Introduction

Land Surface Temperature (LST) is a key indicator in the assessment of ecosystem changes (Guha & Govil, 2021; Kumar et al., 2022; Przeździecki et al., 2023). The organic and inorganic components of the Earth's surface, through the emission and re-radiation of energy, can reflect levels of molecular stability. For instance, the optimal temperature range for plants in temperate climates and arid tropical regions is 15-30 °C and 25-40 °C, respectively. Temperatures exceeding these ranges trigger a critical phase of thermal stress, which can lead to physiological damage and, in several cases, plant mortality (Hamlyn, 2014; Peng et al., 2016).

The urban heat island approach, traditionally applied in urban environments for public health purposes (Abedrabboh et al., 2025; Stewart & Oke, 2012), can be similarly used to create local zoning or climatopes. These classifications rely on local knowledge of wind patterns, temperature, land use, topography, and population density (Nazarenko et al., 2025; Wang et al., 2025; Žgela et al., 2024). However, boundary delineation can be made more precise using remote sensing technologies (Yantao et al., 2024; Žgela et al., 2024).

LST derived from hyperspectral sensors such as those on-board LANDSAT satellites (U.S. Geological Survey [USGS]) can be used to detect areas under stress from extrinsic factors (Peng et al., 2016; Tulandi & Aloanis, 2021). A lack of vegetation or the presence of senescent vegetation in a given area can manifest as elevated LST values (Bindajam et al., 2020), however, in some cases, LST may be more strongly influenced by other factors, such as precipitation, solar radiation, wind speed, and soil characteristics (Twumasi et al., 2021). Although LANDSAT Collection 2 (LC-L2) products from satellites 8 and 9 are calibrated using global climate datasets from MODIS (USGS, 2024) to estimate LST and surface reflectance, these products have certain limitations that can lead to errors in key variables such as the Normalized Difference Vegetation Index (NDVI) and surface emissivity Ɛ _S (USGS, 2024).

The emissivity coefficient (Ԑ) at a specific wavelength refers to the radiation emitted by a surfaces or object as a fraction of the maximum possible radiation emitted by a blackbody at the same wavelength and temperature (Hamlyn, 2014). Both surface emissivity (Ԑ _s ) and NDVI are used in specific algorithms (Mono-Window or Double-Window) for estimating LST. However, the accuracy of these methods may lead to results that are either underestimated or overestimated (Arabi Aliabad et al., 2021; Wang et al., 2019).

The analysis of the causes and spatial distribution patterns of LST should be conducted over multiple time periods (Zhang et al., 2024) to identify variables that show consistent relationships. For instance, the correlation between LST and ambient temperature recorded by Conventional Meteorological Stations (CMS) in correlation analyses may vary throughout the year (Abulibdeh, 2021; Iqbal & Alí, 2022).

In the arid north of Mexico, where 24.94 % or the national population inhabits, heatwaves tend to be particularly severe, especially during the hottest months of the year (Instituto Nacional de Estadística y Geografía [INEGI], 2020; Servicio Meteorológico Nacional (SMN) & Comisión Nacional del Agua [ CONAGUA], 2024). Therefore, zoning based on the heat island approach holds significant potential for guiding the implementation of mitigation and adaptation measures in response to thermal stress-related risks (Lovriha, 2022).

Based on the above, the objective of this study was to analyze the spatial and temporal distribution of LST and its relationship with LANDSAT spectral indices for the zoning of areas potentially adverse to life. This was proposed under the premise that LST shows a correlation greater that 50 % with ambient temperature measured at CMS, and with environmental drivers related to soil and vegetation, such as NDVI and Ɛ _S .

Materials and Methods

Study area delimitation

The study area covered 42.6 % of the total surface (41 290.41 km²) of the Lagunera Region. It is located at an elevation of 1 139 m in north-central Mexico (102.366° - 104.783° W and 24.366° - 26.383° N) and consists of five municipalities in the state of Coahuila and ten in the state of Durango (INEGI, 2018b; Figure 1a).

The study area was defined based on the LANDSAT Worldwide Reference System (WRS); the daytime (descending) scene WRS2 Path:30/Row:42 was adjusted to a quadrangular coverage of the Lagunera Region, with boundary coordinates at 104.2829° W - 26.5649° N, and 102.9269° W - 25.4189° N.

The study area included 80 % of the municipalities (12 out of 15) that make up the Lagunera Region: Francisco I. Madero, San Pedro, Matamoros, Viesca and Torreón in Coahuila and Lerdo, Nazas, San Luis del Cordero, San Pedro del Gallo, Mapimí, Gómez Palacio and Tlahualilo in Durango.

Components determining land surface temperature

Climate

The predominant climate group is dry (B) with two subtypes: semi-arid (BS) and arid (BW), average annual temperature range of 18 to 22 °C. Three subgroups were identified: 1) BWhw (80.4 %), 2) BS₀kw (12.43 %), BS₀hw (4.66 %) and 3) BS₁kw (2.54 %) (INEGI, 2008; Figure 1b).

Main soil groups, land use and cover

A total of 15 soil groups were identified in the Lagunera Region (Figure 1c), with 13 of these present in the study area. Across an approximate area of 16 716 km², the dominant soil types are: Calcisol (35.3 %), Leptosol (21.1 %) and Regosol (15.8 %) (Figure 1c), followed by Luvisol (5.9 %), Phaeozem (4.7 %), Vertisol (4.1 %), Cambisol (4 %), Solonchak (3.8 %), Solonetz (2.3 %), Gypsisol (1.1 %), Fluvisol (0.4 %), Kastanozem (0.4 %) and Chernozem (0.1 %) (INEGI, 2007a-f).

In terms of land use and vegetation cover (Figure 1d), xerophilous scrubland (56.4 %) predominates, followed by induced vegetation (17.4 %) and irrigated agriculture (16.7 %). Other cover types include grassland (4.3 %), human settlements (2.2 %), rainfed agriculture (1.6 %), bare soil (0.8 %), water bodies (0.12 %), deciduous forest (0.02 %) and coniferous forest (0.01 %) (INEGI, 2018a).

Information preprocessing

This phase involved the search and preparation of basic and complementary inputs to establish the geographical scope and application in the study area. The main data sources were INEGI (2018b) and USGS (2023), from which datasets were downloaded and processed as illustrated in Figure 2. The drivers influencing the LST were integrated according to the methodology described by Wan Mohd Jaafar et al. (2020), as shown in Table 1.

Table 1. Inputs used in the evaluation of land surface temperature in the Lagunera Region.

The analysis and monitoring of LST and its relationship with NDVI, based on LANDSAT datasets, covered the years 2017 and 2021. The selection and download of inputs (Figure 2) followed three criteria: 1) WRS2 daytime scenes, 2) cloud cover ≤15 % and 3) high geometric and radiometric quality. Nine LANDSAT collections were identified and downloaded; the months of March, August and December were excluded due to not meeting quality requirements (Appendix 1).

Post-processing: Estimation of Land Surface Temperature and its Spatial Distribution

To estimate the spatial distribution of LST, the thermal band (B10) was used according to the following equation: LST = [(Qcal x VCa) + VCo] - 273.15; where, Qcal represents the digital numbers (DN) contained in the raster image; VCo and VCa are constants obtained from the metadata (*MTL.txt), the first (0.00341802) used to calibrate DN and the second (149.0) used to convert to the Kelvin (K) scale; and 273.15 is the constant for conversion from K to °C. The spatial distribution of LST was classified into six categories corresponding to critical temperature ranges, as shown in Table 2.

Table 2. Categories, soil temperature ranges and optimal values of selected organisms in the Lagunera Region.

Temperature zonation

A geostatistical analysis of LST was conducted for each month of 2017 and 2021 using QGIS v3.30 (2018). Mean (Tavg) and maximum temperature (Tmax) were determined from the frequency analysis. Monthly temperature zonation was performed by adding 5 °C to Tavg for each of the six classes previously defined (Table 2). LST zonation was carried out through segmentation using the raster calculator with the following algorithm: TSTzi = (TST > Tavg + 5) AND (TST < Tmax); where, LSTzi is the monthly LST z (°C) for the i-th month.

Zonation and field verification

Once the six LST categories were segmented, the location and extent of the zones with the highest temperature records were identified, specifically categories C5(44-56 °C) and C6 (>56 °C). Within these zones, four sites were randomly selected, where temperatures were consistently higher during the studied months (Figure 3).

Field verification was carried out at each of the selected sites (A-D, Figure 3) to document six local factors potentially influencing LST: 1) vegetation type, 2) relative vegetation density, 3) physical condition of the topsoil layer, 4) soil color and 5) land use.

Normalized Difference Vegetation Index (NDVI)

The NDVI was used to determine the distribution and condition of vegetation on the same dates for which the LST was estimated. For this purpose, bands 4 (red) and 5 (near-infrared) were employed, using the following formula: NDVI = (b5 - b4) / (b5 + b4). The spatial relationship between LST and NDVI was classified according to the criteria proposed by López-Pérez et al. (2015), who defined five surface condition categories: 1) <0.01 for clouds, water bodies, or other non-vegetative elements; 2) 0.01-0.1 for bare soil; 3) 0.1-0.2 for areas with sparse vegetation; 4) 0.2-0.4 for areas with moderate vegetation cover; and 5) >0.4 for areas with dense vegetation cover.

Soil information

The field data collected were compared with the soil profile identification data provided by INEGI (2013) for the upper soil layers at depths of 0-10 and 0-15 cm. The variables considered included texture, color, and hue, organic matter content, field capacity, permanent wilting point, and electrical conductivity.

Statistical analysis

In addition to field verification, a vector layer was generated in QGIS v 3.30 (2018) with 500 random points, which were used as sampling sites on the LST, Ԑs and NDVI images of the 18 raster scenes analyzed. Using this data, a simple linear regression analysis was carried out, regarding LST as the independent variable, the standard error, and the uncertainty (I = s * 2/LSTa), where s is the standard deviation and LSTa is the average LST. Statistical analyses were carried out in Excel 2016 (Ms Office).

The coefficient of determination (R²) between TA and LST was determined using the monthly extreme maximum temperature data from 10 weather stations located in the study area, obtained from SMN-CONAGUA (2023), and compared with the maximum temperatures of each analyzed scene.

Results

LST - NDVI spatial distribution

An inversely proportional relationship was observed between NDVI and LST. When the NDVI approaches zero, LST increases throughout the study area. According to the five LST ranges established (Table 2), the predominant classes during 2017 and 2021 were C3(20-32 °C), C4(32-44 °C) and C5(44-56 °C), with extreme anomalous values corresponding to category C6(>56 °C). The highest and most persistent LST, maintained throughout the nine months, was detected in the northeastern portion of the study area in the municipalities of Francisco I. Madero (009) and San Pedro (033), and in the southeastern portion in Matamoros (017) and Viesca (036) (Figure 4). According to the NDVI values (0.01-0.1), these areas showed predominantly bare soil conditions.

Reports from weather stations indicate that the hottest months were May and June, with temperatures ranging around 40 °C (Figures 5b and 5d). The LST estimated showed that the temperature ranges of classes C5(44-56 °C) and C6(≥56 °C) had the greatest impact on the territory in both years. In 2017, during June and July, C6 affected 40 to 50 % of the study area. In 2021, C5 was associated with April and May; when the maximum temperature barely reached 35 °C (Figure 5d), LST in May affected nearly 80 % of the study area (Appendix 2; Figure 5).

For most of the study areas in both years, two primary land cover types were observed: a) bare soil (0.01-0.1) and b) sparse vegetation (0.1-0.2), with annual averages of the total area ranging from 40-55 % and 39-55 %, respectively. Areas with semi-evergreen forest (0.2-0.4) and high vegetation cover (>0.4) range between 5 and 6 %. It was also found that during the driest months (November-May), the bare soil conditions in 2017 and 2021 correspond to 46 and 65 %, respectively (Figure 6; Appendix 3).

Field verification

The information collected at the four sites was compiled into five components, as shown in Table 3. Graphical evidence is presented in Figure 7.

Table 3. Characteristics of the field verification sites used to gather evidence on local factors that could influence land surface temperature in the Lagunera Region.

Soil properties of the verified sites

The information collected at the four sites (Table 3), when compared with the soil profiles, confirmed the presence of three soil groups: Leptosol, Calcisol and Solonetz. Except for site D, the textural properties reveal a predominance of fractions ˃0.05 mm, where silts and sands are grouped, resulting in a low available moisture capacity ranging from 3.2 to 6.8 % (Appendix 4).

Degree of relationship among the studied variables

Linear regression between LST and air temperature

Between LST and the air temperature recorded at the 10 weather stations, it was found that in 2017, 95 % of the stations showed R² values ˃ 0.7, except for the weather station 5139 (R² = 0.589); however, in 2021, this station and stations 5180 and 10085 showed a strong relationship (R² > 0.9) (Table 4).

Table 4. Correlation coefficients (R²) estimated between air temperature (data from the weather station and land surfaces temperature in the Lagunera Region.

Linear regression of LST with spectral indices

In general, the linear regression analysis of the LST-Ԑs AND LST-NDVI pairings indicates a good relationship (Table 5), with LST as the independent variable and NDVI or Ԑs as the dependent variables (NDVI = β0 + β1*LST and Ԑs = β0 + β1*LST). Although Ԑs is a better predictor than NDVI; NDVI shows an average uncertainty 2.9 times lower than Ԑ _s when calculating LST for all analyzed months in both years

Table 5. Statistical parameters derived from the linear regression analysis, with estimated land surface temperature in the Lagunera Region as the independent variable.

Discussion

High temperatures and probable causes

In hot deserts, large temperature variations between day and night are common. Morning radiative incidence initiates a surface heating process, with daytime maximum surface temperature reaching up to 80 °C, exceeding air temperature (Farhad & Monger, 2020).

In recent years, arid regions, particularly in North America (Yan et al., 2020), have recorded unusually high temperatures (25-40 °C) that are attributed to changes in precipitation and temperature patterns resulting from prolonged droughts, water scarcity, high solar radiation, and low vegetation cover (Gohain et al., 2021; Tariq et al., 2020). This study documents some of these effects in the year 2017 and 2021, which experienced more intense droughts and higher-than- normal temperatures (CONAGUA SMN-, 2025). Furthermore, the close relationship between elevated temperatures and land surface temperature (LST) observed in this study is consistent throughout the analysis (Figures 4 and 7).

The behavior of LST in terms of its maximum extent (˃12 000 km²) and magnitude for C6(˃56 °C) reflects a regional daytime energy balance, in which sensible heat is primarily concentrated in the upper soil layer and at the soil-atmosphere interface. This pattern agrees with the air temperature values observed in the EMC, suggesting a possible transition from arid to hyper-arid conditions. According to Yan et al. (2020), LST in North America has increased by an average of 1.92 °C in recent years (2002-2018), with maximum values reaching 67.73 °C, while the Baja California Peninsula shows the highest average LST at 32.41 °C. Furthermore, Žgela et al. (2024) report contrasting LST values between urban and natural environments.

Evaluation of indicators: Surface emissivity and NDVI

From a thermal perspective, emissivity Ԑ is an ideal indicator that reveals the intrinsic characteristics and functioning of the surface upon which shortwave radiative energy (R _oc ) is incident and subsequently reemitted as longwave energy (R _ol ). Emissivity can be attributed to the atmosphere (Ԑ_a), bare soil (Ԑ_sd) or vegetation (Ԑ_v). This concept is grounded in blackbody theory, which states that “any body can absorb and emit energy across all wavelengths” (Hamlyn, 2014).

In calibrated LANDSAT (L2-C2) products, as in this case, Ԑs is considerably more accurate due to the sensor’s enhanced radiometric, spectral, spatial, and temporal resolution. This likely provides greater accuracy than studies relying on algorithmic approaches (e. g., ‘Split Window’ and ‘Doble Window’), which are often labor-intensive (Mujabar, 2019; Wang et al., 2020) and generally use default emissivity Ԑ values associated with NDVI (Arabi Aliabad et al., 2021).

Abulibdeh (2021) analyzed the relationship between LST and NVDI in the arid and semi-arid regions of the Persian Gulf and found that bare soils show higher temperatures than urban and vegetated areas, with variations of 1 to 2 °C for the former and 1 to 7 °C for the latter. Meanwhile, Iqbal and Alí (2022), in a semi-arid to arid environment in northwestern Pakistan, demonstrated that linear and quadratic models are effective for estimating air temperature from LST, yielding Pearson correlation coefficients of 0.62 and 0.78 in summer and 0.59 and 0.78 in winter.

Additional indicators

Statistical parameters

The negative coefficients (β₁) representing the slopes of the linear regression models obtained (Table 5), for both NDVI and Ԑs, are consistent with the triangle theory proposed by Goward et al. (1985) and applied in studies of similar conditions (Kumar et al., 2022; Tang et al., 2010) in the study area. In other words, as both indices approach zero, LST increases. This temperature rise is particularly pronounced in the study region, where LST exceeded twice the normal values during the hottest month. These observations are in agreement with the BWhw climate type, which covers nearly 80.4 % of the area according to INEGI (2008).

Soil component

Leptosols are very thin soils, often stony, with a high calcium carbonate (CaCO₃) content and are highly susceptible to erosion. The vegetation that predominantly develops on these soils is rosetophilous desert scrub. Calcisoils are characterized by a high CaCO₃ content (>15 %) in the surface layer, which may be cemented. They are primarily found in arid regions in the northern part of the country. Solonetz soils have a more alkaline pH (>8.5) with subsurface layers commonly hardened. These soils are typically located in hot dry climates with high sodium concentrations (González & Chávez-García, 2025).

On the other hand, Bindajam et al. (2020) reported that in semi-arid regions, the highest LST values are observed in rocky areas and on built-up surfaces. In contrast, Singh et al. (2023) found a low correlation (r = 0.12) between LST and calcium carbonate (CaCO₃). This weak association is attributed to the prevalence of light-colored soils, which reflect a larger proportion of incident solar radiation. The reflected radiation is captured by the thermal bands of satellite sensors, resulting in apparently elevated temperature readings. Additionally, a high clay content alters the specific heat of the soil by reducing the proportion of macropores, thereby limiting air and water circulation within the soil profile (Rucks et al., 2004).

Conclusions

A strong correlation was confirmed between land surface temperature (LST), spectral indicators, and air temperature. The spatiotemporal distribution analysis indicates that the highest LST values are directly associated with soil degradation characteristics and sparse vegetation cover. This underscores the significant influence of anthropogenic factors, such as vegetation cover and land use change, as key controllers of LST, and as their potential impacts of short-term weather events (heat weaves) and medium-term phenomena (droughts), and their likely contribution to climate change, which may threaten the establishment of wild species. This study can serve as a basis for estimating LST at larger scales and support public policy decisions related to sustainable development, climate change mitigation, natural area restoration, and agricultural planning.