Evaluation and Prediction of Ecological Quality Based on Remote Sensing Environmental Index and Cellular Automata-Markov

Abstract

The evaluation and prediction of ecological environmental quality are essential for sustainable development and environmental management, particularly in regions experiencing rapid urbanization and industrial growth like Johor in southern Peninsular Malaysia. This study evaluates the temporal and spatial changes in ecological environmental quality in Johor from 1990 to 2020 using the Remote Sensing Environmental Index (RSEI) and Cellular Automata-Markov (CA-Markov). A CA-Markov model was employed to predict ecological environmental quality for the next 12 months based on historical data. The results reveal significant changes over the 30 years, highlighting the dynamic nature of ecological conditions. The prediction results indicate that areas with excellent ecological quality are primarily focused in the central and northern regions, while the southern and eastern edges show mixed ecological conditions. The western region, characterized by intensive land use, shows significant environmental degradation. The poorest ecological points are mainly distributed in urban and semiurban areas with frequent human activities, such as cities, ports, and villages. These findings highlight the need for targeted environmental policies and management strategies to mitigate ecological degradation and promote sustainable development in Johor State of Peninsular Malaysia.

1. Introduction

The evaluation of ecological environmental quality plays a crucial role in supporting ecosystem restoration and sustainable development. As ecosystems face increasing pressures from human activities, climate change, and land use intensification, it has become essential to conduct scientific and systematic monitoring and assessment of ecological conditions. Such evaluations provide critical insights into the current state and dynamic changes of ecosystems, serving as a foundation for ecological protection and restoration efforts. By identifying areas of environmental degradation and predicting future trends, ecological environmental quality assessments offer valuable scientific evidence for policymakers, enabling informed decision-making and effective management practices.

Johor is a state located in southern Peninsular Malaysia and is a region experiencing rapid economic growth, urbanization, and industrialization. This development, while contributing to economic prosperity, has placed significant pressure on the region’s ecological environment, leading to habitat loss, deforestation, and environmental degradation. As a key area with diverse ecosystems, including forests, wetlands, and coastal zones, Johor’s ecological health is vital for maintaining regional biodiversity, supporting local livelihoods, and ensuring sustainable development. The evaluation of ecological environmental quality in Johor is particularly important for understanding the spatial and temporal impacts of human activities and natural processes on its ecosystems. Such evaluations can help identify critical areas requiring immediate restoration and conservation efforts while providing a scientific basis for managing urban expansion, industrial activities, and resource exploitation.

While ecological environmental quality evaluation is critical for Johor, existing methods often face several limitations that hinder their effectiveness. Traditional evaluation approaches tend to rely on fragmented field surveys and localized data, which can be time-consuming, costly, and insufficient for capturing large-scale spatial and temporal changes. Additionally, these methods often cannot integrate multiple environmental indicators, leading to incomplete or inaccurate assessments of ecological conditions. In Johor, the rapid pace of urbanization and industrialization exacerbates these challenges, as conventional methods struggle to keep up with the dynamic changes in land use and ecosystem health. Furthermore, predictive analysis is rarely incorporated into existing evaluations, limiting the ability to foresee future environmental trends and plan proactive conservation strategies. To address these issues, there is a need for advanced, integrated approaches that utilize remote sensing technology and predictive models to provide comprehensive, efficient, and reliable assessments of ecological environmental quality in Johor.

In previous studies, remote sensing technology has played a significant role in the evaluation of ecological environmental quality, providing an effective means for regional ecological monitoring and dynamic analysis. For instance, vegetation index models, such as the Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI), have been widely used to monitor changes in vegetation coverage, assess forest health, and estimate agricultural yields [1,2,3]. Land cover and land use change detection techniques leverage multi-temporal remote sensing data to analyze dynamic changes such as urban expansion and deforestation [4]. Ecosystem productivity assessment methods combine remote sensing data with photosynthesis models to study carbon cycles and the impacts of climate change on ecosystems [5,6]. In water quality assessments, remote sensing technology extracts parameters such as water transparency and chlorophyll content through spectral analysis to monitor water quality conditions [7]. High-resolution remote sensing imagery, integrated with ground survey data, has also been utilized for biodiversity evaluation and wetland monitoring, supporting ecological restoration planning through the construction of habitat distribution models [8,9,10].

The Remote Sensing Ecological Index (RSEI) is a comprehensive ecological evaluation method that integrates four key ecological factors, such as greenness, humidity, dryness, and heat. By employing principal component analysis (PCA) to construct a composite index, RSEI effectively eliminates the subjectivity associated with weight assignment in traditional methods [11,12,13,14,15]. In recent years, researchers have optimized RSEI to suit the characteristics of different regions, leading to the development of variants such as the Multi-Indicator Remote Sensing Ecological Index (MSRE) and the Integrated Remote Sensing Ecological Index (IRSEI), which are tailored to the specific features of various ecosystems [16,17,18,19]. RSEI has been widely applied in ecological quality assessments across urbanized areas, wetland conservation regions, and desertification-prone zones [20,21,22,23,24].

In order to enhance the predictive capabilities of ecological environment assessments [25,26,27,28,29,30], the CA-Markov model has been widely used to simulate land use changes and their impacts on ecological environments [18,26,31,32,33,34,35]. By combining the strengths of Cellular Automata (CA) and Markov Chain, this model efficiently simulates dynamic ecosystem changes across temporal and spatial scales [36,37,38,39,40,41]. For instance, Li et al. utilized the CA-Markov model to analyze future ecological quality changes in urban clusters in central Yunnan [36]; Yang et al. compared the impacts of urbanization on ecological environments across different regions in China [38]; and Wang et al. simulated land use changes in the Heihe River Basin to evaluate ecological quality variations under different scenarios [41].

Despite significant progress in the application of remote sensing technology and dynamic models in ecological environment assessments, existing research still has several limitations. First, traditional methods often rely on scattered field surveys or single indicators, making it difficult to comprehensively reflect the complexity and dynamic changes of ecosystems. Second, the predictive capabilities of current studies are limited, and they cannot provide strong support for the development of forward-looking conservation strategies. Moreover, in Johor, a region with a complex and diverse ecological environment, existing assessment methods lack the ability to effectively integrate multi-scale analyses and adapt models to the region’s specific characteristics.

This study focuses on Johor State, Peninsular Malaysia, a rapidly developing region, to evaluate and predict its ecological environmental quality using the Remote Sensing Environmental Index (RSEI) and CA-Markov model. This study aims to support targeted ecological restoration strategies, ensuring balanced development and environmental sustainability. This study utilizes the Google Earth Engine (GEE) cloud platform to process multi-temporal remote sensing data, using Landsat 5 satellite images recorded from 1990 to 2013 and Landsat 8 satellite images recorded from 2013 to 2023. The RSEI values for 34 years are calculated to comprehensively analyze the dynamic changes in the ecological quality of Johor. By integrating remote sensing technology and predictive models, this study reveals the spatial distribution, temporal trends, and potential driving factors of Johor’s ecological environment.

The contributions of this study are as follows:

(i)

A comprehensive method for assessing and predicting ecological environment quality is proposed, which integrates the RSEI with the CA-Markov model. By leveraging the spatial characteristics of remote sensing data and the dynamic spatiotemporal simulation capabilities of the CA-Markov model, this approach achieved integrated ecological environment analysis in the Johor region, addressing the research gap in large-scale, long-term time series prediction of ecological environment dynamics.

(ii)

Quantitative spatial distribution and temporal trend analysis reveal long-term ecological change patterns and the relationship between regional differences in ecological quality and spatial aggregation characteristics, providing a theoretical foundation for formulating ecological conservation and restoration strategies.

(iii)

Spatial autocorrelation analysis is introduced to identify spatiotemporal distribution patterns of priority ecological conservation zones, vulnerable areas, and transitional regions through autocorrelation analysis of ecological quality data across different periods. This significantly enhances the understanding of spatial heterogeneity in regional ecological environments.

(iv)

Based on spatiotemporal analysis results, this study proposed targeted ecological protection and management strategies for rapidly urbanizing and industrializing zones in Johor, along with concrete recommendations for balancing economic development and ecological conservation. The policy implications offer valuable references for ecological environment management in similar regions.

2. Materials and Methods

2.1. Materials

2.1.1. Study Area

Johor is located at the southernmost tip of Peninsular Malaysia, with geographical coordinates ranging from 1°20′ N to 2°35′ N latitude and 103°20′ E to 104°20′ E longitude, as shown in Figure 1. The region has a typical tropical rainforest climate, with annual rainfall ranging between 2000 mm and 3000 mm, concentrated mainly during the northeast monsoon period from November to March. The temperature remains relatively high throughout the year, with an average range between 25 °C and 30 °C. The hottest months are May and June, when temperatures can exceed 30 °C, while December and January are relatively cooler, with temperatures ranging from 24 °C to 26 °C. Johor is home to diverse wetland ecosystems, including coastal wetlands, inland wetlands, swamps, and estuarine areas. These wetland ecosystems play a crucial role in maintaining biodiversity, providing wildlife habitats, regulating the water cycle, and controlling flooding. The eastern and southern coastlines of Johor, particularly around Johor National Park and Pulau Kukup National Park, are characterized by extensive mangrove forests. These mangroves are vital in preventing coastal erosion, protecting coastal ecosystems, and supporting fisheries production. In aquaculture, Johor’s activities are mainly concentrated in the coastal and estuarine areas, particularly around Batu Pahat and Mersing. The main aquaculture species in these areas include tilapia, prawns, and various shellfish. Wetlands and mangrove forests provide essential ecological conditions for aquaculture while playing an irreplaceable role in maintaining ecological balance and protecting biodiversity.

2.1.2. Data Sources

The remote sensing data collected for this study were collected from cloud-free images captured by the Landsat 5 and Landsat 8 satellites, as shown in

Table 1. Data sources used in this study.

Data	Time	ee.ImageCollection	Resolution (m)
Landsat 5	1990–2012	(“LANDSAT/LT05/C02/T1_L2”)	30
Landsat 8	2013–2023	(“LANDSAT/LC08/C02/T2_L2”)	30
MODIS		(“MODIS/061/MOD11A1”)	LST_1000
Water(mask)		JRC/GSW1_3/Yearly History	30

. Landsat data were primarily used to calculate Normalized Difference Vegetation Index (NDVI), Wetness Index (WET), Normalized Difference Built-up Space Index (NDBSI), and Land Surface Temperature (LST), Principal Component Analysis (PCA), and Remote Sensing Environmental Index (RSEI), providing the foundation for analyzing trends in ecological landscape patterns, conducting spatial autocorrelation analysis, and preparing predictive models. MODIS data were used as a substitute when the ST_10 band for a specific year was damaged and unable to compute LST, and the data fusion method is given in Equation (4) listed in Table 1. The JRC/GSW1_3 Yearly Water Classification History dataset was employed to perform water masking.

2.2. Methods

2.2.1. Remote Sensing Ecological Index (RSEI)

The Remote Sensing Ecological Index (RSEI) is a comprehensive indicator designed to effectively assess regional ecological environmental quality, providing a holistic reflection of ecosystem health. The calculation of RSEI integrates multiple ecological factors, including the Normalized Difference Vegetation Index (NDVI), Wetness Index (WET), Normalized Difference Built-up Space Index (NDBSI), and Land Surface Temperature (LST), as outlined in

Table 2. The calculation methods for RSEI.

Index				Sources			No.
$\mathrm{NDVI}={\displaystyle \frac{\left(B_{\mathrm{NIR}}- {\mathrm{B}}_{\mathrm{RED}}\right)}{\left(B_{\mathrm{NIR}} -{\mathrm{B}}_{\mathrm{RED}}\right)}}$				Xu et al., 2013 [11]; Crist, 1985 [42]; Chen et al., 2022 [43].			(1)
L5_Wet = 0.0315 ∗ B1 + 0.2021 ∗ B2 + 0.3012 ∗ B3 + 0.1594 ∗ B4 − 0.6806 ∗ B5 − 0.6109 ∗ B7				Xu et al., 2013 [11]; Crist, 1985 [42]; Baig et al., 2014 [44]			(2)
L8_Wet = 0.1509 ∗ B2 + 0.1973 ∗ B3 + 0.3279 ∗ B4 + 0.3406 ∗ B5 − 0.7122 ∗ B6 − 0.4572 ∗ B7
$\mathrm{IBI}={\displaystyle \frac{2\frac{{\mathrm{B}}_{\mathrm{SWIR1}}}{B_{\mathrm{SWIR1}}+{\mathrm{B}}_{\mathrm{NIR}}}-\frac{B_{\mathrm{NIR}}}{B_{\mathrm{NIR}}+{\mathrm{B}}_{\mathrm{RED}}}+\frac{B_{\mathrm{GREEN}}}{B_{\mathrm{GREEN}}+{\mathrm{B}}_{\mathrm{SWIR1}}}}{2\frac{{\mathrm{B}}_{\mathrm{SWIR1}}}{B_{\mathrm{SWIR1}}+{\mathrm{B}}_{\mathrm{NIR}}}+\frac{B_{\mathrm{NIR}}}{B_{\mathrm{NIR}}+{\mathrm{B}}_{\mathrm{RED}}}+\frac{B_{\mathrm{GREEN}}}{B_{\mathrm{GREEN}}+{\mathrm{B}}_{\mathrm{SWIR1}}}}}$				Xu et al., 2013 [11]; Hu et al., 2018 [45]; Jiménez-Muñoz et al., 2014 [46]			(3)
$\mathrm{SI}={\displaystyle \frac{\left(B_{\mathrm{SWIR1}}+{\mathrm{B}}_{\mathrm{RED}}\right)-\left(B_{\mathrm{NIR}}+{\mathrm{B}}_{\mathrm{BLUE}}\right)}{\left(B_{\mathrm{SWIR1}}+{\mathrm{B}}_{\mathrm{RED}}\right)+\left(B_{\mathrm{NIR}}+{\mathrm{B}}_{\mathrm{BLUE}}\right)}}$
$\mathrm{NDBSI}={\displaystyle \frac{\mathrm{IBI}+\mathrm{SI}}{2}}$
$L_{\lambda }=\mathrm{Gain}\ast \mathrm{DN}+\mathrm{bias}$				Xu et al., 2013 [11]; Hu et al., 2018 [45];			(4)
$T_{\mathrm{BT}}={\displaystyle \frac{K_2}{\ln (\frac{K_1}{L_{\lambda }}+1)}}$
$\mathrm{pv}=\left({{\displaystyle \frac{NVDI-NDV{I}_{min}}{NDV{I}_{max}}}}^2\right)$
ε = 0.004 ∗ pv + 0.986
$\mathrm{LST}={\displaystyle \frac{T_{\mathrm{BT}}}{1+ ( \lambda {\mathrm{T}}_{\mathrm{BT}}/\rho \left) \ast \mathrm{In}\right(\epsilon )}}$
ρ = 1.438 ∗ 10−2	DN	λ	gain	bias	K1	K2
Landsat5	BSWIR1=B6	11.450 µm	0.0552	1.2378	607.76	1260.56
Landsat8	BSWIR1=B10	10.895 µm	-	-	774.89	1321.08
$\mathrm{RES}{\mathrm{I}}_0=1-\beta 1\ast \mathrm{NDVI}+\beta 2\ast \mathrm{W}\mathrm{E}\mathrm{T}+\beta 3\ast \mathrm{N}\mathrm{D}\mathrm{B}\mathrm{S}\mathrm{I}+\beta 4\ast \mathrm{L}\mathrm{S}\mathrm{T}$							(5)
β1, β2, β3, and β4 are the main components of PCA
$\mathrm{RSEI}={\displaystyle \frac{\mathrm{RSE}{\mathrm{I}}_0-\mathrm{RES}{\mathrm{I}}_{0-\min }}{\mathrm{RES}{\mathrm{I}}_{0-\max }-\mathrm{RES}{\mathrm{I}}_{0-\min }}}$							(6)

. The RSEI methodology is particularly well-suited for evaluating Johor’s ecological environment. Its inclusion of the Wetness Index enables precise characterization of the dynamic changes in Johor’s wetland and mangrove ecosystems, while the Land Surface Temperature and Built-up Space Index provide critical support for quantifying the impacts of urbanization. Through Principal Component Analysis (PCA), the primary information (PC1) from these four indicators is extracted, enabling a dimensional reduction and comprehensive analysis of ecological environmental quality. Standardization and normalization of the data further allow the classification of RSEI into five levels (from Poor to Excellent), laying the groundwork for spatiotemporal change analysis and spatial clustering studies.

Johor, with its diverse ecosystems, relies heavily on the ecological services provided by its wetlands, mangroves, and coastal areas for regional biodiversity conservation and environmental management. Based on Johor’s ecological characteristics, this study utilizes remote sensing imagery from 1990 to 2023 to calculate RSEI values for different years, identifying the spatial distribution patterns and temporal trends in ecological environmental quality. By conducting spatial autocorrelation analysis of RSEI levels, this study uncovers the spatial clustering patterns and dynamic changes in ecological quality. Additionally, the CA-Markov model is applied to predict future ecological quality, simulating the potential impacts of urbanization and land-use changes on the ecosystem. The RSEI assessment framework designed in this study aims to provide a scientific basis for ecological conservation and land-use optimization in Johor, while also offering a theoretical model and practical approach for ecological monitoring and management in similar regions.

2.2.2. Spatial Auto-Correlation Analysis

The diversity and regional characteristics of Johor’s ecological environment highlight the importance of local spatial autocorrelation analysis. By integrating local Moran’s I, Z-values, and p-values, it is possible to accurately identify ecological priority zones, vulnerable areas, and transitional zones within Johor’s complex ecological landscape.

Spatial autocorrelation methods, particularly through global indicators such as Moran’s I, are highly effective in evaluating the spatial relationships of ecological quality across different regions in Johor. These methods provide insights into whether the spatial distribution of ecological quality is uniform [47,48]. Johor’s ecosystems, including wetlands, mangroves, and coastal zones, often exhibit significant ecological gradients when juxtaposed with adjacent urban and industrial areas. Using Moran’s I, as expressed in Equation (7), the spatial clustering effect of these ecological differences can be quantified, along with the strength of the spatial relationships between high-quality and low-quality ecological areas [49]. A Moran’s I value approaching +1 indicates that high-quality ecological areas are spatially clustered, potentially forming critical ecological conservation zones or corridors. Conversely, a value nearing −1 may reflect pronounced ecological quality differentiation, such as the negative impact of industrial activities on surrounding wetland ecosystems.

The analyses for the Johor area are instrumental in identifying high-priority areas for ecological conservation as well as ecological vulnerability zones or areas requiring restoration efforts. The application of spatial autocorrelation methods thus deepens the understanding of the spatial dependencies of ecological quality in Johor, providing a robust foundation for designing targeted ecological protection and management policies [47,48,49].

$$I={\displaystyle \frac{n}{{\sum }_{i=1}^n{\sum }_{j=1}^n{w}_{ij}}}{\displaystyle \frac{{\sum }_{i=1}^n{\sum }_{j=1}^n{w}_{ij}(x_i-\stackrel{-}{x})(x_j-\stackrel{-}{x})}{{\sum }_{i=1}^n(x_i-\stackrel{-}{x}{)}^2}}$$

(7)

where n represents the number of observations; x_i and x_j are the attribute values of spatial units i and j, respectively; $\overline{x}$ is the mean of all attribute values across spatial units; and w_ij is the weight between units i and j in the spatial weight matrix.

The index of local indicators of spatial association (LISA) is shown in Equation (8), which calculates the relationship between each grid cell and the ecological environment quality of its neighboring areas, providing a more detailed perspective on the spatial heterogeneity of ecological quality within Johor [50]. For instance, wetland and mangrove areas may exhibit H-H clustering types (high–high), indicating regions of ecological importance and good conditions for conservation. Conversely, areas undergoing rapid urbanization, such as Johor Bahru and its surroundings, may form L-L clustering types (low–low), reflecting the negative impacts of urbanization on the ecological environment. Additionally, L-H or H-L types can identify ecological boundaries or transition zones, such as the potential threats posed by urban expansion to wetland ecosystems.

$$\mathrm{LISA}={\displaystyle \frac{x_i-\stackrel{-}{x}}{{\sum }_{i=1}^n(x_i-\stackrel{-}{x}{)}^2/n}}\sum _{j=1}^n{w}_{ij}(x_j-\stackrel{-}{x})$$

(8)

LISA cluster maps categorize local spatial autocorrelation of different regions in Johor into the following five types: high–high (H-H), low–low (L-L), low–high (L-H), high–low (H-L), and not significant. These classifications describe the relative ecological quality of a specific area compared to its neighboring areas. H-H indicates that both the selected area and its neighboring regions exhibit high ecological quality; L-L signifies low ecological quality for both; L-H implies that the selected area has low ecological quality while its neighbors are relatively better; and H-L represents the opposite, where the selected area has high ecological quality, but the surrounding areas are relatively poorer [51].

A positive LISA index value (typically indicating positive autocorrelation) corresponds to high–high or low–low cluster regions, whereas a negative LISA index value (typically indicating negative autocorrelation) corresponds to high–low or low–high neighbors. For instance, aquaculture areas in Batu Pahat and Mersing may fall under the L-H category, suggesting low ecological quality in these regions but relatively better conditions in adjacent wetlands or mangroves. Such classifications assist policymakers in identifying key areas where low ecological quality significantly impacts neighboring regions, enabling the prioritization of ecological restoration measures.

In analyzing the ecological quality of Johor, the Z-value can be calculated using Equation (9), while the p-value can be computed using Equation (10). When the Z-value exceeds 1.96, it indicates the presence of spatial autocorrelation in certain regions of Johor, such as H-H clustering patterns in coastal mangroves or wetland conservation areas. Conversely, urbanized areas may exhibit L-L clustering. If the p-value falls below typical significance levels, such as 0.05 or 5%, the observed pattern is considered significant, confirming the existence of spatial autocorrelation.

$$Z(I)={\displaystyle \frac{I-E(I)}{\sqrt{\mathrm{VAR}(I)}}}$$

(9)

where E(I) represents the mathematical expectation and VAR(I) denotes the variance.

The p-value is calculated based on the Z score, assuming that the Z score follows a standard normal distribution, as follows:

$$\mathrm{p}=2 \times \mathsf{\Phi } (-\mid \mathrm{Z}\mid )$$

(10)

where Φ is the cumulative distribution function (CDF) of the standard normal distribution. For a one-sided test, when the p-value is less than 0.01, it indicates that the Moran’s I index has strong statistical significance.

2.2.3. Cellular Automata-Markov (CA-Markov)

The CA-Markov model is an integrated model combining Cellular Automata (CA) and Markov Chain methodologies. It is widely used to simulate and predict the spatiotemporal evolution of spatial data, particularly in land use change studies within the field of Geographic Information Systems (GIS) [52]. By integrating the state transition probabilities of Markov Chains with the spatial neighborhood rules of CA, the model enables accurate prediction and simulation of dynamic change processes in both temporal and spatial dimensions [53,54].

CA is a discrete model consisting of a regular grid and simple rules, where each grid cell (referred to as a cell) updates its state at each time step based on predetermined rules and the states of its neighboring cells. This spatial dependency allows CA to reflect the spatial relationships and neighborhood effects in geographic regions. The Markov Chain, however, is a statistical model that describes the probabilities of transitioning from one state to another depending only on its current state without any memory of the previous states. The CA-Markov model combines these two features, utilizing the Markov Chain to predict the transition probabilities for each cell’s state while using the CA rules to simulate the spatial change process.

In this study, the ecological quality levels derived from RSEI are used as the cell states for the CA-Markov model. The RSEI is categorized into the following five levels with corresponding labels: Poor (1), Fair (2), Moderate (3), Good (4), and Excellent (5). The model leverages the state changes among these levels to build a Transition Probability Matrix (TPM), which is used to predict future trends in ecological quality.

The first step involves classifying the RSEI results for Johor from 1990 to 2020 into five levels (Poor to Excellent) and assigning a cell state to each year based on these categories. At each time step, the total number of transitions between different states is counted—for example, transitions from Poor to Poor (1→1), Poor to Fair (1→2), and so on. The transition probability for a particular state can be calculated as follows:

$$\begin{gathered} P=\left[\begin{array}{ccccc}p11& p12& p13& p14& p15\\ p21& p22& p23& p24& p25\\ p31& p32& p33& p34& p35\\ p41& p42& p43& p44& p45\\ p51& p52& p53& p54& p55\end{array}\right] \\ {P}_{\mathrm{ij}}\in \left[0, 1\right), \sum _{i=1}^n{P}_{ij}=1, i, j=1, 2, 3\dots \dots n. \end{gathered}$$

(11)

$$\mathrm{PRE}\_\mathrm{RES}{\mathrm{I}}_{2024}=\mathrm{RSE}{\mathrm{I}}_{2023}•P$$

(12)

where P is a probability matrix for the entire map, and the prediction probability p in the Markov model represents the probability P^ij for each pixel.

As shown in Equation (11), the prediction of the ecological index for the next year can be given by the ecological index of the current year multiplied by P, as follows:

$$\mathrm{PRE}\_{\mathrm{RESI}}_{2025}={\mathrm{RSEI}}_{2024}•P_{\mathrm{new}}$$

(13)

$${\lim }_{n\to \infty }\pi =\pi •P_{1990-2020}$$

(14)

where Pnew is an updated probability matrix P, RSEI2024 is the RSEI in the year 2024, PRE_RSEI2025 is the predicted RSEI in the year 2025, and P1990–2025 is the transfer probability matrix obtained from the years 1990~2020. The probability matrix P is calculated using the fixed ecological indices from 1990 to 2023 by statistically analyzing the transition probabilities of each pixel. After predicting the ecological data for 2024 using Equation (12), the transition probabilities from 1990 to 2024 are recalculated to obtain a new probability matrix P_new, which is then used to predict the ecological index for 2025, as shown in Equation (13).

Markov predictions cannot be extended indefinitely, as they are subject to a limit in predictive capability. It is necessary to obtain new data for the study area to continue making accurate forecasts, and therefore the actual data from 1990 to 2024 must be obtained to ensure the accuracy and reliability of the predictions. Consequently, the CA-Markov model effectively simulates the spatiotemporal changes in ecological quality in Johor from 1990 to 2020 and predicts future ecological states, providing scientific decision support for ecological conservation and land use planning in the region [52,53,54]. The CA-Markov model plays a significant role in the evaluation of Johor’s ecological quality, particularly in dynamic land-use change simulation, future ecological quality prediction, and multi-scenario analysis for decision-making support.

The CA-Markov model precisely captures the spatiotemporal dynamics of land-use changes in Johor, simulating the transformation processes between different land-use types such as urban land, agricultural land, forests, and wetlands. This capability helps to reveal the impacts of land-use changes on ecological quality. By analyzing historical data and current patterns, the CA-Markov model can predict future land-use change trends in Johor, providing a scientific basis for regional ecological planning and policymaking. For example, the model can predict the impact of urban expansion on critical ecosystems like wetlands and forests, supporting the delineation of priority conservation areas.

Moreover, the CA-Markov model enables simulations under various development scenarios, such as economically driven, ecologically prioritized, or integrated development approaches. This allows the assessment of the potential impacts of land-use decisions on ecological quality, providing valuable insights for optimizing land-use strategies and balancing development with environmental sustainability.

2.2.4. Rolling Forecast Validation

Rolling forecast validation is a commonly used method for evaluating time series forecasting models, primarily aimed at assessing the model’s performance and stability through a gradual expansion of the training dataset. The core concept of this method is to use all available data from the starting point to the current time point for model training at each prediction step, followed by forecasting the data for the next time point. The actual values are then compared with the predicted values to evaluate the model’s forecasting accuracy [55,56].

In this study, the model is first trained using data from 1990 to 2018 to predict the ecological quality values for 2019. The training data is then extended to include data from 1990 to 2019 to predict the values for 2020, and so on until the data from 1990 to 2020 is used to predict the values for 2023. The predicted values are compared with the actual values to calculate the mean squared error (MSE) and the coefficient of determination (R²) for each year, evaluating the model’s stability and accuracy during the rolling forecast process [55,56] such that

$$\mathrm{MSE}={\displaystyle \frac{1}{n}}{\sum _{i=1}^n(y_i-\stackrel{\Lambda }{y_i})}^2$$

(15)

$$\begin{gathered} R^2=1-{\displaystyle \frac{{\sum }_{i=1}^n(y_i-\stackrel{\Lambda }{y_i}{)}^2}{{\sum }_{i=1}^n(y_i-\stackrel{-}{y}{)}^2}} \\ i, j=1, 2, 3\dots \dots ,\mathrm{n}. \end{gathered}$$

(16)

where $\stackrel{\mathsf{\Lambda }}{y_i}$ is the predicted value,$\stackrel{-}{y}$ is the average value, and $y_i$ is the true value. A smaller MSE value and an R² value closer to one indicate a higher precision.

The advantage of this validation method lies in its ability to dynamically reflect the model’s forecasting performance over an expanding time series. Specifically, in forecasting the ecological quality of Johor, rolling forecast validation not only effectively tests the stability of the CA-Markov model but also identifies potential biases in different periods. This provides a basis for model optimization and parameter adjustments, ensuring more accurate and reliable predictions.

3. Results and Analysis

3.1. Factor Attributes

After calculating the NDVI, WET, NDBSI, and LST data for the period from 1990 to 2023 using Equations (1)–(4), this study conducted Principal Component Analysis (PCA) on the results for four different time spots, such as 1990, 2000, 2010, and 2020. The PCA analysis decomposed the covariance matrix of the standardized ecological indicators to extract the principal components that explain the variance in the data. Principal components (e.g., PC1, PC2, etc.) were selected based on their eigenvalues, and their loadings and contribution rates across the indicators were calculated.

Table 3. Results of the PCA of the indices in the years 1990, 2000, 2010, and 2020.

Year	Principal Component	NDVI	WET	NDBSI	LST	Percent Correlation Eigenvalue (%)	Eigenvalue
		β1	β2	β3	β4
1990	PC1	0.171	0.249	−0.033	−0.952	80.01	0.0064
	PC2	0.372	0.301	−0.859	−0.175	12.49	0.0049
	PC3	0.329	−0.081	−0.467	−0.247	7.35	0.0018
	PC4	0.850	0.485	−0.202	0.017	0.15	0.0011
2000	PC1	0.818	−0.074	−0.425	0.088	81.55	0.033
	PC2	0.669	0.030	−0.713	−0.203	10.23	0.004
	PC3	0.702	0.213	−0.530	−0.423	7.81	0.003
	PC4	0.147	−0.097	−0.171	0.032	0.41	0.001
2010	PC1	0.103	0.450	−0.319	−0.930	81.25	0.033
	PC2	0.708	−0.361	−0.361	−0.180	9.61	0.004
	PC3	0.678	−0.129	−0.649	−0.317	8.64	0.003
	PC4	0.168	−0.091	−0.374	−0.036	0.50	0.001
2020	PC1	0.136	0.343	−0.076	−0.925	84.17	0.026
	PC2	0.683	−0.696	−0.141	−0.169	10.89	0.003
	PC3	0.705	−0.624	−0.061	−0.331	4.57	0.001
	PC4	0.131	0.088	−0.985	0.068	0.37	0.001

shows the calculation results with the contribution rates of the principal components and the variable loadings for each time spot.

As shown in Table 3, for the imagery from 1990, 2000, 2010, and 2020, the proportion of variance explained by the first principal component (PC1) exceeded or equaled 80.01%, with the highest being 84.17% in 2020. Therefore, this study selected the loadings of PC1 (β₁, β₂, β₃, and β₄) as weight coefficients to calculate the Remote Sensing Ecological Index (RSEI) for each corresponding year using Equation (5). Furthermore, the results were normalized according to Equation (6) to obtain standardized RSEI values (Normalized RSEI, NRSEI). Finally, the Normalized RSEI values were used to generate spatial landscape maps for the respective years, visually depicting the spatial distribution characteristics and trends in ecological environment quality in Johor for 1990, 2000, 2010, and 2020.

PCA was employed to extract the most explanatory components from ecological indicators, including the NDVI, WET, NDBSI, and LST. The contribution weights of these indicators exhibited significant variations across the distinct years of 1990, 2000, 2010, and 2020, reflecting dynamic shifts in regional ecological conditions and the progressive impact of human activities.

Around 1990, the primary drivers of ecological quality were low urbanization and limited industrialization. NDVI held a relatively high weight, indicating that vegetation coverage and green spaces played a dominant role in sustaining ecological health. Meanwhile, WET and NDBSI contributed minimally to the principal components, suggesting negligible impacts from moisture fluctuations or urban expansion at this stage.

By 2000, accelerated urbanization and industrialization, particularly in coastal regions, had marked this period. As human activities intensified, the weight of LST gradually increased, highlighting the growing influence of temperature variations on ecological quality. Concurrently, NDVI’s weight slightly declined, likely due to reduced green spaces caused by urban sprawl and agricultural expansion.

Around 2010, the urbanization and industrialization in regions like Johor reached a mature phase. Rising temperatures, exacerbated by industrial emissions and transportation activities in urban areas, became more pronounced. LST’s weight surged further, dominating the principal components. NDBSI also began to show increased significance, reflecting the escalating ecological pressure from urban infrastructure development, particularly in industrialized zones.

In 2020, the implementation of ecological conservation policies, such as protections for northwestern wetlands and mangrove reserves, led to slight rebounds in NDVI and WET weights. This signaled improved roles of vegetation and moisture conditions in ecological restoration. However, LST retained a high weight, underscoring persistent temperature impacts from climate change and industrial activities.

These temporal shifts in PCA weights illustrate the interplay between natural ecological processes and anthropogenic forces, offering critical insights for sustainable environmental management.

3.2. Spatiotemporal Distribution and Changes in Ecological Environmental Quality

Based on the five RSEI classification standards mentioned earlier, the RSEI indices for the study area were classified in ArcGIS, producing spatiotemporal landscape pattern maps for 1990, 2000, 2010, 2020, and the 30-year average, as shown in Figure 2. These maps clearly illustrate the spatial distribution and changing trends of RSEI in Johor during the study period. Overall, regions with “Excellent” and “Good” ecological status are primarily located in the central, northern, western, and eastern areas, while “Poor” RSEI values are mainly concentrated in the southern urban areas and their surroundings.

Specifically, in 1990, the areas with “Poor” RSEI values were concentrated in the southern urban regions, reflecting that urban planning and environmental protection measures at the time were primarily focused on the city and its surrounding areas. By 2000, the “Poor” RSEI value areas gradually expanded from the southern urban areas to the surrounding towns and rural regions, indicating the broader impact of urbanization and the increased ecological pressure in rural areas. In 2010, the “Poor” RSEI values further extended to include water bodies and their riverbank areas, such as reservoirs and rivers, indicating significant human interference with water resources and aquatic ecosystems. By 2020, the “Poor” RSEI values covered more wetland parks and mangrove protection areas, reflecting increased disturbance in the watershed’s surrounding areas, while also suggesting that ecological protection zones and nature parks may have been established and expanded in an attempt to mitigate this trend.

Table 4. Ecological level area statistics of Johor.

Year	1990		2000		2010		2020
RESI Level	Area (km2)	Pct. (%)	Area (km2)	Pct. (%)	Area (km2)	Pct. (%)	Area (km2)	Pct. (%)
Excellent (0.8–1)	11,476	60.16	10,772	56.47	10543	55.27	10,376	54.39
Good (0.6–0.8)	5024	26.34	5288	27.72	5627	29.50	5776	30.28
Mod (0.4–0.6)	1552	8.14	1820	9.54	1736	9.10	1580	8.28
Fair (0.2–0.4)	612	3.21	704	3.69	721	3.78	731	3.84
Poor (0–0.2)	412	2.16	492	2.58	449	2.35	613	3.21
	19,076	100.00	19,076	100.00	19,076	100.00	19,076	100.00

presents the area and proportion changes of different RSEI levels in Johor from 1990 to 2020. The data indicates that over the past three decades, the spatial distribution of ecological environment quality in Johor has undergone some changes. First, the area of the highest quality (RSEI 0.8–1) decreased from 11,476 square kilometers in 1990 to 10,376 km² in 2020, a reduction of 1100 km². This suggests that the highest-quality ecological areas have shrunk over the past 30 years, possibly due to urbanization, land use changes, and ecosystem degradation. In contrast, the area of the “Good” category (RSEI 0.6–0.8) gradually increased from 5024 km² in 1990 to 5776 km² in 2020, an increase of 752 km², which reflects improvements in some medium-quality ecological areas at certain times. The area of the “Moderate” category (RSEI 0.4–0.6) increased by 268 km² between 1990 and 2000 but then decreased each year, showing only a 28 km² increase by 2020 compared to 1990, indicating a more fluctuating change in ecological quality at this level. The area of the “Fair” category (RSEI 0.2–0.4) showed a consistent annual increase, growing from 621 km² in 1990 to 731 km² in 2020, an increase of 110 km². Furthermore, the area of the lowest quality category (RSEI 0–0.2) significantly increased from 412 km² in 1990 to 612 square kilometers in 2020, a total increase of 200 km². This indicates a clear deterioration in the ecological environment quality in some areas, especially those more heavily impacted by human activities and land use.

Figure 3 illustrates the percentage change in the area of different ecological environment quality levels in Johor from 1990 to 2020. The proportion of excellent quality (RSEI 0.8–1) decreased from 60.16% in 1990 to 54.39% in 2020, with a decrease of 5.77%. A reduction in the share of high-quality ecological areas in Johor’s total area indicates that high-quality ecological regions are gradually shrinking due to the impacts of rapid urbanization and land use changes. The proportion of good quality (RSEI 0.6–0.8) increased from 26.34% in 1990 to 30.28% in 2020. An increase of 3.94% suggests that areas with better environmental quality have gained a larger share, likely due to ecological protection and restoration measures in some regions. The proportion of the lowest quality (RSEI 0–0.2) increased from 2.16% in 1990 to 3.21% in 2020. A rise of 1.05% shows an increase in the proportion of areas with the worst environmental quality, reflecting the ecological degradation issue in certain areas, which is a concerning negative trend.

3.3. Spatial Autocorrelation Analysis of Ecological Quality

By calculating the Normalized RSEI average values for Johor from 1990 to 2020, this study uses the UTM-Malaysia-48N projection, resampling the 30 m resolution imagery to a 1 km resolution. The spatial autocorrelation analysis function of GIS is then used to compute the Moran’s I value in Equation (7) and the Z value in Equation (9). It can be seen that the Moran’s I value is 0.40, the Z value is 38.11, and the p-value is less than 0.001, suggesting a significant positive spatial autocorrelation in the ecological quality distribution, as shown in Figure 4. The ecological conditions exhibit a clustered spatial distribution rather than a random one. From the spatial perspective, as shown in Figure 5, forested areas, urban regions, and plantations present significant clustering characteristics, providing a foundation for subsequent clustering and outlier analysis.

Through the ArcGIS spatial analysis of the RSEI data for 1990, 2000, 2010, and 2020, the results reveal the spatiotemporal evolution characteristics of the ecological quality in Johor, as shown in Figure 6. From the analysis, it can be seen that HH and LL clusters exhibit significant spatial clustering characteristics, reflecting the polarization of ecological conditions within the region. Among them, the HH clusters are mainly concentrated in urbanized areas with intense economic activity, while the LL clusters are primarily found in areas with relatively untouched natural environments or lower levels of development. Furthermore, the HL and LH outlier distributions show a certain spatial dispersion, indicating that these areas may have special ecological or resource utilization features. Further scatter plot analysis in Figure 7 presents the spatial correlation of the RSEI values, particularly highlighting the dominance of HH and LL clusters, which underscores the ecological quality disparities across different regions of Johor.

The spatial pattern changes in Johor from 1990 to 2020, as shown in Figure 6, clearly illustrate the dynamic distribution characteristics of LL and HH clusters, as well as HL and LH outliers. The LL clusters were initially concentrated in the central region of Johor in 1990, and over the past thirty years, they have gradually expanded towards the northern and eastern coastal areas, forming a large-scale aggregation in the northern region by 2020. The central region of Johor serves as an important link between the economically developed southern region and the less urbanized northern region, primarily characterized by agricultural activities and medium-scale industrial development. The northern region, bordering Pahang, Melaka, and Negeri Sembilan, has relatively low urbanization but plays a significant role in agricultural production, providing essential support for Johor’s economic development.

The HH clusters in 1990 were primarily distributed in the southern and central-western regions of Johor. Over the past 30 years, these clusters have gradually concentrated in the southern, western coastal, and northwestern areas. The southern region, centered around Johor Bahru, is the most economically active area in Johor. The western coastal region, bordering the Strait of Malacca, has become a key economic hub due to its important port facilities and industrial parks, such as large-scale development projects like the Forest City. The eastern region is renowned for its beautiful coastline and vibrant maritime activities, particularly in the Johor River Basin, where the government has actively promoted tourism and fisheries industries by leveraging its natural landscapes and geographical advantages.

It can also be seen from Figure 7 that the distribution of HL and LH outliers is more scattered. The HL areas are mainly found in the southwestern part of Johor, while the LH areas are concentrated in the northeastern part. These regions do not show significant clustering characteristics. The spatial characteristics of these areas reflect a more dispersed distribution of villages or lower levels of resources.

Figure 7 shows the scatter plot of the Moran’s I, and it can be seen that the data points from 1990 to 2020 are primarily concentrated in the first quadrant (HH) and the third quadrant (LL), indicating the presence of spatial clusters where geographically adjacent areas have similar attribute values. There are more data points distributed in the third quadrant (LL) than in the first quadrant (HH), reflecting that LL areas are mostly natural or undisturbed, while HH clusters represent urbanized or economically developed regions. Additionally, the density of data points in the second quadrant (LH) is higher than in the fourth quadrant (HL), suggesting that low-value areas are more commonly surrounded by high-value areas. This phenomenon indicates that the geographical characteristics where urban green spaces, parks, and other natural areas are surrounded by highly developed regions, or areas with scarce resources are encircled by agricultural land or plantations.

3.4. One-Year Distribution of RSEI for Johor Area

In the experiment, the CA-Markov model was applied to calculate the dynamic changes in the ecological environment distribution of Johor, based on Equations (12), (13), (15), and (16). To evaluate the performance of the CA-Markov model, different amounts of training data were used for calculation. In order to better analyze the relationships within the ecosystem, we sampled data from 1990 to 2020 at varying intervals, including 4-year, 5-year, 2-year, and 1-year intervals. The detailed sampling strategy and corresponding results are presented in

Table 5. Evaluation of the CA-Markov model for ecological outcomes with varying amounts of data.

Interval of Years	10	5	1	0
No. of years	4	7	16	30
MSE	0.4939	0.2433	0.1638	0.0778
R2	0.9686	0.8219	0.8792	0.8823

.

When the CA-Markov model was trained using four years of data, it achieved a high R² value of 0.9686; however, due to the limited dataset, the model struggled to effectively capture the long-term dynamic characteristics of Johor’s ecological environment. The prediction results exhibited poor stability and signs of overfitting. This indicates that models trained with short-term data may perform well on training datasets but lack reliability for forecasting complex ecological system changes.

As the time span of the training data increased, the model’s performance improved progressively. With seven years of data, the model began to capture the long-term trends in the ecological environment. The MSE value significantly decreased, and the R² value stabilized at 0.8219, achieving a balance between training and testing data. However, limitations in data volume still constrained the model’s ability to fully describe the complexities of the ecosystem, with a slight risk of underfitting or overfitting.

When the dataset was extended to 16 years, the model’s performance improved significantly, offering a more comprehensive reflection of Johor’s long-term ecological trends and complex dynamics. The MSE value dropped to 0.1638, while the R² value increased to 0.8792. At this stage, the model’s predictive capacity for new data was greatly enhanced, and the risk of overfitting was substantially reduced, demonstrating that a longer period of data enhances the model’s generalization ability and the credibility of its predictions.

Further extending the training data to 30 years yielded the best predictive performance, with an MSE value of 0.0778 and an R² value of 0.8823. The ample data volume allowed the model to accurately capture the dynamic changes in Johor’s ecological environment, achieving high precision and robustness in prediction results.

By analyzing the RSEI data from 1990 to 2020 and constructing the CA-Markov model based on Equation (14) for final predictions, the geographical distribution map of Johor’s ecological quality was generated, as shown in Figure 8.

It can be seen from Figure 8 that the green areas (Excellent, 0.8–1) dominate most of the study region, primarily concentrated in areas with high forest coverage. These zones exhibit low pollution levels, reflecting the effectiveness of natural resource management and conservation measures. The ecological health in these regions stands out, indicating minimal human interference and high environmental resilience.

Yellow areas (Good, 0.6–0.8) indicate regions with slightly affected but generally stable ecological conditions. These areas might experience limited human activity or have benefited from mitigation measures that prevented further ecological degradation.

Orange areas (Moderate, 0.4–0.6) represent zones with moderate ecological conditions, often associated with light industrial activities, agricultural runoff, or early-stage urbanization. While these areas have not entirely lost their natural landscape, they exhibit signs of ecological stress and potential vulnerability.

Red areas (Fair, 0.2–0.4) and pink areas (poor, 0–0.2) are predominantly located in regions with intense human activity, such as cities, ports, and settlements. These areas face significant ecological challenges, including high pollution levels, habitat degradation, and weakened ecosystem functionality.

Geographically, the ecological quality in Johor’s southern and eastern edge regions is highly variable, likely influenced by urban runoff, emissions from aquaculture, or inadequate waste management systems. Central and northern areas are predominantly green, interspersed with small patches of yellow, reflecting low land-use intensity. The western region is mainly yellow, indicating the impact of intensive land use on ecological quality. Additionally, water bodies are often surrounded by red and pink patches, underscoring the environmental pressures faced by these areas.

4. Discussion

In this study, the calculated RSEI values have been influenced by the absence of 2010 Landsat 8 Band 10 data, as shown in Figure 9. To address this issue, MODIS Land Surface Temperature (LST) data were used as a substitute. However, since MODIS data has a resolution of only 1000 m, even after resampling to 30 m, the clustering effect of the data persisted, affecting the spatial accuracy of the distribution. This limitation may have introduced localized errors in the RSEI results, particularly in the densely populated southern and eastern regions, reducing the model’s ability to capture subtle ecological changes. Additionally, the data substitution strategy could have introduced instability in the CA-Markov model’s prediction results, especially when processing data with a 10-year interval, causing fluctuations. This highlights the need to prioritize higher-resolution and temporally complete datasets in future research to improve the accuracy and consistency of RSEI calculations and related predictive modeling.

Through spatial analysis of Johor’s RSEI from 1990 to 2020, significant spatial heterogeneity in ecological quality across the region was observed. The LL clusters were primarily located in the central, northern, and eastern coastal areas, dominated by agriculture and medium-scale industries. These areas exhibited relatively stable ecological conditions with minimal human disturbance. In contrast, the HH clusters were concentrated in the southern urban areas and western coastal industrial zones, reflecting the strong impact of economic activities on regional ecology. The HL and LH outliers were sporadically distributed in the southwestern and northeastern regions, indicating potential for development or localized ecological pressures. This spatial pattern highlights the complex interplay between economic development, land use, and ecological conservation in Johor, underscoring the pronounced regional imbalances in development.

From a temporal perspective, Johor’s ecological quality underwent notable changes over the past 30 years. The area classified as having excellent ecological conditions decreased by 1100 km², while the area with good conditions increased by 752 km². Simultaneously, the area categorized as poor grew from 412 km² to 613 km², demonstrating the environmental pressures brought by urbanization and industrialization. The southern and eastern edge areas displayed uneven ecological quality, likely influenced by urban runoff, industrial pollution, and inadequate waste management. In contrast, the central and northern regions maintained predominantly high ecological quality, possibly due to lower land use intensity and protective measures. These trends reveal that the dynamic changes in Johor’s ecological environment are driven by a combination of factors, including economic activities, land use policies, and resource management practices.

The CA-Markov model’s predictions for Johor’s future ecological quality indicate that increasing the amount of data significantly enhances the model’s accuracy and robustness. However, its limitations remain notable. First, the model relies on a static transition probability matrix, which does not fully account for dynamic factors such as policy adjustments, climate change, and economic activities, potentially leading to insufficient predictions of future complex ecological changes. Second, the use of MODIS data as a substitute in 2010 introduces accuracy issues, particularly in areas with high spatial heterogeneity. Additionally, the computational demands and memory consumption of large-scale datasets limit the model’s efficiency when handling high-resolution data. Enhancing the dynamic adaptability of the model’s transition probabilities and optimizing computational resources will be critical to improving prediction accuracy.

This study adapted and optimized the RSEI with the CA-Markov model, particularly for wetland and mangrove ecosystems in Johor. The introduction of the WET and NDBSI enhanced the RSEI model’s ability to accurately reflect dynamic changes in the local ecological environment, offering a novel methodological framework for similar regions. By integrating remote sensing data with the CA-Markov model, we further improved its spatiotemporal prediction capabilities. The application of a rolling prediction validation method optimized the model’s performance in Johor, demonstrating its effectiveness in capturing spatiotemporal patterns of regional ecological changes. Notably, the proposed model successfully revealed future ecological quality trends in long-term time series predictions, providing a scientific basis for ecological conservation in the region.

Through spatial autocorrelation analysis—specifically the combined use of local Moran’s I and LISA indices—the spatial aggregation patterns of ecological quality in Johor have been identified. This approach effectively delineated priority ecological conservation zones, vulnerable areas, and transitional regions, offering critical decision-making support for regional ecological management. Based on spatiotemporal analysis results, targeted ecological protection and management policies can be formulated, particularly for urbanized areas and industrial belts with poorer ecological quality. Concrete restoration and conservation measures can be tailored to these findings. These policy recommendations hold significant practical value, providing theoretical support for ecological management in Johor and contributing to sustainable development in similar regions.

5. Conclusions

This study utilized the GEE platform to calculate the RSEI for Johor from 1990 to 2020 and applied the CA-Markov model to predict future ecological environment quality. The analysis revealed the spatiotemporal evolution characteristics and trends of ecological quality in Johor. The results indicated significant regional differences in ecological quality over the past 30 years. High-quality ecological areas decreased by 1100 km², while medium-quality areas fluctuated, and poor-quality areas significantly increased, reflecting a notable trend of localized ecological degradation. In spatial dimension, the HH clusters of ecological quality were primarily located in the southern and western regions, characterized by dense urban and industrial activities, while the LL clusters were concentrated in the northern and eastern areas, dominated by low-interference agricultural zones. The scattered HL and LH outliers suggest potential development opportunities in the future. The CA-Markov model predictions demonstrated that the model’s robustness and accuracy improved significantly with increased data volume, as indicated by optimized MSE and R² values. The forecast showed that future high-quality ecological areas would mainly concentrate in central and northern regions, while poor-quality areas would be concentrated in the southern and western regions, where human activities are most intensive. Remarkably, ecological patches near water sources exhibited a higher risk of degradation, warranting particular attention for conservation efforts.