Next Article in Journal
Analyses of MODIS Land Cover/Use and Wildfires in Italian Regions Since 2001
Previous Article in Journal
How Do Natural Environmental Factors Influence the Spatial Patterns and Site Selection of Famous Mountain Temple Complexes in China? Quantitative Research on Wudang Mountain in the Ming Dynasty
Previous Article in Special Issue
Mapping Habitat Suitability of Migratory Birds During Extreme Drought of Large Lake Wetlands: Insights from Crowdsourced Geographic Data
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Land
Volume 14
Issue 7
10.3390/land14071442
land-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Historical Land Cover Dynamics and Projected Changes in the High Andean Zone of the Locumba Basin: A Predictive Approach Using Remote Sensing and Artificial Neural Network—Cellular Automata Model

by
German Huayna
1,
Victor Pocco
1,
Edwin Pino-Vargas
1,*,
Pablo Franco-León
2,
Jorge Espinoza-Molina
3,
Fredy Cabrera-Olivera
4,
Bertha Vera-Barrios
5,
Karina Acosta-Caipa
3,
Lía Ramos-Fernández
6 and
Eusebio Ingol-Blanco
7
1
Department of Civil Engineering, Jorge Basadre Grohmann National University, Tacna 23000, Peru
2
Laboratory of Ecological Processes, Research Group of Arid Zones, Deserts and Climate Change (ADERIZA), Jorge Basadre Grohmann National University, Tacna 23000, Peru
3
Department of Architecture, Jorge Basadre Grohmann National University, Tacna 23000, Peru
4
Department of Geological Engineering-Geotechnics, Jorge Basadre National University, Tacna 2300, Peru
5
Faculty of Mining Engineering, National University of Moquegua, Moquegua 18001, Peru
6
Departament of Water Resources, Universidad Nacional Agraria La Molina, Lima 15024, Peru
7
Department of Civil Engineering, New Mexico State University, Las Cruces, NM 88003, USA
*
Author to whom correspondence should be addressed.
Land 2025, 14(7), 1442; https://doi.org/10.3390/land14071442
Submission received: 30 April 2025 / Revised: 3 July 2025 / Accepted: 8 July 2025 / Published: 10 July 2025
(This article belongs to the Special Issue Land Use/Land Cover Mapping Changes Using Earth Observations Techniques)
Browse Figures
Versions Notes

Abstract

The conservation and monitoring of land cover represent crucial elements for sustainable regional development, especially in fragile high Andean ecosystems. This study evaluates the spatiotemporal changes in land use and land cover (LULC) in the Locumba basin over the period of 1984–2023. A hybrid modeling approach combining artificial neural networks (ANN) and cellular automata (CA) was employed to project future changes for 2033, 2043, and 2053. The results reveal a significant reduction in glaciers and lagoons throughout the Locumba basin, with notable declines from 1984 to 2023, while vegetated areas, particularly grasslands and wetlands, experienced substantial expansion. Specifically, grasslands increased by 273.7% relative to their initial coverage, growing from 57.87 km2 in 1984 to over 220.31 km2 in 2023, with projections indicating continued growth to over 331.62 km2 by 2053. This multitemporal analysis provides crucial information for anticipating future land dynamics and underscores the urgent need for strategic conservation planning to mitigate the continued loss of strategic ecosystems in the high Andean region of Tacna.

1. Introduction

Natural reserves worldwide are critical hotspots of biodiversity and multiple land uses, typically corresponding to areas where farming, grazing, tourism, and other human activities occur [1]. Among these, high Andean ecosystems maintain exceptionally high levels of flora and fauna diversity across diverse topographic gradients [2]. These ecosystems provide crucial services, notably the regulation of water supply for cities and agricultural valleys downstream [2,3,4].
Mountain landscapes provide multiple ecosystem services that are highly vulnerable to soil changes [5,6]. In degraded Andean basins, changes in forest cover have been associated with increased annual water yield and reduced sediment production, with reforestation and vegetation recovery contributing significantly to these improvements [7]. However, they are increasingly vulnerable to anthropogenic pressures and environmental change.
In the South American Andes, recent studies have highlighted strong spatial and temporal variability in precipitation and temperature [8,9]. Long-term observations reveal a decrease in precipitation since the 1980s [10], while mean precipitation patterns are highly influenced by geographic location and topographic conditions [4,11]. Climate change, driven by increased greenhouse gas concentrations, is expected to further modify temperature and precipitation regimes [12]. Globally, precipitation patterns are already shifting in quantity, intensity, and frequency [13]. Within the Locumba Basin, located in a hyper-arid zone, hydric availability issues are becoming increasingly critical, with direct impacts on the sustainability of agriculture [14,15,16].
The growing dependence on groundwater to address water deficits has generated serious environmental challenges in the region, such as marine intrusion into coastal aquifers and the desertification of high-altitude wetlands due to overexploitation [17,18]. Notably, the Caplina aquifer presents fossil characteristics and very low recharge rates [19,20], while the high Andean aquifers are exposed to intense natural contamination processes associated with the region’s volcanic geology [20].
Monitoring and anticipating land use and land cover (LULC) dynamics are crucial for safeguarding fragile ecosystems. Recent research on land use and land cover (LULC) dynamics has increasingly focused on understanding the interplay between anthropogenic pressures and environmental drivers in mountainous and fragile ecosystems [21]. For example, Yang et al. (2025) analyzed rangeland degradation in Pakistan’s semi-arid highlands from 2000 to 2020, revealing substantial declines in vegetative cover linked to agricultural expansion using Landsat time series and machine learning classification methods [22]. Likewise, a broad review by Shadmehri Toosi et al. (2025) emphasized the importance of multi-temporal LULC datasets and classifier selection in hydrological modeling frameworks, highlighting how inconsistent land cover products may misrepresent surface processes unless classification schemes are carefully aligned [23]. Studies in the Andes and other highland regions have demonstrated how LULC changes are closely linked to climate variability, topographic gradients, and local land management practices [5]. These findings demonstrate both the methodological and ecological value of integrating long-term LULC analysis into landscape planning in fragile mountain ecosystems.
In this sense, remote sensing technologies have emerged as indispensable tools for detecting and monitoring LULC dynamics over extended temporal and spatial scales. The integration of multi-spectral satellite imagery with geospatial modeling techniques enables detailed classification of land cover types and the identification of subtle changes across complex terrains [24,25]. Platforms such as Google Earth Engine (GEE) have democratized access to satellite archives and processing capabilities, significantly enhancing the scalability and replicability of LULC studies [26]. This platform has recently been considered a powerful tool for processing and analyzing large volumes of remote sensing data [27]. In particular, spectral indices such as NDVI, NDWI, and NDSI have proven effective in delineating vegetation, water bodies, and snow-covered areas, while elevation data and terrain metrics further refine the spatial resolution of classification outputs. This synergy between remote sensing tools and geostatistical modeling provides a robust foundation for predictive land change analysis in high mountain ecosystems.
In this context, time series analysis emerges as a powerful tool to characterize dynamic dependencies and predict future values in complex systems, utilizing approaches such as autoregressive models, complex networks, and deep learning [28,29,30,31]. Most time series methods assume that any trend will continue unabated, regardless of the forecast time frame; however, recent findings suggest that forecast accuracy can be improved by attenuating or altogether ignoring trends with a low probability of persistence [32,33].
Among the modeling approaches for LULC change, the CA–Markov method has gained popularity by combining cellular automata, Markov chain analysis, and multi-criteria evaluation [34]. Recent studies suggest that the CA–Markov model is effective in predicting changes in use and land cover [35,36]. These models also have applications in analyzing urban growth and spatial trends, with implications for urban planning, hydrological impact assessments, and environmental management [36,37,38]. Cellular automata and Markov chain modeling are effective in geospatial environmental modeling, aiding in land use and watershed management, but limitations must be addressed and advancements are needed for future research [39,40,41]. It is important to note that there are multiple ways in which a stochastic system can deviate from statistical equilibrium [42,43].
Exploratory data analysis consists of detecting and describing patterns, trends, and relationships in the data, motivated by certain research purposes [44,45]. The inference, prediction, and control of complex dynamical systems from time series is important in many areas of knowledge, including climate and water modeling [43].
Cellular automata (CA) are discrete simulation models that produce spatiotemporal data through experiments, as well as stochastic models, which generate data from multiple runs [46,47]. The performance and scalability of cellular automata, in some cases, are limited by the need to synchronize all nodes at each time step; one node can be executed only after the execution of the previous step at all other nodes [47,48]. These studies suggest that cellular automata can be effectively used for time series trend analysis by modeling interactions, trend detection, and forecasting with methods such as stochastic cell automata, two-dimensional cell automata, and the analysis of seasonally adjusted fluctuations-based multifractal that show promising results [42,43,44,46,47,48].
While numerous studies have examined LULC changes in mountainous regions across the globe, there is still a noticeable gap in future-oriented simulations specifically focused on semi-arid high Andean basins. Research in the Colombian and Ecuadorian Andes has extensively documented historical transitions between cropland, pasture, and vegetation, underscoring links to demographic, socioeconomic, and topographic drivers [49]. In coastal Peruvian watersheds, studies contrasting MODIS and ESA-CCI datasets have highlighted discrepancies in detected LULC trends, emphasizing the need for spatially consistent, high-resolution analyses [50]. Similarly, hydrology-focused LULC projections in Chilean basins have shown the practical value of CA–Markov models for water balance applications, but these have not been extended to adjacent Andean contexts [51]. Moreover, global reviews of Andean ecosystem dynamics demonstrate strong interest in how LULC affects hydrological services, yet they seldom include basin-scale future simulations integrating both land cover change and ecosystem function [52].
Despite the recognized ecological importance of high Andean ecosystems, long-term analyses of land cover evolution in vulnerable basins such as the Locumba remain scarce. This lack of forward-looking analyses limits the ability of planners and conservationists in southern Peru to anticipate landscape trajectories under ongoing climate and land use pressures. In this way, a spatiotemporal variation in LULC is expected to happen in the future, yet no study has attempted to simulate the future LULC dynamics and their effect on the Locumba basin. Addressing this gap is essential for anticipating future environmental challenges and informing sustainable management practices.
Given this context, the main objective of this study is to evaluate the spatial and temporal dynamics of wetlands, biomass, and water bodies in the high Andean region of the Locumba Basin over the period of 1984–2023. Additionally, we develop a predictive modeling framework combining artificial neural networks (ANN) and cellular automata (CA) to forecast LULC changes for the years 2033, 2043, and 2053. This approach aims to support strategic planning for ecosystem conservation and water resource sustainability in a region increasingly threatened by climate variability and anthropogenic pressures.

2. Materials and Methods

2.1. Study Area

The Province of Candarave is located in the northeastern corner of the Tacna Region between the geographic coordinates 16°17′04″ S and 17°27′56″ S latitude and 70°03′32″ W and 70°34′52″ W longitude. Its altitude varies from 2400 to 5500 m.a.s.l (Figure 1a). Most of the study area has a semi-arid climate with a dry winter and spring (between 10 mm and 15 mm of annual rainfall) and a semi-frigid climate with a dry winter (between 158 mm of annual rainfall). The study area presents an average maximum temperature of 20 °C and an average minimum temperature of 3 °C. Additionally, evapotranspiration increases progressively throughout the year, starting with an average of 100 mm in January and reaching about 145 mm by December (Figure 1b).

2.2. Methodology

This study applied a hybrid remote sensing and predictive modeling approach to evaluate LULC dynamics between 1984 and 2023 and to project future changes for 2033, 2043, and 2053. Figure 2 presents a flowchart summarizing the main methodological steps, including (a) satellite data acquisition, (b) image preprocessing, (c) LULC classification, (d) accuracy assessment, (e) the cellular automata model, and (f) prediction.
Regarding the data used, satellite images from the Landsat-5 TM, Landsat-7 ETM+, and Landsat-8 OLI sensors were accessed through the Google Earth Engine (GEE) platform (https://earthengine.google.com, accessed on 20 March 2025) (Table 1). Scenes were selected based on low cloud coverage and their suitability for representing hydrological conditions during the dry season (Julian days 135–227). The GEE function imageCollection.reduce() was used to calculate the pixel-wise average for the defined period. Subsequently, cloud/shadow reduction algorithms were applied, followed by the calculation of normalized spectral indices oriented toward identifying each LULC class within the study area. The accuracy of the generated maps was evaluated through statistical validation using the AcATaMa plugin in QGIS software 3.42.2 Münster, released in April 2025, for the representative years.

2.3. Climatic and Topographic Datasets

In order to describe the topographic and climatic conditions of the study area, a variety of geospatial datasets were utilized. Elevation-derived variables—including slope, aspect, shape index, horizontal curvature, vertical curvature, topographic position index, and terrain ruggedness index—were extracted from the Shuttle Radar Topography Mission (SRTM) dataset at a spatial resolution of 30 m, available through the Google Earth Engine platform. Additionally, several climate-related variables were incorporated from different sources. Land surface temperature (LST) was estimated from Landsat 8 imagery using a code repository developed by Sofia Ermida on the Earth Engine platform [53]. For broader-scale climatic variables, such as annual precipitation and surface shortwave radiation, the TerraClimate dataset was employed, offering a spatial resolution of approximately 4.6 km (Table 2).

2.4. Identification of Land Cover Types: Glacier, Lagoon, Wetland, Agriculture, and Grassland

The identification of glaciers, lagoons, wetlands, agriculture, and grasslands involves a combination of normalized spectral indices and thresholding methods to accurately classify these land cover types. For glaciers, the Normalized Difference Snow Index (NDSI) was applied to distinguish glacial surfaces from other land types. The NDSI was calculated using an empirical threshold of 0.4, based on spectral profiles of glacier surfaces in the study area. To further improve glacier detection and reduce false positives, additional constraints were applied using spectral bands. Specifically, pixels were selected where the near-infrared (NIR) reflectance was >0.11 and the red band reflectance was >0.1. To address classification errors caused by debris-covered glaciers, visual inspections and manual corrections were made.
For lagoon identification, the Normalized Difference Water Index (NDWI) was employed to extract water bodies from satellite imagery. The NDWI was calculated with the formula shown in Table 3. The threshold for identifying lagoons was set at NDWI > 0.2, and additional restrictions were applied to exclude shadowed areas (slope ≤ 20 degrees and DEM > 3800 m.a.s.l.).
In the case of wetlands, classification was refined using two normalized indices: the NDVI (Normalized Difference Vegetation Index) and NDII (Normalized Difference Infrared Index). The NDVI threshold was set at NDVI > 0.43, while the NDII threshold was defined as 0.02 < NDII < 0.76. These indices, along with elevation data (SRTM DEM > 3800 m), further helped to delineate wetland boundaries more accurately, ensuring a precise classification of wetland areas.
For agriculture, classification was performed by identifying areas with NDVI > 0.20 and located at elevations lower than 3800 m.a.s.l., which excludes high Andean ecosystems like wetlands and grasslands. This combination highlights productive, vegetated zones typically associated with cultivated land.
Grasslands were detected using a composite approach by selecting areas with NDVI > 0.2, located at elevations equal to or higher than 3800 m.a.s.l., and explicitly excluding wetland zones. This method allowed for the identification of dry, high-altitude vegetated zones that do not meet the criteria for wetlands but still present substantial vegetative cover.

3. Results

3.1. Historical Land Cover Analysis

3.1.1. Markov Transition Matrix Analysis

The transition matrix is a key tool for analyzing the evolution of land use and land cover (LULC) classes over time [54,55]. This matrix provides a quantitative summary of transformations between categories by comparing classified maps from different time periods through a pixel-by-pixel evaluation. Its utility lies not only in identifying net changes by class but also in revealing specific transition trajectories. This approach is particularly valuable in land dynamics studies that employ satellite imagery to assess LULC change. The procedure incorporates four core probabilistic components: the state transition matrix and a set of images representing conditional probabilities for the periods 1984–1993, 1993–2003, 2003–2013, and 2013–2023, based on existing LULC conditions and associated driving variables (Table 4).
During the 1984–1993 period, the classes of bare soil, grassland, and agriculture exhibited high stability, with persistence probabilities of 0.980, 0.891, and 0.799, respectively. However, notable transitions occurred: 1.6% of bare soil converted to grassland, 10.3% of grassland changed to bare soil, and 20.1% of agricultural areas were transformed into bare soil. Between 1993 and 2003, bare soil, wetlands, glaciers, lagoons, and grassland maintained high stability, with persistence values above 0.781. Still, relevant transitions were detected: 1.9% of bare soil changed to grassland; 10.5% of wetlands and 10.8% of lagoons also shifted to grassland; in addition, 15.2% of glaciers and 15.0% of grassland converted to bare soil.
In the 2003–2013 period, bare soil, wetlands, lagoons, grasslands, and agriculture continued to show high stability, with persistence probabilities exceeding 0.838. Nonetheless, significant transitions were observed: 4.9% of bare soil and 16.2% of wetlands were converted into grassland, while 2.9% of lagoons and 16.2% of agricultural land changed to bare soil. Additionally, 5.0% of grassland transitioned into wetlands. Between 2013 and 2023, bare soil, wetlands, grassland, and agriculture again showed high stability, with persistence probabilities above 0.771. However, notable transitions occurred: 3.2% of bare soil and 21.7% of wetlands converted to grassland, while 8.2% of grassland and 22.9% of agriculture changed to bare soil.

3.1.2. Accuracy of the Methodology in Land Cover Mapping

To validate the evaluation, LULC maps were assessed for the representative years 1984, 1993, 2003, 2013, and 2023 (Table 5). A total of 390 random points were selected within the study area, considering a 95% confidence level (z = 1.96), a correct proportion (p) of 95%, and a maximum tolerable error (e) of 1.4%. The validation process was carried out using QGIS software 3.42.2 Münster with the AcATaMa plugin for the accuracy assessment of each representative map.
The results of the validation show us that on each map, there is an overall accuracy ranging from 0.93 to 0.96, indicating that the results were successful in the classification of Wetland (W), Glacier (GL), Lagoons (L), Grassland (GH), and Agriculture (A). Additionally, the Kappa coefficient values ranged from 0.91 to 0.95, which, according to the standard interpretation scale (Table 6), correspond to a strong to near-perfect level of agreement, further supporting the robustness and consistency of the classification results.
The distribution of the 390 sampling points was carried out using a proportional stratified approach, prioritizing land cover classes with greater spatial extent and/or thematic relevance within the context of the study. In particular, classes such as bare soil and grassland were assigned a larger number of points, as their representation is critical for analyzing land use and land cover (LULC) change. This strategy enhances the statistical robustness of the analysis by improving the accuracy of commission and omission error estimates, especially for dominant classes, which are typically more prone to spectral variability and classification errors.

3.1.3. Temporal Evolution of Coverage

Figure 3 shows the temporal evolution of our indices, as well as the fluctuations and trend curves of each LULC category from 1984 to 2023, including projected predictions for the years 2033, 2043, and 2053. In the case of wetlands, these have shown an increase since their initial period in 1984, growing from an area of 2.93 km2 to 10.96 km2 in 2023. It is projected that over the next 30 years, wetlands will experience a slight increase, reaching approximately 11.32 km2 by 2053. In the case of glaciers, they have experienced a severe decrease in their areas, going from an initial area of 19.74 km2 to a minimum of 3.95 km2 in 2023. It is projected that over the next 30 years, glaciers will maintain approximately this minimum coverage, reaching about 3.13 km2.
According to Table 7, which presents the land cover change results expressed in square kilometers (km2), notable trends are observed across various land cover classes throughout the historical and projected periods. Regarding lagoons, showing a similar trend to glaciers, they have presented a considerable decrease in their total initial area, decreasing from 16.87 km2 to 11.45 km2 in 2023. It is projected that by the year 2053, lagoon areas will maintain a total of 11.24 km2. Focusing now on grasslands, this land cover type has shown a considerable increase in its available areas, growing from an initial area of 57.87 km2 to a total of 220.31 km2 in 2023. It is projected that these areas will continue increasing over the next 30 years, reaching totals of 257.65 km2, 294.8 km2, and 331.62 km2 for the years 2033, 2043, and 2053, respectively. In the case of agricultural areas, these have maintained approximately the same dimensions since 1984, remaining in the range between 32.8 km2 and 37.29 km2 of total area. It is projected that over the next 30 years, agricultural areas will maintain the same level of coverage within the Locumba basin.
Thus, it is confirmed that the trend of each variable has maintained a consistent trajectory from the past to the present and is expected to continue along the same path over the next 30 projected years (Table 7). Agricultural areas, wetlands, and grasslands will continue to expand relative to their current extent, while lagoon and glacier areas will experience further reductions over the same period. This finding regarding glacier retreat in the Barroso Mountain Range reveals that vast areas of terrain previously covered by ice have been exposed, creating new surfaces that, in many cases, have been colonized by vegetation. This revegetation process has been documented in various regions, where vegetation expansion has been attributed to glacial melting and the subsequent stabilization of the soil by emerging vegetation [56,57].

3.1.4. Time Series Classification

The time series maps reveal the evident changes that different land cover types have undergone. To analyze in greater detail the changes within our study area, three representative zones, A, B, and C, were selected, covering the most important transformations identified in our research (Figure 4). Zone A encompasses areas of grasslands, wetlands, and, most notably, the location of Lake Vizcachas. Between the years 1984 and 1993, Lake Vizcachas underwent an almost complete disappearance, a condition that remained stable until 2003. Between 2003 and 2013, a reappearance of the lake occurred, with its dimensions nearly returning to those observed back in 1984. However, over the following ten years, a new phase of disappearance took place, as evidenced by the fact that by 2023, and continuing to the present, Lake Vizcachas has completely lost its presence within the headwaters of the Locumba basin. The disappearance of lagoons in mountainous regions can be attributed to a combination of factors, among which accelerated glacial retreat plays a significant role [58,59].
Zone B mainly consists of glacier areas located at the peaks of the mountains belonging to the Barroso Mountain Range, along with grassland zones and small wetland areas situated at lower elevations. The primary change observed over the past 40 years was the reduction of glacier-covered areas at the mountain summits. Regarding this phenomenon, an increase in grasslands and wetlands was observed in areas with gentler slopes. Regarding Zone C, this area is predominantly composed of glaciers, agricultural zones, and, most notably, abundant areas of grasslands. The changes observed between 1984 and 2023 include a clear increase in grassland areas, along with a considerable reduction in the small glacier remnants located in this zone. Additionally, agricultural land has shown a slight decline from its initial dominance. The substantial expansion of grasslands observed in our study area can be primarily attributed to the increase in mean annual temperature and the interannual variability of precipitation, factors that have been identified globally as significant drivers of vegetation change in other study areas [60,61].
Additionally, Figure 5 shows the gain vs. loss metrics of the covers in the historical period 1984–2023. This graph reveals how, for the period 1984–1993, agriculture experienced a positive net increase of 9.2% in new areas compared to its initial dimension, grasslands had an increase of 69.2% in total areas, lagoons suffered a reduction of −33.0% in relation to their initial areas, glaciers reduced their initial areas by −80.0%, wetlands suffered a reduction of −55.9% in their total areas, and in the case of unoccupied bare land, it decreased by −1.2%.
In summary, the historical analysis conducted between 1984–2023 shows that the covers that have experienced the greatest increase in initial areas were the green areas of grasslands and wetlands, with a total net change of 280.7% and 273.7%, respectively, followed by agriculture, which only had a net change of 13.7% compared to its initial dimension. On the other hand, the covers that drastically reduced their initial areas were the water sources of glaciers and lagoons, with a total net change of −80.0% and −32.1%, respectively, followed by bare soil, which experienced a reduction of −8.9% in its initial area.

3.2. Projection of Future Scenarios

3.2.1. Spatial Variable Evaluation for Predictive Modeling

To predict the future trends of the different LULC classes for the next 30 years, the hybrid methodology of artificial neural networks (ANN) and cellular automata (CA) was used. To create the prediction model, it is necessary to establish predictor variables; these variables were processed within the MOLUSCE plugin usable in QGIS to estimate the changes and trends for the proposed future projections. Figure 6 shows the correlation matrix between the 11 predictor spatial variables. The figure presents pairwise relationships, with the scale ranging from values between −1.0 to 1.0, where negative values in blue indicate a negative correlation and the higher values in red represent a positive correlation.
The correlation matrix provides a comprehensive assessment of the interactions between topographic and climatic variables that influence land use and land cover dynamics. Beyond the individual coefficients, a consistent structure is observed between geomorphological and climatic factors, allowing the identification of ecologically relevant patterns. Strong positive correlations—such as the one between slope and terrain ruggedness index—suggest potential redundancy among predictors, which may be considered during variable selection or dimensionality reduction processes. In contrast, negative correlations between elevation and climatic variables such as land surface temperature and shortwave radiation highlight well-defined altitudinal environmental gradients. These gradients play a key role in the spatial distribution of ecologically sensitive land cover types, such as high Andean wetlands and grasslands. Finally, the presence of weak correlations between certain variable pairs indicates statistical independence—a desirable condition for inclusion in multivariate models, as it minimizes the risk of multicollinearity.
Additionally, Figure 7 visually shows the 11 predictor variables considered. These variables have been divided according to their influence factor, having Topographic Variables (elevation, aspect, slope, shape index, horizontal curvature, vertical curvature, position index, and rigidity index) and Climatic Variables (land surface temperature, precipitation, and surface shortwave radiation). The strong relationship between topographic and climatic factors has been widely recognized as a key driver for modeling and forecasting LULC dynamics, particularly in mountainous regions where elevation-related gradients strongly influence vegetation patterns [62]. In this sense, elevation is one of the most determining variables in the spatial distribution of LULC, as it regulates factors such as temperature, humidity, and radiation, thus conditioning the type of vegetation and the development of certain land uses [63]. Similarly, land surface temperature acts as a crucial indicator for identifying coverages such as wetlands, agricultural areas, or grassland zones, as its variability can reflect changes in soil moisture and vegetation cover [64].

3.2.2. Validation of Future Scenario Models

The ANN–CA–Markov modeling framework applied in this study integrates artificial intelligence and spatial simulation techniques to produce accurate projections of land use and land cover (LULC) change. The modeling process began with the preparation and standardization of historical LULC maps (2003, 2013, and 2023), along with 11 topographic and climatic predictor variables, ensuring consistent spatial resolution across all inputs.
Using this dataset, a multilayer perceptron (MLP) artificial neural network was trained with 10,000 stratified sampling points. The ANN model was configured with a 1 × 1 neighborhood, a learning rate of 0.001, a momentum of 0.005, 12 hidden layers, and a maximum of 700 iterations. Both the calibration process and the adjustment of model parameters were carried out through iterative trial and error to ensure optimal performance. The output of this stage was a set of transition potential maps representing the likelihood of LULC change in each pixel based on environmental conditions.
To capture spatial dynamics, a cellular automata (CA) model was applied to allocate land cover changes according to the calculated transition probabilities, incorporating neighborhood effects and spatial patterns observed in the historical data. To simulate temporal dynamics, a Markov chain model was used to estimate class-to-class transition probabilities over time. This hybrid structure leverages the strengths of each component: the ANN captures complex nonlinear relationships among predictors, the CA simulates spatial diffusion processes, and the Markov chain governs long-term transition trends.
Model validation was conducted by simulating the 2023 land cover map using the historical data from 2003 and 2013. The MOLUSCE plugin in QGIS was used to compare the simulated map with the actual classified 2023 map, resulting in a prediction accuracy of 77% and a general Kappa coefficient of 0.81—indicating near-perfect agreement and high predictive reliability. Following validation, the model was used to simulate future LULC scenarios for the years 2033, 2043, and 2053, providing a robust basis for long-term territorial and environmental planning in the Locumba basin.

3.2.3. Assessment of Historical and Future Changes

The results of the projected LULC change prediction for the next 30 years show that for the three representative zones A, B, and C, each of them presents a common evident change in the increase of the area covered by grasslands, while the other land types maintain constant dimensions, with slight increases in their areas for wetlands and agriculture (Figure 8). Thus, it can be observed that a grassland area increase is projected from 220.31 km2 in 2023 to about 257.65 km2 in 2033, 294.8 km2 in 2043, and 331.62 km2 in 2053. For wetlands, there is a more modest increase in their areas, going from 10.96 km2 to 10.94 km2, 11.2 km2, and 11.32 km2 for the years 2033, 2043, and 2053, respectively. Another coverage projected to increase in the next 30 years is the areas occupied by agriculture, with a total of 37.29 km2 in 2023, projected to 37.73 km2, 38.15 km2, and 38.59 km2 for the years 2033, 2043, and 2053.
The gain and loss graph for the projected years 2023–2053 shows a continuous trend with the current historical time series (Figure 9). For example, for the first predicted period of 2023–2033, it is projected that agriculture and grasslands will increase in area by 1.2% and 16.9%, respectively, while wetlands, lagoons, and glaciers will experience a decrease in their areas by −0.2%, −3.0%, and −13.4%, respectively. For the second period of 2033–2043, the increase in lagoons, agriculture, wetlands, and grasslands will continue by 0.2%, 1.1%, 2.3%, and 14.4% of their initial areas, respectively, while glaciers will continue their decline in total areas with a negative change of −6.7%. Finally, for the third predicted period of 2043–2053, there will be an increase in the area covered by lagoons, agriculture, wetlands, and grasslands by 0.9%, 1.1%, 1.1%, and 12.5%, respectively, while only glaciers will continue to reduce their dimensions with a decrease of −1.9%.
This analysis clearly shows us how the covers that play a greater role within the Locumba basin are grasslands and glaciers. Grasslands will continue their aggressive increase in area, which is related to water loss in the glaciers. Hence, the projected land use and land cover changes for the next 30 years align with historical trends; this consistency underscores the reliability of the predictive model and its applicability for future land management planning. Moreover, incorporating future scenario analysis alongside historical trends provides a broader and more strategic perspective, allowing not only the identification of past dynamics but also the anticipation of potential future landscape transformations. Such an integrated approach is essential for developing proactive conservation and management strategies in fragile high Andean ecosystems like the Locumba basin.
Additionally, Figure 10 provides a comprehensive visualization of the historical and projected surface areas of the main LULC classes across the study period from 1984 to 2053. The graphical representation offers an integrated view of both past dynamics and future projections, reinforcing the spatial and temporal patterns discussed earlier and serving as a final synthesis of the major landscape transformations expected in the Locumba basin. This summary highlights the clear trends identified throughout the analysis, emphasizing the continuous expansion of grasslands and agricultural areas, alongside the steady reduction of glacial and lagoon surfaces.
The results provide valuable insights for territorial planning and environmental management in high Andean regions. The observed patterns of glacier retreat, wetland stabilization, and aggressive grassland expansion emphasize the urgent need to reassess current land use policies. For instance, areas projected to undergo sustained vegetation growth could be prioritized for pasture restoration or conservation zoning to preserve ecological resilience. Similarly, zones identified as vulnerable to water loss—especially those downstream of shrinking glaciers and disappearing lagoons—may require integrated water resource management plans to mitigate future scarcity. Local governments and regional planning authorities can use the spatial projections from this study to guide infrastructure development, agricultural zoning, and the protection of strategic water recharge areas.
Moreover, this research establishes a methodological framework that can be extended to other basins facing similar environmental pressures. The combination of time series analysis with ANN–CA modeling not only enables more accurate forecasting of LULC dynamics but also highlights the influence of topographic and climatic variables in shaping future landscape scenarios. As such, the study serves as a reference for future interdisciplinary research that seeks to model socio-ecological interactions, assess ecosystem service trade-offs, or quantify climate change impacts in mountainous ecosystems.

4. Discussion

The Locumba Basin has undergone significant land use and land cover (LULC) changes over the past four decades, with a pronounced retreat of glaciers and a marked expansion of vegetated areas, particularly grasslands and wetlands. These transformations align with broader trends observed in high Andean ecosystems, where rising temperatures and altered precipitation patterns are driving rapid ecological transitions [3,4].
The sharp decline in glacial coverage has exposed new terrain, which appears to be rapidly occupied by vegetation cover. This process is especially notable in the upper reaches of the basin, where grasslands have expanded by more than 280%, surpassing 220.31 km2 by 2023. This aggressive growth trend is supported by the projections for the next 30 years, which show a continued increase, potentially reaching over 331.62 km2 by 2053. Wetlands and agricultural areas also display slight gains, although to a much lesser extent.
This revegetation process, documented in other Andean zones, is often facilitated by increased meltwater availability and soil stabilization mechanisms associated with emergent vegetation [3,4]. In this study, the aggressive expansion of grassland areas, especially above 3700 m.a.s.l, suggests a gradual ecological transition catalyzed by warming temperatures and reduced snow coverage, two variables strongly correlated with elevation [9,10]. Nevertheless, the expansion of green cover should be interpreted with caution. While it may suggest increased productivity or ecological resilience, it could also represent a transitory phase. As glacier volumes continue to diminish, the long-term availability of meltwater is likely to decline, potentially threatening the stability of wetlands and high-altitude pastures.
In this context, the results highlight that the major driving factors that control the spatial and temporal patterns observed in LULC within the Locumba basin are likely driven by a complex interplay of climatic, topographic, and socioeconomic factors. In this sense, the regional climate variability—particularly fluctuations in temperature and rainfall across seasons—strongly governs water availability, affecting glacier persistence and wetland dynamics, which in turn modulate vegetation expansion and contraction [65]. On the other hand, the topography in the basin dictates microclimatic gradients: steep slopes and variations in elevation create distinct niches where vegetation and snow cover respond differently to the same climatic forcing [66]. Finally, water-driven land use changes (e.g., irrigation-linked pasture expansion and related soil compaction) further feedback on hydrological regimes, exacerbating LULC transitions in semi-arid settings [67]. These factors align with the observed retreat of glaciers, wetland shrinkage, and grassland encroachment in our study area, suggesting that integrating climatic, topographic, and socio-economic variables is essential for understanding—and forecasting—the future landscape trajectory in the Locumba basin.
The hybrid artificial neural network–cellular automata predictive modeling suggests that current LULC trends will persist over the next three decades. The forecasts for 2033, 2043, and 2053 indicate a continued expansion of grasslands, alongside more modest increases in wetlands and agriculture. Glaciers and lagoons, in contrast, are projected to stabilize at minimal surface areas, reinforcing the long-term trend of cryosphere loss [5,36]. These projections are consistent with historical time series analyses and reinforce the model’s credibility, as supported by the validation phase, where the Kappa coefficient exceeded 0.81, indicating near-perfect agreement. Similar findings have been reported in other Andean regions using CA-Markov and ANN models for landscape forecasting [36,37,38]. Although the model provides a valuable approximation of future trends, its predictive capacity is inherently limited by uncertainties related to climate extremes, land use decisions, and socio-environmental factors not included in this study. Continuous monitoring and periodic updates to the input datasets are essential to improve long-term forecasting.
In a similar sense, while this study chooses to use the ANN–CA–Markov approach for its robust predictive capacity, several alternative land-use change models have been widely applied in similar contexts. The CLUMondo approach predicts future land changes by assigning them to areas with the greatest suitability for a particular land system, based on the spatial relationships between that system and relevant location-specific factors [68]. FLUS (Future Land Use Simulation) advances CA–Markov models by embedding a top-down demand module, adaptive inertia, and stochastic competition mechanisms, allowing accurate spatial dynamics prediction even with non-linear land change processes, as demonstrated in studies from China [69]. Newer models, such as PLUS, build upon FLUS by integrating rule-mining and multi-type random patch seed generation to better capture patch-level growth patterns under complex human–ecological interactions [70]. Nevertheless, comparing the methodologies mentioned to the ANN–CA–Markov hybrid approach, this one leverages the non-linear mapping capabilities of artificial neural networks to calibrate transition potentials, coupled with Markov chain projection and spatially explicit cellular automata [71]. This synergistic framework enhances model realism by simultaneously accounting for historical trends, driving factor complexity (e.g., topography, hydrology), and spatial context. As a result, it delivers superior predictive performance in fragmented, high-relief landscapes like the high Andean Locumba basin.
In summary, the sustained transformations observed in this study highlight the need for integrated land management strategies that account for the fragility of Andean ecosystems. Planning measures should consider the ongoing loss of glacial reserves, the shifting distribution of water resources, and the implications for agriculture, biodiversity, and local livelihoods. For future studies, integrating dynamic socioeconomic scenarios, stakeholder input, and climate model projections could improve the realism and policy utility of LULC simulations. This would allow for exploring trade-offs between development and conservation under plausible future trajectories, helping decision-makers plan for long-term sustainability in fragile mountain environments.

5. Conclusions

This study demonstrated substantial transformations in land use and land cover (LULC) in the Locumba Basin over the past four decades, characterized by a significant reduction in glacial and lagoon areas and a pronounced expansion of grasslands and wetlands. The predictive modeling results indicate that these trends will likely continue over the next 30 years, with grasslands exhibiting the most aggressive growth.
Between 1984 and 2023, wetland areas expanded significantly, from 2.93 km2 to 10.96 km2, while glaciers decreased from 19.74 km2 to 3.95 km2. Meanwhile, grasslands have expanded considerably, from 57.87 km2 in 1984 to 220.31 km2 in 2023, illustrating the strong influence of climatic variability on high Andean ecosystems.
Predictive modeling, employing a hybrid artificial neural network–cellular automata (ANN–CA) approach, suggests that these trends will likely persist over the next 30 years. Grassland areas are projected to increase to over 331.62 km2 by 2053, while glacier coverage is expected to stabilize at minimal levels. Wetland areas are anticipated to experience slight growth, whereas agricultural areas are projected to remain relatively stable.
The strong performance of the classification and predictive models supports the reliability of these findings and highlights the importance of topographic and climatic variables in driving landscape changes in high-altitude ecosystems. Given the projected persistence of glacial retreat and vegetation expansion, future land and water resource management strategies must consider the changing dynamics of these fragile high-altitude ecosystems. Long-term monitoring, adaptive planning, and proactive conservation initiatives will be essential to mitigate the risks associated with declining water availability and ecosystem stability. The insights gained from this study provide valuable input for developing sustainable conservation strategies and for anticipating future environmental scenarios in vulnerable mountainous regions.
Likewise, future research should incorporate climate model data, stakeholder input, and field validation to refine simulation accuracy and broaden the relevance of projections for planning and policymaking.

Author Contributions

Conceptualization, G.H., E.P.-V., V.P., P.F.-L. and F.C.-O.; software, G.H., V.P., J.E.-M., F.C.-O. and E.I.-B.; data curation, G.H., E.P.-V., V.P., F.C.-O., B.V.-B. and L.R.-F.; validation, E.P.-V., P.F.-L., F.C.-O., E.I.-B. and G.H.; formal analysis, V.P., B.V.-B., G.H., K.A.-C., P.F.-L., E.I.-B. and E.P.-V.; writing—original draft preparation, G.H., E.P.-V., V.P., F.C.-O., L.R.-F. and E.I.-B.; writing—review and editing, G.H., E.P.-V., V.P., P.F.-L., J.E.-M., F.C.-O., K.A.-C., B.V.-B., L.R.-F. and E.I.-B.; project administration, P.F.-L. and E.P.-V. All authors have read and agreed to the published version of the manuscript.

Funding

This work was financed by funds from the mining royalties, IGIN, VIIN of the UNJBG, within the framework of the research project “Water availability and conservation status of water-dependent ecosystems in the upper basin of the Locumba River”, Resolution of the University Council N° 7747-2020-UN/JBG.

Data Availability Statement

The data that support the findings of this study are available upon request from the corresponding author.

Acknowledgments

To the Jorge Basadre Grohmann National University and especially to the H2O’UNJBG Research Group, Water Research Group.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Tsai, Y.H.; Stow, D.; Chen, H.L.; Lewison, R.; An, L.; Shi, L. Mapping Vegetation and Land Use Types in Fanjingshan National Nature Reserve Using Google Earth Engine. Remote Sens. 2018, 10, 927. [Google Scholar] [CrossRef]
  2. Yaranga, R.; Custodio, M.; Chanamé, F.; Pantoja, R. Floristic Diversity in Grasslands According to Plant Formation in the Shullcas River Sub-Basin, Junin, Peru. Sci. Agropecu. 2018, 9, 511–517. [Google Scholar] [CrossRef]
  3. Pizarro, S.E.; Pricope, N.G.; Vargas-Machuca, D.; Huanca, O.; Ñaupari, J. Mapping Land Cover Types for Highland Andean Ecosystems in Peru Using Google Earth Engine. Remote Sens. 2022, 14, 1562. [Google Scholar] [CrossRef]
  4. Buytaert, W.; Celleri, R.; Willems, P.; De Bièvre, B.; Wyseure, G. Spatial and Temporal Rainfall Variability in Mountainous Areas: A Case Study from the South Ecuadorian Andes. J. Hydrol. 2006, 329, 413–421. [Google Scholar] [CrossRef]
  5. Madrigal-Martínez, S.; Miralles i García, J.L. Land-Change Dynamics and Ecosystem Service Trends across the Central High-Andean Puna. Sci. Rep. 2019, 9, 9688. [Google Scholar] [CrossRef]
  6. Duque, A.; Peña, M.A.; Cuesta, F.; González-Caro, S.; Kennedy, P.; Phillips, O.L.; Calderón-Loor, M.; Blundo, C.; Carilla, J.; Cayola, L.; et al. Mature Andean Forests as Globally Important Carbon Sinks and Future Carbon Refuges. Nat. Commun. 2021, 12, 2138. [Google Scholar] [CrossRef]
  7. Molina, A.; Vanacker, V.; Balthazar, V.; Mora, D.; Govers, G. Complex Land Cover Change, Water and Sediment Yield in a Degraded Andean Environment. J. Hydrol. 2012, 472–473, 25–35. [Google Scholar] [CrossRef]
  8. Mora, D.E.; Willems, P. Decadal Oscillations in Rainfall and Air Temperature in the Paute River Basin—Southern Andes of Ecuador. Theor. Appl. Climatol. 2012, 108, 267–282. [Google Scholar] [CrossRef]
  9. Laraque, A.; Ronchail, J.; Cochonneau, G.; Pombosa, R.; Guyot, J.L. Heterogeneous Distribution of Rainfall and Discharge Regimes in the Ecuadorian Amazon Basin. J. Hydrometeorol. 2007, 8, 1364–1381. [Google Scholar] [CrossRef]
  10. Espinoza Villar, J.C.; Ronchail, J.; Guyot, J.L.; Cochonneau, G.; Naziano, F.; Lavado, W.; De Oliveira, E.; Pombosa, R.; Vauchel, P. Spatio-temporal Rainfall Variability in the Amazon Basin Countries (Brazil, Peru, Bolivia, Colombia, and Ecuador). Int. J. Climatol. 2009, 29, 1574–1594. [Google Scholar] [CrossRef]
  11. Pino-Vargas, E.; Chávarri-Velarde, E. Evidencias de Cambio Climático En La Región Hiperárida de La Costa Sur de Perú, Cabecera Del Desierto de Atacama. Tecnol. Cienc. Agua 2022, 13, 333–376. [Google Scholar] [CrossRef]
  12. Rind, D.; Goldberg, R.; Ruedy, R. Change in Climate Variability in the 21st Century. Clim. Change 1989, 14, 5–37. [Google Scholar] [CrossRef]
  13. Kazemzadeh, M.; Hashemi, H.; Jamali, S.; Uvo, C.B.; Berndtsson, R.; Huffman, G.J. Linear and Nonlinear Trend Analyzes in Global Satellite-Based Precipitation, 1998–2017. Earth’s Future 2021, 9, e2020EF001835. [Google Scholar] [CrossRef]
  14. Pino-Vargas, E.; Chávarri-Velarde, E.; Ingol-Blanco, E.; Mejía, F.; Cruz, A.; Vera, A. Impacts of Climate Change and Variability on Precipitation and Maximum Flows in Devil’s Creek, Tacna, Peru. Hydrology 2022, 9, 10. [Google Scholar] [CrossRef]
  15. Machaca-Pillaca, R.; Pino-Vargas, E.; Ramos-Fernández, L.; Quille-Mamani, J.; Torres-Rua, A. Estimación de La Evapotranspiración Con Fines de Riego En Tiempo Real de Un Olivar a Partir de Imágenes de Un Drone En Zonas Áridas, Caso La Yarada, Tacna, Perú. Idesia 2022, 40, 55–65. [Google Scholar] [CrossRef]
  16. Pino Vargas, E.M.; Huayna, G. Spatial and Temporal Evolution of Olive Cultivation Due to Pest Attack, Using Remote Sensing and Satellite Image Processing. Sci. Agropecu. 2022, 13, 149–157. [Google Scholar] [CrossRef]
  17. Pino-Vargas, E. Conflictos Por El Uso Del Agua En Una Región Árida: Caso Tacna, Perú. Diálogo Andin. 2021, 405–415. [Google Scholar] [CrossRef]
  18. Pino-Vargas, E.; Ramos, L.; Mejía, J.; Chávarri, E.; Ascensios, D. Medidas de Mitigación Para El Acuífero Costero La Yarada, Un Sistema Sobreexplotado En Zonas Áridas. Idesia 2020, 38, 21–31. [Google Scholar] [CrossRef]
  19. González-Domínguez, J.; Mora, A.; Chucuya, S.; Pino-Vargas, E.; Torres-Martínez, J.A.; Dueñas-Moreno, J.; Ramos-Fernández, L.; Kumar, M.; Mahlknecht, J. Hydraulic Recharge and Element Dynamics during Salinization in an Overexploited Coastal Aquifer of the World’s Driest Zone: Atacama Desert. Sci. Total Environ. 2024, 954, 176204. [Google Scholar] [CrossRef]
  20. Pocco, V.; Mendoza, A.; Chucuya, S.; Franco-León, P.; Huayna, G.; Ingol-Blanco, E.; Pino-Vargas, E. Assessment of Potential Aquifer Recharge Zones in the Locumba Basin, Arid Region of the Atacama Desert Using Integration of Two MCDM Methods: Fuzzy AHP and TOPSIS. Water 2024, 16, 2643. [Google Scholar] [CrossRef]
  21. Ayana, B.; Senbeta, F.; Seyoum, A. Analyses of LULC Dynamics in a Socio-Ecological System of the Bale Mountains Eco Region of Southeast Ethiopia. Environ. Monit. Assess. 2024, 196, 644. [Google Scholar] [CrossRef] [PubMed]
  22. Yang, Y.; Li, Z.; Quddoos, A.; Aslam, R.W.; Naz, I.; Khalid, M.B.; Afzal, Z.; Liaquat, M.A.; Abdullah-Al-Wadud, M. Decadal Dynamics of Rangeland Cover Using Remote Sensing and Machine Learning Approach. Rangel. Ecol. Manag. 2025, 100, 1–13. [Google Scholar] [CrossRef]
  23. Shadmehri Toosi, A.; Batelaan, O.; Shanafield, M.; Guan, H. Land Use-Land Cover and Hydrological Modeling: A Review. WIREs Water 2025, 12, e70013. [Google Scholar] [CrossRef]
  24. Gharahbagh, A.A.; Hajihashemi, V.; Machado, J.J.M.; Tavares, J.M.R.S. Land Cover Classification Model Using Multispectral Satellite Images Based on a Deep Learning Synergistic Semantic Segmentation Network. Sensors 2025, 25, 1988. [Google Scholar] [CrossRef]
  25. Cuypers, S.; Nascetti, A.; Vergauwen, M. Land Use and Land Cover Mapping with VHR and Multi-Temporal Sentinel-2 Imagery. Remote Sens. 2023, 15, 2501. [Google Scholar] [CrossRef]
  26. Stromann, O.; Nascetti, A.; Yousif, O.; Ban, Y. Dimensionality Reduction and Feature Selection for Object-Based Land Cover Classification Based on Sentinel-1 and Sentinel-2 Time Series Using Google Earth Engine. Remote Sens. 2019, 12, 76. [Google Scholar] [CrossRef]
  27. Belay, H.; Melesse, A.M.; Tegegne, G. Scenario-Based Land Use and Land Cover Change Detection and Prediction Using the Cellular Automata–Markov Model in the Gumara Watershed, Upper Blue Nile Basin, Ethiopia. Land 2024, 13, 396. [Google Scholar] [CrossRef]
  28. Tiao, G.C. An Introduction to Multiple Time Series Analysis. Med. Care 1993, 31, YS71–YS74. [Google Scholar]
  29. Souto, H.G. Time Series Forecasting Models for S&P 500 Financial Turbulence. J. Math. Financ. 2023, 13, 112–129. [Google Scholar]
  30. Yang, X.; Liu, B. Uncertain Time Series Analysis with Imprecise Observations. Fuzzy Optim. Decis. Mak. 2019, 18, 263–278. [Google Scholar] [CrossRef]
  31. Bandt, C. Ordinal Time Series Analysis. Ecol. Modell. 2005, 182, 229–238. [Google Scholar] [CrossRef]
  32. Gardner, E.S.; Mckenzie, E. Forecasting Trends in Time Series. Manage. Sci. 1985, 31, 1237–1246. [Google Scholar] [CrossRef]
  33. Pinheiro, A.C.; Rodrigues, P.C. Hierarchical Time Series Forecasting of Fire Spots in Brazil: A Comprehensive Approach. Stats 2024, 7, 647–670. [Google Scholar] [CrossRef]
  34. Eastman, J.R.; Toledano, J. A Short Presentation of CA_MARKOV. In Geomatic Approaches for Modeling Land Change Scenarios; Springer: Cham, Switzerland, 2018; pp. 481–484. [Google Scholar]
  35. Huayna, G.; Pino-Vargas, E.; Espinoza-Molina, J.; Cruz-Rodríguez, C.; Cabrera-Olivera, F.; Ramos-Fernández, L.; Vera-Barrios, B.; Acosta-Caipa, K.; Ingol-Blanco, E. Glacier, Wetland, and Lagoon Dynamics in the Barroso Mountain Range, Atacama Desert: Past Trends and Future Projections Using CA-Markov. Hydrology 2025, 12, 64. [Google Scholar] [CrossRef]
  36. Farhan, M.; Wu, T.; Anwar, S.; Yang, J.; Naqvi, S.A.A.; Soufan, W.; Tariq, A. Predicting Land Use Land Cover Dynamics and Land Surface Temperature Changes Using CA-Markov-Chain Models in Islamabad, Pakistan (1992–2042). IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2024, 17, 16255–16271. [Google Scholar] [CrossRef]
  37. Yulianto, F.; Maulana, T.; Khomarudin, M.R. Analysis of the Dynamics of Land Use Change and Its Prediction Based on the Integration of Remotely Sensed Data and CA-Markov Model, in the Upstream Citarum Watershed, West Java, Indonesia. Int. J. Digit. Earth 2019, 12, 1151–1176. [Google Scholar] [CrossRef]
  38. Asif, M.; Kazmi, J.H.; Tariq, A.; Zhao, N.; Guluzade, R.; Soufan, W.; Almutairi, K.F.; El Sabagh, A.; Aslam, M. Modelling of Land Use and Land Cover Changes and Prediction Using CA-Markov and Random Forest. Geocarto Int. 2023, 38, 2210532. [Google Scholar] [CrossRef]
  39. Fathizad, H.; Rostami, N.; Faramarzi, M. Detection and Prediction of Land Cover Changes Using Markov Chain Model in Semi-Arid Rangeland in Western Iran. Environ. Monit. Assess. 2015, 187, 629. [Google Scholar] [CrossRef]
  40. Ghosh, P.; Mukhopadhyay, A.; Chanda, A.; Mondal, P.; Akhand, A.; Mukherjee, S.; Nayak, S.K.; Ghosh, S.; Mitra, D.; Ghosh, T.; et al. Application of Cellular Automata and Markov-Chain Model in Geospatial Environmental Modeling- A Review. Remote Sens. Appl. Soc. Environ. 2017, 5, 64–77. [Google Scholar] [CrossRef]
  41. Feng, Y.; Liu, Y.; Batty, M. Modeling Urban Growth with GIS Based Cellular Automata and Least Squares SVM Rules: A Case Study in Qingpu–Songjiang Area of Shanghai, China. Stoch. Environ. Res. Risk Assess. 2016, 30, 1387–1400. [Google Scholar] [CrossRef]
  42. Koltai, P.; Wu, H.; Noé, F.; Schütte, C. Optimal Data-Driven Estimation of Generalized Markov State Models for Non-Equilibrium Dynamics. Computation 2018, 6, 22. [Google Scholar] [CrossRef]
  43. Wu, H.; Noé, F. Variational Approach for Learning Markov Processes from Time Series Data. J. Nonlinear Sci. 2020, 30, 23–66. [Google Scholar] [CrossRef]
  44. Schöttler, S.; Yang, Y.; Pfister, H.; Bach, B. Visualizing and Interacting with Geospatial Networks: A Survey and Design Space. Comput. Graph. Forum 2021, 40, 5–33. [Google Scholar] [CrossRef]
  45. Gu, Y.; Kraak, M.-J.; Engelhardt, Y.; Mocnik, F.-B. A Classification Scheme for Static Origin–Destination Data Visualizations. Int. J. Geogr. Inf. Sci. 2023, 37, 1970–1997. [Google Scholar] [CrossRef]
  46. Giabbanelli, P.J.; Baniukiewicz, M. Visual Analytics to Identify Temporal Patterns and Variability in Simulations from Cellular Automata. ACM Trans. Model. Comput. Simul. 2019, 29, 1–26. [Google Scholar] [CrossRef]
  47. Giordano, A.; de Rango, A.; Spataro, D.; D’Ambrosio, D.; Mastroianni, C.; Folino, G.; Spataro, W. Parallel Execution of Cellular Automata through Space Partitioning: The Landslide Simulation Sciddicas3-Hex Case Study. In Proceedings of the 2017 25th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP), St. Petersburg, Russia, 6–8 March 2017; pp. 505–510. [Google Scholar]
  48. Folino, G.; Giordano, A.; Mastroianni, C. Scalable Asynchronous Execution of Cellular Automata. AIP Conf. Proc. 2016, 1776, 080006. [Google Scholar]
  49. Rodríguez Eraso, N.; Armenteras-Pascual, D.; Alumbreros, J.R. Land Use and Land Cover Change in the Colombian Andes: Dynamics and Future Scenarios. J. Land Use Sci. 2013, 8, 154–174. [Google Scholar] [CrossRef]
  50. Castillón, F.; Rau, P.; Bourrel, L.; Frappart, F. Dynamics and Patterns of Land Cover Change in the Piura River Basin (Peruvian Pacific Slope and Coast) in the Last Two Decades. Front. Remote Sens. 2025, 6, 1529044. [Google Scholar] [CrossRef]
  51. Pereira, C.O.; Escanilla-Minchel, R.; González, A.C.; Alcayaga, H.; Aguayo, M.; Arias, M.A.; Flores, A.N. Assessment of Future Land Use/Land Cover Scenarios on the Hydrology of a Coastal Basin in South-Central Chile. Sustainability 2022, 14, 16363. [Google Scholar] [CrossRef]
  52. Dai, Z.; Chang, S.; Zhu, Z.; Duan, J.; Jiang, T.; Wu, W.; Feng, Y.; Yang, G.; Wang, X. Assessment and Multiscenario Simulation of Land Use and Ecosystem Services Interactions in Inner Mongolia. Land Degrad. Dev. 2024, 35, 5611–5625. [Google Scholar] [CrossRef]
  53. Ermida, S.L.; Soares, P.; Mantas, V.; Göttsche, F.-M.; Trigo, I.F. Google Earth Engine Open-Source Code for Land Surface Temperature Estimation from the Landsat Series. Remote Sens. 2020, 12, 1471. [Google Scholar] [CrossRef]
  54. Romero Rodríguez, J.; Aguilar-Ávila, J.; Santoyo Cortés, V.H.; Diakite, L. Transiciones Del Cambio de Uso de Suelo En El Estado de Puebla (1980–2016), México. Investig. Geográficas 2022. [Google Scholar] [CrossRef]
  55. González Soto, F.; Ullón, D.; Yaguachi Alarcón, A.L.; Ramos Alcivar, J.A.; Montenegro Benalcázar, V.E.; Loján Córdova, J.I. Análisis Multitemporal de Cambios de Uso Del Suelo En La Isla Santa Cruz, Archipiélago de Las Galápagos, Periodo 1991–2023. Rev. Cienc. Tecnol. 2024, 17, 1–9. [Google Scholar] [CrossRef]
  56. Scott, B.; Baldwin, A.H.; Yarwood, S.A. Quantification of Potential Methane Emissions Associated with Organic Matter Amendments Following Oxic-Soil Inundation. Biogeosciences 2022, 19, 1151–1164. [Google Scholar] [CrossRef]
  57. Robison, A.L.; Deluigi, N.; Rolland, C.; Manetti, N.; Battin, T. Glacier Loss and Vegetation Expansion Alter Organic and Inorganic Carbon Dynamics in High-Mountain Streams. Biogeosciences 2023, 20, 2301–2316. [Google Scholar] [CrossRef]
  58. Emmer, A.; Vilímek, V. New Method for Assessing the Susceptibility of Glacial Lakes to Outburst Floods in the Cordillera Blanca, Peru. Hydrol. Earth Syst. Sci. 2014, 18, 3461–3479. [Google Scholar] [CrossRef]
  59. Zhang, M.; Chen, F.; Guo, H.; Yi, L.; Zeng, J.; Li, B. Glacial Lake Area Changes in High Mountain Asia during 1990–2020 Using Satellite Remote Sensing. Research 2022, 2022, 9821275. [Google Scholar] [CrossRef]
  60. Gao, J.-G.; Zhang, Y.-L.; Liu, L.-S.; Wang, Z.-F. Climate Change as the Major Driver of Alpine Grasslands Expansion and Contraction: A Case Study in the Mt. Qomolangma (Everest) National Nature Preserve, Southern Tibetan Plateau. Quat. Int. 2014, 336, 108–116. [Google Scholar] [CrossRef]
  61. Zhu, K.; Song, Y.; Lesage, J.C.; Luong, J.C.; Bartolome, J.W.; Chiariello, N.R.; Dudney, J.; Field, C.B.; Hallett, L.M.; Hammond, M.; et al. Rapid Shifts in Grassland Communities Driven by Climate Change. Nat. Ecol. Evol. 2024, 8, 2252–2264. [Google Scholar] [CrossRef]
  62. Li, P.; Feng, Z. Extent and Area of Swidden in Montane Mainland Southeast Asia: Estimation by Multi-Step Thresholds with Landsat-8 OLI Data. Remote Sens. 2016, 8, 44. [Google Scholar] [CrossRef]
  63. Körner, C. The Use of ‘Altitude’ in Ecological Research. Trends Ecol. Evol. 2007, 22, 569–574. [Google Scholar] [CrossRef] [PubMed]
  64. Weng, Q. Thermal Infrared Remote Sensing for Urban Climate and Environmental Studies: Methods, Applications, and Trends. ISPRS J. Photogramm. Remote Sens. 2009, 64, 335–344. [Google Scholar] [CrossRef]
  65. Carilla, J.; Aráoz, E.; Foguet, J.; Casagranda, E.; Halloy, S.; Grau, A. Hydroclimate and Vegetation Variability of High Andean Ecosystems. Front. Plant Sci. 2023, 13, 1067096. [Google Scholar] [CrossRef] [PubMed]
  66. Deng, C.; Ma, X.; Xie, M.; Bai, H. Effect of Altitude and Topography on Vegetation Phenological Changes in the Niubeiliang Nature Reserve of Qinling Mountains, China. Forests 2022, 13, 1229. [Google Scholar] [CrossRef]
  67. Carbajal, M.; Ramírez, D.A.; Turin, C.; Schaeffer, S.M.; Konkel, J.; Ninanya, J.; Rinza, J.; De Mendiburu, F.; Zorogastua, P.; Villaorduña, L.; et al. From Rangelands to Cropland, Land-Use Change and Its Impact on Soil Organic Carbon Variables in a Peruvian Andean Highlands: A Machine Learning Modeling Approach. Ecosystems 2024, 27, 899–917. [Google Scholar] [CrossRef]
  68. Malek, Ž.; Verburg, P.H.; Geijzendorffer, I.R.; Bondeau, A.; Cramer, W. Global Change Effects on Land Management in the Mediterranean Region. Glob. Environ. Change 2018, 50, 238–254. [Google Scholar] [CrossRef]
  69. Luan, C.; Liu, R. A Comparative Study of Various Land Use and Land Cover Change Models to Predict Ecosystem Service Value. Int. J. Environ. Res. Public Health 2022, 19, 16484. [Google Scholar] [CrossRef]
  70. Liu, J.; Liu, B.; Wu, L.; Miao, H.; Liu, J.; Jiang, K.; Ding, H.; Gao, W.; Liu, T. Prediction of Land Use for the next 30 Years Using the PLUS Model’s Multi-Scenario Simulation in Guizhou Province, China. Sci. Rep. 2024, 14, 13143. [Google Scholar] [CrossRef]
  71. Gharaibeh, A.; Shaamala, A.; Obeidat, R.; Al-Kofahi, S. Improving Land-Use Change Modeling by Integrating ANN with Cellular Automata-Markov Chain Model. Heliyon 2020, 6, e05092. [Google Scholar] [CrossRef]
Land 14 01442 g001
Figure 1. (a) Localization of the Locumba Basin. (b) Climatological Data of the Candarave Meteorological Station.
Figure 1. (a) Localization of the Locumba Basin. (b) Climatological Data of the Candarave Meteorological Station.
Land 14 01442 g001
Land 14 01442 g002
Figure 2. Methodology flowchart.
Figure 2. Methodology flowchart.
Land 14 01442 g002
Land 14 01442 g003
Figure 3. Temporal trends and predictions of LULC for 1984–2053.
Figure 3. Temporal trends and predictions of LULC for 1984–2053.
Land 14 01442 g003
Land 14 01442 g004
Figure 4. Time series maps of LULC changes (1984–2023).
Figure 4. Time series maps of LULC changes (1984–2023).
Land 14 01442 g004
Land 14 01442 g005
Figure 5. LULC loss and gain during the periods (a) 1984–1993, (b) 1993–2003, (c) 2003–2013, (d) 2013–2023, and (e) 1984–2023.
Figure 5. LULC loss and gain during the periods (a) 1984–1993, (b) 1993–2003, (c) 2003–2013, (d) 2013–2023, and (e) 1984–2023.
Land 14 01442 g005
Land 14 01442 g006
Figure 6. Correlation matrix of predictor variables.
Figure 6. Correlation matrix of predictor variables.
Land 14 01442 g006
Land 14 01442 g007
Figure 7. Visualization of predictor variables for the prediction model. (a) Elevation. (b) Slope. (c) Aspect. (d) Shape. (e) Curvature Horizontal. (f) Curvature Vertical. (g) Topographic Position Index. (h) Terrain Ruggedness Index. (i) Annual Precipitation. (j) Shortwave Radiation. (k) Land Surface Temperature.
Figure 7. Visualization of predictor variables for the prediction model. (a) Elevation. (b) Slope. (c) Aspect. (d) Shape. (e) Curvature Horizontal. (f) Curvature Vertical. (g) Topographic Position Index. (h) Terrain Ruggedness Index. (i) Annual Precipitation. (j) Shortwave Radiation. (k) Land Surface Temperature.
Land 14 01442 g007
Land 14 01442 g008
Figure 8. Time series maps of LULC changes (2023–2053).
Figure 8. Time series maps of LULC changes (2023–2053).
Land 14 01442 g008
Land 14 01442 g009
Figure 9. LULC loss and gain during the periods (a) 2023–2033, (b) 2033–2043, (c) 2043–2053, and (d) 2023–2053.
Figure 9. LULC loss and gain during the periods (a) 2023–2033, (b) 2033–2043, (c) 2043–2053, and (d) 2023–2053.
Land 14 01442 g009
Land 14 01442 g010
Figure 10. Surface area evolution of major land cover types from 1984 to 2053.
Figure 10. Surface area evolution of major land cover types from 1984 to 2053.
Land 14 01442 g010
Table 1. Downloaded scenes from Landsat collections.
Table 1. Downloaded scenes from Landsat collections.
DataYearProduct IdentifierSensing Time
(hh:mm:ss)		Cloud Cover
%		Patch/Row
Landsat 51984LANDSAT/LT05/C02/T1_TOA/LT05_002072_1984071514:10:211002/72
LANDSAT/LT05/C02/T1_TOA/LT05_003072_1984062014:16:124603/72
Landsat 51993LANDSAT/LT05/C02/T1_TOA/LT05_002072_1993060614:04:12002/72
LANDSAT/LT05/C02/T1_TOA/LT05_002072_1993070814:04:021
LANDSAT/LT05/C02/T1_TOA/LT05_003071_1993062914:09:51003/71
LANDSAT/LT05/C02/T1_TOA/LT05_003071_1993071514:09:494
Landsat 52003LANDSAT/LT05/C02/T1_TOA/LT05_002072_2003061814:17:31602/72
LANDSAT/LT05/C02/T1_TOA/LT05_002072_2003070414:17:520
LANDSAT/LT05/C02/T1_TOA/LT05_003071_2003071114:23:46103/71
LANDSAT/LT05/C02/T1_TOA/LT05_003071_2003072714:24:044
Landsat 82013LANDSAT/LC08/C02/T1_TOA/LC08_002072_2013062914:43:362.7202/72
LANDSAT/LC08/C02/T1_TOA/LC08_002072_2013071514:43:3718.09
LANDSAT/LC08/C02/T1_TOA/LC08_003071_2013062014:49:202.6503/71
LANDSAT/LC08/C02/T1_TOA/LC08_003071_2013070614:49:248.82
Landsat 82023LANDSAT/LC08/C02/T1_TOA/LC08_002072_2023060914:41:034.19
LANDSAT/LC08/C02/T1_TOA/LC08_002072_2023062514:41:105.5602/72
LANDSAT/LC08/C02/T1_TOA/LC08_002072_2023071114:41:205.13
LANDSAT/LC08/C02/T1_TOA/LC08_003071_2023053114:46:440.22
LANDSAT/LC08/C02/T1_TOA/LC08_003071_2023061614:46:530.2203/71
LANDSAT/LC08/C02/T1_TOA/LC08_003071_2023071814:47:070.26
Table 2. Geospatial data sources and thematic variables used.
Table 2. Geospatial data sources and thematic variables used.
DatasetSourceResolutionThematic Layers
Shuttle Radar Topography MissionGoogle Earth Engine (https://developers.google.com/earth-engine/datasets, accessed on 15 January 2020)30 mElevation, Slope, Aspect, Shape Index, Curvature Horizontal, Curvature Vertical, Topographic Position Index, Terrain Ruggedness Index
Landsat 8Google Earth Engine Código Fuente (https://code.earthengine.google.com/?accept_repo=users/sofiaermida/landsat_smw_lst, accessed on 15 January 2020)30 mLand Surface Temperature
TerraClimateGoogle Earth Engine (https://developers.google.com/earth-engine/datasets, accessed on 15 January 2025)4638.3 mAnnual precipitation, Shortwave Radiation
Table 3. Land cover index calculations.
Table 3. Land cover index calculations.
Land Cover TypeIndex UsedIndex FormulaThreshold AppliedAdditional Criteria
GlacierNDSI N D S I = G R E E N − S W I R G R E E N + S W I R NDSI > 0.4Empirical thresholds adjusted per sensor (NIR > 0.11, RED > 0.10); visual correction for debris-covered areas
LagoonNDWI N D W I = G R E E N − N I R G R E E N + N I R NDWI > 0.2Slope ≤ 20 degrees and DEM > 3800 m.a.s.l.
WetlandNDVI + NDII N D V I = N I R − R E D N I R + R E D
N D I I = N I R − S W I R N I R + S W I R 		NDVI > 0.43 0.02 < NDII < 0.76DEM > 3800 m.a.s.l. (SRTM); combined condition (NDVI ∧ NDII)
AgricultureNDVI + DEM N D V I = N I R − R E D N I R + R E D NDVI > 0.20DEM < 3800 m.a.s.l. (excludes high Andean ecosystems)
GrasslandNDVI + DEM N D V I = N I R − R E D N I R + R E D NDVI > 0.20DEM ≥ 3800 m.a.s.l.; excludes wetland areas (wetland mask)
Table 4. Transition probability matrix calculated using land use in 1984–2023.
Table 4. Transition probability matrix calculated using land use in 1984–2023.
PeriodCategoryBare SoilWetlandGlacierLagoonsGrasslandAgriculture
1984–1993Bare soil0.9800.0000.0000.0000.0160.004
Wetland0.0010.3200.0000.0000.6790.000
Glacier0.8020.0000.1980.0000.0000.000
Lagoons0.3230.0000.0000.6680.0090.000
Grassland0.1030.0060.0000.0000.8910.000
Agriculture0.2010.0000.0000.0000.0000.799
1993–2003Bare soil0.9710.0000.0050.0020.0190.002
Wetland0.0000.8950.0000.0000.1050.000
Glacier0.1520.0000.8480.0000.0000.000
Lagoons0.0470.0000.0000.8440.1080.000
Grassland0.1500.0570.0000.0130.7810.000
Agriculture0.4020.0000.0000.0000.0000.598
2003–2013Bare soil0.9380.0000.0030.0010.0490.009
Wetland0.0000.8380.0000.0000.1620.000
Glacier0.4000.0000.5990.0010.0000.000
Lagoons0.0290.0000.0000.9700.0010.000
Grassland0.0330.0500.0010.0260.8910.000
Agriculture0.1620.0000.0000.0000.0000.838
2013–2023Bare soil0.9620.0000.0000.0000.0320.006
Wetland0.0000.7830.0000.0000.2170.000
Glacier0.7140.0000.2860.0000.0000.000
Lagoons0.3950.0000.0000.5730.0310.000
Grassland0.0820.0120.0000.0000.9060.000
Agriculture0.2290.0000.0000.0000.0000.771
Table 5. Validation matrix of land cover mapping.
Table 5. Validation matrix of land cover mapping.
YearLULCBSWGLLGHATotalAccuracy
User		Omission
Error
2023Bare soil (BS)8400060900.930.07
Wetland (W)0590010600.980.02
Glacier (GL)0060000601.000.00
Lagoons (L)0005000501.000.00
Grassland/Herbaceous (GH)1000690700.990.01
Agriculture (A)1000653600.880.12
Total865960508253390
Producer’s Accuracy0.981.001.001.000.841.000.96 = Overall Accuracy
Omission Error0.020.000.000.000.160.000.95 = Kappa Coefficient
2013Bare soil (BS)8601030900.960.04
Wetland (W)0510090600.850.15
Glacier (GL)1059000600.980.02
Lagoons (L)0005000501.000.00
Grassland/Herbaceous (GH)5001640700.910.09
Agriculture (A)1000851600.850.15
Total935160518451390
Producer’s Accuracy0.921.000.980.980.761.000.93 = Overall Accuracy
Omission Error0.080.000.020.020.240.000.91 = Kappa Coefficient
2003Bare soil (BS)8200062900.910.09
Wetland (W)0580020600.970.03
Glacier (GL)2058000600.970.03
Lagoons (L)0005000501.000.00
Grassland/Herbaceous (GH)1201660700.940.06
Agriculture (A)0000159600.980.02
Total856058517561390
Producer’s Accuracy0.960.971.000.980.880.970.96 = Overall Accuracy
Omission Error0.040.030.000.020.120.030.95 = Kappa Coefficient
1993Bare soil (BS)8300070900.920.08
Wetland (W)0590010600.980.02
Glacier (GL)0060000601.000.00
Lagoons (L)0005000501.000.00
Grassland/Herbaceous (GH)3100660700.940.06
Agriculture (A)0000456600.930.07
Total866060507856390
Producer’s Accuracy0.970.981.001.000.851.000.96 = Overall Accuracy
Omission Error0.030.020.000.000.150.000.95 = Kappa Coefficient
1984Bare soil (BS)8500032900.940.06
Wetland (W)0560040600.930.07
Glacier (GL)0060000601.000.00
Lagoons (L)0005000501.000.00
Grassland/Herbaceous (GH)4100650700.930.07
Agriculture (A)0000456600.930.07
Total895760507658390
Producer’s Accuracy0.960.981.001.000.860.970.95 = Overall Accuracy
Omission Error0.040.020.000.000.140.030.94 = Kappa Coefficient
Table 6. Kappa coefficient values.
Table 6. Kappa coefficient values.
Kappa Hat ValueInterpretation of Agreement
0.81–1.00Near-perfect agreement
0.61–0.80Strong level of agreement
0.41–0.60Moderate agreement
0.21–0.40Acceptable but limited agreement
0.01–0.20Minimal agreement
<0.00Discordance or poor match
Table 7. Area cover changes from 1984–2053, expressed in km2.
Table 7. Area cover changes from 1984–2053, expressed in km2.
YearsWetlandVar.GlacierVar.LagVar.GrassVar.AgricVar.
19842.93 19.74 16.87 57.87 32.8
19853.25−0.3222.6−2.8619.12−2.25110.89−53.0248.48−15.68
19863.73−0.4815.826.7823.57−4.45105.295.631.7316.75
19876.7−2.971.6214.219.34.2756.5148.7822.529.21
19884.791.914.17−2.5516.672.6381.91−25.423.83−1.31
19895.31−0.5213.32−9.1516.69−0.0287.57−5.6633.77−9.94
19903.461.852.510.829.766.9357.1930.3823.5810.19
19913.390.076.88−4.3812.77−3.0190.49−33.329.29−5.71
19920.233.160.066.828.654.1253.536.9927.761.53
19931.29−1.063.94−3.8811.29−2.6497.89−44.3935.83−8.07
19943.53−2.248.47−4.5315.2−3.91122.86−24.9740.54−4.71
19954.73−1.22.346.138.716.49133.6−10.7439.321.22
19963.181.552.75−0.419.87−1.16101.7831.8227.2512.07
19973.96−0.7812.69−9.9416.55−6.68120.21−18.4334.31−7.06
19985.55−1.590.2512.449.077.48121.9−1.6925.189.13
19992.962.5925.23−24.9817.74−8.67106.5715.3336.26−11.08
20003.96−111.2513.9817.560.18106.59−0.0229.296.97
20013.490.4724.83−13.5821.38−3.82109.41−2.8236.56−7.27
200210.39−6.923.791.0420.70.68143.42−34.0139.55−2.99
20036.743.6512.211.5914.915.79111.0132.4125.5414.01
20044.242.56.016.1911.293.62108.452.5623.052.49
20053.910.332.453.5610.380.91105.153.336.09−13.04
20065.69−1.788.77−6.3218.8−8.42124.99−19.8431.085.01
20075.620.074.594.1815.743.06105.319.6926.194.89
20084.211.410.963.6310.525.22111.9−6.631.81−5.62
20094.090.125.45−4.4913.46−2.94104.227.6833.71−1.9
20103.930.160.355.19.344.1271.0333.1928.974.74
20114.58−0.6510.58−10.2311.57−2.23102.53−31.532.16−3.19
20128.3−3.7221.39−10.8119.44−7.87175.39−72.8645.12−12.96
201311.18−2.8812.029.3719.69−0.25182.85−7.4636.918.21
20148.163.020.0711.9511.488.21153.3629.4929.377.54
201511.53−3.379.49−9.4214.61−3.13210.84−57.4849.41−20.04
201610.161.372.27.299.894.72194.6616.1834.1415.27
201782.164.77−2.579.260.63177.3217.3434.19−0.05
20188.32−0.324.510.269.35−0.09119.857.5217.7616.43
20199.58−1.262.232.2810.79−1.44202.02−82.2230.43−12.67
202012.35−2.775.94−3.7118.35−7.56253.37−51.3546.86−16.43
202110.781.577.11−1.1714.363.99216.2737.126.8420.02
20228.72.081.565.5514.68−0.32159.7356.5426.590.25
202310.96−2.263.95−2.3911.453.23220.31−60.5837.29−10.7
Mean −0.20 0.40 0.13 −4.16 −0.11
203310.940.023.420.5311.120.33257.65−37.3437.73−0.44
204311.2−0.263.190.2311.14−0.02294.8−37.1538.15−0.42
205311.32−0.123.130.0611.24−0.1331.62−36.8238.59−0.44
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Huayna, G.; Pocco, V.; Pino-Vargas, E.; Franco-León, P.; Espinoza-Molina, J.; Cabrera-Olivera, F.; Vera-Barrios, B.; Acosta-Caipa, K.; Ramos-Fernández, L.; Ingol-Blanco, E. Historical Land Cover Dynamics and Projected Changes in the High Andean Zone of the Locumba Basin: A Predictive Approach Using Remote Sensing and Artificial Neural Network—Cellular Automata Model. Land 2025, 14, 1442. https://doi.org/10.3390/land14071442

AMA Style

Huayna G, Pocco V, Pino-Vargas E, Franco-León P, Espinoza-Molina J, Cabrera-Olivera F, Vera-Barrios B, Acosta-Caipa K, Ramos-Fernández L, Ingol-Blanco E. Historical Land Cover Dynamics and Projected Changes in the High Andean Zone of the Locumba Basin: A Predictive Approach Using Remote Sensing and Artificial Neural Network—Cellular Automata Model. Land. 2025; 14(7):1442. https://doi.org/10.3390/land14071442

Chicago/Turabian Style

Huayna, German, Victor Pocco, Edwin Pino-Vargas, Pablo Franco-León, Jorge Espinoza-Molina, Fredy Cabrera-Olivera, Bertha Vera-Barrios, Karina Acosta-Caipa, Lía Ramos-Fernández, and Eusebio Ingol-Blanco. 2025. "Historical Land Cover Dynamics and Projected Changes in the High Andean Zone of the Locumba Basin: A Predictive Approach Using Remote Sensing and Artificial Neural Network—Cellular Automata Model" Land 14, no. 7: 1442. https://doi.org/10.3390/land14071442

APA Style

Huayna, G., Pocco, V., Pino-Vargas, E., Franco-León, P., Espinoza-Molina, J., Cabrera-Olivera, F., Vera-Barrios, B., Acosta-Caipa, K., Ramos-Fernández, L., & Ingol-Blanco, E. (2025). Historical Land Cover Dynamics and Projected Changes in the High Andean Zone of the Locumba Basin: A Predictive Approach Using Remote Sensing and Artificial Neural Network—Cellular Automata Model. Land, 14(7), 1442. https://doi.org/10.3390/land14071442

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop