineering to carbon sequestration in the Heihe River watershed in Northwest China.The developed methodology, with a relative error of 0.22%, predicts these circumstances will contribute to a significant upsurge in carbon preservation by 10.27 Tg from 2015 to 2029 within the studied area.[Hong et al. 2024] examined the carbon sequestration potential of urban forests in the Harbin-Changchun urban agglomeration from 2000 to 2020.The researchers emphasized a gradual increase in carbon sequestration capacity, particularly in large built-up areas.
In Notably, this study represents the first comprehensive investigation into the environmental challenges facing Agadir city.This deficiency serves as the cornerstone of the conducted research, aiming to address the question of how LULC changes impact the local climate-related challenges by examining the correlation between both LST and terrestrial carbon dynamics and urban development patterns in Agadir city over the period spanning from 1984 to 2019.This examination is undertaken through a comprehensive approach that involves: • extraction of LULC change data from Landsat products, utilizing the support vector machine SVM algorithm.

• evaluation of the spatial variations in Land
Surface Temperature (LST) derived from the thermal bands of Landsat imagery.• the identification of carbon stocks and sequestration levels to elucidate their dynamic responses to urbanization and land use transformations.

STUDY AREA
Agadir city is located on the Souss plateau at latitude 30° 25′ North and longitude 9° 36′ West along the Atlantic Ocean (Figure 1).The prevailing climate is semi-arid with mainly Mediterranean oceanic influences characterized by a hot summer season (maximum temperature is recorded in July and August) and a cool winter season (January is the coldest month).The mean temperatures range between 20.04 and 26.04°C throughout the year with the highest in September and the lowest in January [WorldClim 2020].With an mean yearly rainfall of 291.9 mm, the rain falls in Agadir for 39.4 days and the wettest month is December (60.7 mm) while the driest month is July (0.The city is known as a tourist destination of excellence and a seaside resort of choice, the Agadir Regional Tourism Council (CRT) recorded an excess of 5,300.000overnight stays in.In other hands, the developed infrastructures of the port of Agadir, the fishery sector (coastal, deep sea and artisanal) provides a total of 660 000 direct and indirect employments, and it produces 36% of the national added value.These statistics are increased with the installation of the first Halieutic pole in Morocco named 'Haliopolis'.The industrial zone of Agadir city employs more than 33 000 people and comprises more than 138 operational industrial units.The sector of commerce in Agadir is endowed with one of the largest urban markets in Morocco and Africa, the Souk El Had market covers an area of 9 ha, with more than 2 000 shops and 1 200 outlets and its visitors exceeds 50 000 in a normal day [HCP.2020].

Methodology
As mentioned in the following sections, the process of this study will be described in three principal sections.The first one consists of preparing all Landsat scenes downloaded for the years 1984, 1994, 2004, 2014 and 2019.Then, the study field has been subseted from the corrected scenes using a shapefile in which the city is included.The second section is about estimating normalized indices -normalized difference vegetation index (NDVI) and normalized difference built up index (NDBI)) and developing the LULC change maps of the study years.SVM classifier was used, which requires Global positioning system (GPS) points as training samples for the classification process and others as ground truth (GT) for validation.The third section consists of retrieving the LST using thermal bands of the used Landsat products.Finally, the terrestrial carbon model developed by the InVEST was performed by considering the values of three carbon pools for each LULC class obtained from literature reviews.Figure 2 illustrates the research approach utilized in this study.

Data pre-processing
All Landsat scenes of the above-mentioned dates consisted of Level-One Terrain data (L1T), which are orthorectified and radiometrically calibrated [Cao, Gao, and Li 2024].However, atmospheric conditions of each Landsat scene might be different since they were taken on different dates.Hence, fast line-of-sight atmospheric analysis of hypercubes (FLAASH) algorithm was executed on spectral bands of each satellite scene in order to enhance it atmospherically.Further, histogram equalization and resampling techniques have been performed to enhance the quality of the preprocessed images and to bring all images to the same pixel resolution, respectively.

Derivation of LULC maps and normalized indices
First, the obtained images from the pre-processing step are used to create LULC maps by applying the SVM algorithm as a supervised classifier using training samples obtained through field survey and observation of Google Earth archives.The SVM algorithm was chosen, because it performs better than other classical ones, such as maximum likelihood and minimum distance [Aydda et al. 2019].As shown in Table 1, four LULC classifications were studied in this paper.Secondly, NDVI and NDBI were calculated.Both indices range among −1 and +1.Positive NDVI indicates high to moderate green surface, while they emphasize urbanized areas and bare lands covered by more than 5% of earthen material for NDBI.On the other hands, negative values of NDVI express the absence of vegetation and domination of constructed materials or water bodies, while they represent water bodies and areas with vegetation area for the NDBI [Idoumskine et al. 2022].Each index was calculated using Eq. 1 and Eq. 2. (1) (2)

Retrieval of LST
In this part, TM and TIRS thermal bands are employed to extract the spatial variations of LST.First, top of atmospheric (TOA) spectral radiance was calculated using data available from metadata of each scene, by applying Eq. 3 for TM bands and Eq. ( 4 ) − 273.15 °C (5) (2) ) − 273.15 °C (5) where: L λ and L λ1 are the TOP spectral brightness in W/ (m²/sr/mm) of TM and TIRS thermal bands, respectively.L λmax and L λmin are the highest value and the lowest value of spectral radiance, respectively.(Q CalMAX) is the highest quantized calibrated information of pixel (Q Cal ).M L and A L are the radiance multiplier and the radiance supplement scaling element for each TIRS wavelength, respectively.
Subsequently, the effective factor at-satellite (brightness temperature) which is the sensormeasured apparent temperature of the surface was derived from TOA radiance by applying the inverted Planck's Law [Lin, Wei, and Guan 2024].According to [Lin, Wei, and Guan 2024], the Eq. 5 was used for both Landsat products, by using the TOA spectral radiance obtained from the Eq. 3, and the TOA spectral radiance calculated using Eq. 4. The formula of conversion is as following: (5) where: the effective at-satellite temperature is expressed by T sat , K 1 and K 2 are fixed values of the prelaunch calibration ( To retrieve the LST, land surface emissivity (LSE) was computed utilizing the emissivity described by [Sobrino et al. 2004] (Equation 6): where: ε is the LSE, m and n are constant values, and P v is the vegetation proportion, it was computed using least amount and the upper amount value of NDVI applying the Eq. ( 7).
Finally, LST was retrieved from the effective at-satellite temperature using the Eq. ( 8 3).A diachronic analysis was done using the outputs  of this model in order to assess the dynamics of the terrestrial carbon for all the study years.

Regression analysis
The extract point tool of ArcGIS was applied to generate testing value samples of LST, NDVI, NDBI and terrestrial carbon storage for each study year.Then, statistical analysis was carried out using scatter plots regression analysis of Microsoft Office Excel software for all five time-points, i.e., 1984, 1994, 2004, 2014 and 2019 to conduct an analysis of the interrelationships between LST, indices (NDBI and NDVI), and terrestrial carbon storage.

Classification and examination of LULC maps
Five LULC maps were generated from Landsat dataset, all with the maximum overall accuracy that could be obtained after performing the classification iteratively.The spatial expansion of each LULC class and its statistical changes over 35 years are displayed in Figure 3.
The built-up class exhibited a substantial growth with a net increase of 22.16 km², whereas bare land and vegetation cover witnessed a decrease in their areas with a net decrease of -25.17The total accuracy of the Landsat-derived LULC maps is given in Table 4.

LST retrieval
Figure 6 showcases the spatial distribution of LST captured across different time intervals throughout the research.The uppermost records of LST were recorded in 2004 and 2014 with a temperature of 40°C, while the temperatures less than 25°C were identified during 1994 and 2004 in large extent compared to other years.The results showed an average increase of temperature by 3.50°C for the period 1984-2014 and then a decrease by 2.60°C between 2014 and 2019.

Estimation of terrestrial carbon stocks and segregation
According to the output report generated by the model, the amount of carbon storage across Agadir LULC showed an overall stability (Figure 7).The highest rate of land-based carbon storage is reported in the aboveground biomass pool with an amount of 0.10 Tg C in 1984 and 3.33 Tg C in 2019 followed by the belowground biomass (0.001 to 0.8 Tg C) and soil organic.The mean carbon storage of the vegetation landscape in 1984 was estimated at In contrast, the carbon storage of urban areas is reduced from 0.45 Tg C in 1984 to 0.13 Tg C in 2019, due to the significant conversion in LULC types and particularly the loss of vegetation cover.The values about the aboveground biomass during 2019 can be explained by the implantation of a significant quantity of the Florida palm trees over the city.The three pools illustrated positive values of stocking and sequestering terrestrial carbon during all the study years, except for the period 2004-2014 that recorded a negative value of sequestration (Figure 8).

DISCUSSION LULC change detection
Post-classification analysis was carried out on each LULC map for the purpose of calculation the rates of changes of every class of LULC.Three intervals were chosen, namely, 1984-1994, 1994-2004, and 2004-2019.As per the results, in 1984, 59.67% of the total city area was covered by bare land.These surfaces decreased to 31.48% in 2019 in favor of the built-up class that witnessed a significant growth from 31.93% in 1998 to 57.38% in 2019.The green cover showed a low decrease between 1984 and 2019 (from 12.72% to 9.18%), while water class remained its total area between 0.81 and 3.22%.
Agadir LULC changes are due to two significant phenomena, the considerable demographic growth and its rapid economic development.The city witnessed an important urban growth of the city after the homologation of the new urban management plan of the city and the establishment of many urban subdivision plans in order to create new neighborhoods (Salam, Tikiouine, Al Houda, Hay Mohammadi) To address the requirements of the expanding population [Agence Urbaine Agadir 2011].After the earthquake of 1960, significant efforts have been made to bring the shine re-develop the touristic sector along the coastline its natural landscapes.By the end of 1985, 46 hotels with about 2000 to 3000 bed capacity were built in the touristic zone, providing a considerable direct and indirect employment equal to 8000 workers.In addition, a completely new sector called Founty was built over 300 ha in the southernmost of the touristic zone towards the mouth of the Oued Souss.These numbers have continued to increase from 21586 beds in 2001 to 27904 in 2007.In 2019, the capacity of the touristic zone reached more than 40000 beds with an estimation of 550000 employment [HCP. 2020].
Other economic sectors led to the consumption of bare lands in the northernmost part of the city are the industrial activities and fishing industry.The industrial zone of 60 ha and Anza, that lies along the coast to the north of the city; are the two main locations of the industrial activities.Hence, many new residential areas were created next to the two industrial zones and the port in order to respond to the large flows of immigrants/workers to Agadir since the 1970s [Ben Attou and Semmoud 2015].

LULC change and its impact on LST change
The LST values within each LULC class were computed utilizing a statistic tool implemented in ArcGIS software.It was perceived that bare areas exhibited the highest temperatures, followed by urbanized regions.Specifically, the recorded maximum temperatures were 40.96°C in 2014 and 37.91°C in 2019, respectively.The presence of construction materials used in urban development can be seen because of the increase of social and economic needs of a population in exponential growth, therefore having positive regression coefficients of the LST-NDBI model (Figure 9).These results revealed that noticeable up surging of LST is recorded in the areas with high urban footprints, consistently with [Pannunzio et al. 2024] who linked the increasing LST to the areas that witnessed rapid urban growth or bare soil.With the analysis of [Min et al. 2019] who highlighted how the intensity and type of human activities that produced high emissions of pollutants can affect the urban micro-climate and create artificial heat sources.
By way of explanation, the green area and water body showed the lowest LST values with a minimum temperature equal to 20.75°C and 16.52°C, respectively, both recorded in 1984.Furthermore, an unfavorable association between LST and NDVI was identified for all the research time frame (Figure 10), in which the weaker LST were associated with higher values of NDVI.These values are due essentially to the presence of the eucalyptus forest of Aghroud nestled in the southwest side of the field study, covering 12 km² of the surface area, and the increase of urban vegetation especially in gardens, road network and private properties.
These outcomes are dependable with many authors such as [Iqbal and Ali 2022; Kumar et al. 2022] that evaluated the importance of vegetation partitions in regulating the thermal conditions of urbanized zones.[Lin, Wei, and Guan 2024] found that built-up area morphology causes changes in surface heat, so that its albedo decreases than green area, and causes a high LST.

Effects of LULC change on terrestrial carbon dynamic
The outcome of carbon model analysis revealed the effects of the rapid growth of the study area on terrestrial carbon stocks, by calculating the quantity of sequestered or emitted carbon for each LULC class over the period 1984-2019.
Overall, the areas dominated by green cover or maintained with urban vegetation (coastal side) and zones with less road traffic with the   Moreover, Table 5 shows the amount of carbon sequestration for the same period.The maximum capacity of carbon sequestration was estimated to be 0.103 Tg C and it was recorded for the period 1994-2004, whilst the maximum carbon loss recorded in the converted area (from vegetation to built-up area) reaches to -0.127 Mg C during the period 2004-2014.LST values are negatively correlated with terrestrial carbon storage.Thus, the increases of anthropogenic interventions (energy consumption, urbanization, economic activities) and the alarming decline in terrestrial carbon pools such as vegetation coverage will massively increase the amount of greenhouse gas released to the atmosphere.Especially the carbon dioxide (CO 2 ) that affects the LST, harming human wellness, and reducing the climate sustainability of the city.This finding is similar to [Chaddad et al. 2022] and [Chalchissa and Kuris 2024] who devoted to quantifying the consequences of LULC alterations on ground temperature and carbon dynamics.

CONCLUSIONS
Due to the increase in anthropogenic activities over the last few decades, the composition of Earth's biosphere has undergone significant changes.The aim of this study was to discern and evaluate the LULC changes and its impact on the spatiotemporal dynamics of terrestrial carbon stocks and thermal characteristics of Agadir city, utilizing multi-temporal Landsat datasets spanning from 1984 to 2019.
The results illustrate a notable conversion of vegetated and bare lands into urbanized areas.This urban expansion demonstrates a significant linear correlation with LST records, indicating that the weather conditions of this city are becoming more extreme over time.Additionally, this study highlights the impact of urbanization on the physical characteristics of the land surface, contributing to a decline in terrestrial carbon stocks.
The obtained findings indicate that improving vegetation areas has the potential to reduce LST while also offering an opportunity to sequester carbon, emphasizing the imperative of increasing green spaces and implementing appropriate measures to promote afforestation, particularly in barren land areas.
Hence, the findings offer valuable insights for guiding future research efforts on the impact of various policies on LULC changes in semi-arid regions.By utilizing high-resolution data and advanced algorithms, future studies could delve deeper into monitoring the effects of the growing trend in settlement areas on local environmental health and vulnerability.This would enable informed decision-making and facilitate proactive measures to address emerging challenges.
Morocco, studies have attempted to discern LULC impacts on LST [Garouani et al. 2021; Rahimi et al. 2021] and carbon safeguarding patterns [Mosaid et al. 2024; Moussadek et al. 2014] utilizing multi-temporal remote sensing data and focusing on the aspects related to urbanization.Neither of the aforementioned studies analyzed the impact of the land occupation on surface thermal and terrestrial carbon stocks conjunctly.Despite growing concerns regarding LULC changes and its effects on the global ecosystem and its sustainable development, the urban growth of cities in southern Morocco are less broadly studied.Thus, only few attempts have been made to study the LULC change of Agadir city in the Souss Massa region [Idoumskine et al. 2022].
1 mm) [AB-HSM n.d.].According to [HCP 2014], Agadir contains 421 844 inhabitants spread across an area of 112 km².Historically, Agadir showed a significant demographic growth.Its population escalated from 110479 inhabitants in 1982 to 203000 inhabitants in 2006 [Brahim et al. 2021].The total superficies of the city is subdivided in 13 districts and it includes residential areas, industrial areas, touristic areas and the port facilities [Agence Urbaine Agadir 2011].The development economy of Agadir city is based on three sectors, namely fishing industry, tourism, as well as industrial and commercial.

Figure 1 .
Figure 1.Localization of the Agadir city

Figure 2 .
Figure 2. Workflow of the study ) for TIRS bands [Y.Jiang et al. 2023]: NDVI =   −    +  (1) NDBI = ℎ   −   ℎ   −   (2) −   ℎ   − −   ℎ   −     = (   −     )   +    1 =   •   +   ( sat indicates the effective at-satellite temperature, λ express the value of the wavelength emitted by the thermal bands, ρ is h×c/σ [Du et al. 2014], and ε is the LSE already calculated.Terrestrial carbon change Originated by the Natural Capital Project group, the InVEST carbon model is a GIS open tool.It provides information about the amounts of carbon stock in various terrestrial pools [Sharp et al. 2016].The model requires two primary inputs, namely, LULC maps come out from the classification task and quantities of terrestrial carbon stocks in each carbon pool collected from literature reviews [Nel et al. 2022; Sharp et al. 2016] (Table

Figure 3 .
Figure 3. Spatial distribution and percentage of LULC classes for each study year

Figure 4
Figure 4 depicts the spatial distribution and the percentage of each NDVI class.The values

Figure 4 .
Figure 4. Spatial distribution and percentage of NDVI classes for each study year

Figure 5 .
Figure 5. Spatial pattern and percentage of NDBI classes for the study years

Figure 6 .
Figure 6.Spatial pattern of LST

Figure 7 .Figure 8 .
Figure 7. Terrestrial carbon (Tg C) in three land use classes The solar radiation received by LULC types can alike lead to disparity in the LST recorded [Adeyeri et al. 2024].The evapotranspiration of plants and water surfaces raise the atmospheric humidity and thereby abate the ambient temperature[Tao et al. 2024]

Figure 9 .
Figure 9. LST and NDBI correlation Figure 10.LST and NDVI correlation presence of vegetation cover (Identified notably in the northeast and southwest sectors of the region studied) showed high levels of capacity of sequestration.On the other hands, a loss of 0.127 Tg of carbon was observed in dense urbanized areas located essentially at the central part of the city during the period 2004-2014.Several studies discussed the link between LULC dynamic and terrestrial carbon over the world are confirming the finding of this analysis, citing for example [Nguemhe Fils et al. 2018], [Kusi et al. 2020] and [Piyathilake et al. 2022].

Figure 11
Figure 11  highlights the regression analysis of LST distribution in relation with the terrestrial carbon.Numerous studies have examined the correlation between high temperature and global carbon cycle, the key measure of ecosystem mechanism and a critical intersection between the terrestrial biosphere and the Earth's climate[Xiao et al. 2019].Cross-validation output showed that

Figure 11 .
Figure 11.Correlation between LST and terrestrial carbon

Table 2 )
, and the TOA spectral radiance is indicated by L λ in W/(m²/r/μm).The T sat is converted from the Kelvin (K) to Celsius (°C) by multiplying it by −273.15 [Avdan and Jovanovska 2016].

Table 1 .
LULC classes description

Table 3 .
Quantities of terrestrial carbon pools of different LULC classes (Tg/C)

Table 4 .
Evaluation of the LULC maps

Table 5 .
Amount of carbon sequestration