Mineral assemblage records in global copper deposit databases contain information about ore-forming environments and geological context, but analysis is complicated by sparse observations, variable data quality, unequal documentation and overlapping mineralogical signals. Here, we introduce a hierarchical geological context topic model with sparse deviations (HGCTM-S). This model represents global copper deposits through context-conditioned assemblage modes and estimates deposit-level topic attribution by combining geological context with observed mineral-family records. Applied to 1335 deposits from the global copper deposit dataset, with 1205 mineral species grouped into 35 geologically defined families, the model recovered seven modes derived from the dataset. Independent comparison with deposit type labels most strongly supported a Cu–Mo mixed mode enriched in porphyry deposits and a Ni–Co–As mode enriched in magmatic-sulfide deposits; the other modes represented shared sulfide backgrounds, secondary overprints or residual components rather than direct genetic classes. Within the fitted topic-composition component, regional lithology-associated effective deviations were larger than broad age-associated deviations across alternative priors and lithology proxies, although the magnitude varied, and the proxies did not reliably reproduce deposit scale or hosts. Progressive initialization improved aggregate topic stability, but geographic performance was heterogeneous and did not generally exceed a baseline using geological context alone. Therefore, the HGCTM-S provides a way to compare overlapping mineral assemblage components and their regional geological associations rather than a universal classifier or regional predictor.
The integration of machine learning (ML) techniques into mineral prospectivity mapping (MPM) has significantly improved the efficiency of exploration targeting. However, conventional data-driven approaches often insufficiently incorporate geological processes, limiting model interpretability and geological consistency. To address this gap, this study utilized a knowledge–data dual-driven framework for MPM targeting orogenic gold deposits in Guangxi, Southwest China. Geological knowledge of orogenic gold mineral system is explicitly embedded at two levels: (1) feature engineering, with particular emphasis on structural characteristics, especially lineament density, as controlling factor; and (2) training data are formed by positive-unlabeled (PU) samples guided by knowledge. The strategy restricts the selection of negative samples by using structural, stratigraphic and geochemical criteria to reduce sampling bias and enhance geological representativeness. Four representative ML algorithms—random forest, support vector machine, deep neural network, and deep forest (DF)—were systematically evaluated. The results indicated that the DF model achieved the best prediction accuracy, robustness and generalization. Feature importance analysis and SHapley additive explanations revealed that the control sequence is consistent, with structural parameters dominating, followed by geochemical anomalies and stratigraphic units. Independent verification of the strategy in the Youjiang Basin and Qin–Hang Belt—two domains hosting orogenic gold deposits with contrasting geological characteristics—showed that the structural architecture is broadly transferrable, whereas lithological traps and geochemical signatures limit model generalization. Based on the prediction results of the knowledge–data dual-driven DF model, this study identified several new prospective targets in the Youjiang Basin and Qin–Hang Belt, providing clear direction for the next stage of gold exploration. This paper demonstrates that integrating mineral systems knowledge—particularly structural controls and sampling constrains—improves both model interpretability and geological credibility of ML-based MPM, and provides a practical workflow for MPM in structurally complex regions.
Groundwater resources in semi-arid regions face mounting stress, yet the deep structural controls on aquifer geometry remain poorly constrained in many North African sedimentary basins. This study integrated 2D seismic reflection data, four boreholes, Bouguer gravity analysis, and lineament mapping to characterize the subsurface structural framework of the Essaouira Basin, western Morocco and its controls on aquifer geometry and hydraulic connectivity. Bouguer anomaly data (− 65 to + 35 mGal) were processed through polynomial regression to isolate a residual gravity field (− 38 to + 19 mGal), reflecting density contrasts linked to basement topography, salt diapirism, and lateral variations in sedimentary thickness. Improved logistic filter, total horizontal gradient and UC resolved 13 gravity fault contacts (FG1–FG13) with dominant NE–SW, E–W, and N–S trends. The seismic data from five seismic profiles constrained five key regional reflectors: the top Paleozoic basement, the Triassic–Jurassic boundary, the top of the Jurassic aquifer (Oxfordian–Kimmeridgian), the top of the Lower Cretaceous aquifer (Barremian–Albian), and the base of the Upper Cretaceous (Albian–Cenomanian transition) aquifer. The integrated interpretation delineated five negative gravity anomalies (N1–N5) and three structural sub-basins (SB1–SB3) bounded by steeply dipping faults. Salt diapirs imaged on seismic profiles, together with the Jbel Hadid and Jbel Amsittene salt anticlines, act as impermeable lateral seals that compartmentalize aquifer units into hydraulically discrete cells. The Upper Jurassic, Lower Cretaceous, and Upper Cretaceous aquifer units attain their greatest thickness within fault-bounded structural lows, where fracture-enhanced permeability maximizes recharge potential, and fault intersections are identified as priority zones for targeted artificial recharge and strategic borehole siting. The resulting structural synthesis map represents the first gravity-seismic integrated hydrostructural framework for the Essaouira Basin, providing basin scale, quantitative exploration targets and a transferable methodology applicable to comparable semi-arid basins across Morocco and North Africa.
Detecting deep weak mineralization-related anomalies remains challenging because geophysical responses from depth are commonly attenuated, low in amplitude, and obscured by near-surface backgrounds and inversion non-uniqueness. This challenge is further amplified by the scarcity of deep drilling labels and the severe imbalance between mineralized and non-mineralized samples. Here, a label-scarce reliability-aware graph anomaly detection framework is presented for multi-source geophysical interpretation under limited supervision. Three-dimensional inversion voxels are represented as graph nodes, and local spatial relationships are encoded through neighborhood aggregation using a graph sample and aggregate network. A small number of labeled borehole samples is combined with unlabeled anchor nodes, while a weighted binary cross-entropy loss is used to reduce majority-class bias. Monte Carlo dropout is incorporated to estimate predictive probability, standard deviation, and entropy, providing reliability indicators for target delineation. An anchor-graph-based block-wise inference strategy enables scalable prediction over million-scale voxel volumes. Synthetic experiments, ablation tests, and label-ratio analyses show that the graph structure, sample weighting, dropout regularization, and uncertainty-aware inference jointly improve target continuity and reduce missed detections under sparse-label conditions. Application to the Kalatongke G21 Cu–Ni district integrates electrical, magnetic, and gravity inversion models with 906 labeled samples among 2,023,623 voxels. The predicted high-probability anomaly near line 805 at approximately 1.0 km depth is spatially consistent with a gabbroic interval intersected by borehole ZK2017-6. The proposed framework provides probability and uncertainty information for risk-aware deep exploration and offers a reusable strategy for intelligent exploration of complex concealed ore bodies.
Blast-induced flyrock is one of the most critical geo-environmental hazards in surface mining, posing serious risks to human safety, surrounding ecosystems, and nearby infrastructure. Accurate prediction of flyrock distance is essential for effective environmental risk management, sustainable mine planning, and safer blasting operations. This study proposes an intelligent predictive framework based on extreme gradient boosting (XGBoost) integrated with five advanced optimization algorithms: Bayesian optimization, geometric mean optimization (GMO), osprey optimization algorithm, reptile search optimization, and Archimedes optimization algorithm (AOA). A dataset of 252 blasting events collected from the Sungun Copper Mine, Iran, was used to develop and validate the hybrid models. To ensure a robust performance comparison, multiple statistical indicators, radar plots, and Taylor diagrams were employed. The results showed that GMO-XGBoost achieved the best generalization performance and highest prediction accuracy on testing data, demonstrating strong reliability for practical applications, while AOA-XGBoost exhibited superior performance during the training phase. To improve model transparency and engineering applicability, Shapley additive explanations analysis was conducted, identifying powder factor, burden, and stemming as the most influential parameters controlling flyrock behavior, with powder factor showing the strongest effect. The main innovation of this study lies in combining optimized hybrid machine learning with explainable artificial intelligence to move beyond accuracy-based prediction toward interpretable and decision-oriented flyrock assessment. The proposed framework provides practical guidance for blast design optimization, reduces environmental and operational risks, and supports more sustainable mining practices.
With advancement in the exploration and development of tight sandstone gas reservoirs, the classification of these reservoirs has emerged as a critical aspect of research in oil and gas exploration and development. Traditional reservoir classification is fundamentally limited by its reliance on discrete analytical samples, which fails to adequately address the complex heterogeneity in non-cored wells with insufficient core coverage. This is especially true for dual-medium reservoirs, as point-based analysis is insufficient to capture the continuous variation of fractured reservoirs. This paper proposes a novel approach for classifying dual-medium tight sandstone reservoirs using well logging curves, combined with a multi-ensemble learning method. This approach enables the efficient and accurate recognition of reservoir types while capturing high-frequency, continuous data that reflect the vertical heterogeneity of reservoir properties. First, a “three-major and six-minor” classification scheme for dual-medium tight sandstone reservoirs is established based on pore structure. The fracture identification factor is employed to distinguish between fractured reservoirs and matrix-fractured reservoirs. The issue of imbalanced sample data is addressed using ADASYN oversampling methods. Subsequently, an ensemble architecture is developed based on voting theory, integrating optimized base classifiers (RF, SVM, MLP, XGBoost). The results indicate that ADASYN superiorly enhances the recognition of minority reservoir types by strengthening learning near decision boundaries. Compared with hard-voting, the soft-voting ensemble demonstrates greater stability and accuracy, particularly for reservoirs with ambiguous log signatures. Notably, the ADASYN-voting-soft (AVS) achieved the highest mean accuracy of 90.25
With the global shift toward green and low-carbon energy, geothermal resources in abandoned coal mines have emerged as a renewable, stable, and highly valuable energy source. China hosts over 12,000 abandoned coal mines, with projections indicating an increase to approximately 15,000 by 2030. In such environments, mine water is typically heated to 20–50 °C, forming stable low- to medium-temperature geothermal reservoirs. This paper systematically reviews the progress in geothermal development from abandoned mines in China, analyzing fundamental principles, technical approaches, key challenges, and application prospects. This paper identifies mining activities as shaping unique geothermal systems, where mine water is continuously heated by high-temperature surrounding rocks, forming underground thermal reservoirs. Both open- and closed-loop circulation systems are evaluated, highlighting their advantages and limitations. Resource assessment methods, exploration techniques, and numerical simulations are discussed, alongside challenges such as geological complexity, system design, equipment maintenance, and environmental risks. Finally, the paper proposes future research directions focused on technological innovation, multi-energy integration, and policy support to enable efficient commercialization of geothermal energy from abandoned mines.
Traditional drill-core analysis relies on manual observation and laboratory chemical analyses to characterize the distribution of alteration minerals, but these procedures are time-consuming, costly, and subjective. Hyperspectral remote sensing enables rapid, nondestructive, and low-cost acquisition of large volumes of drill-core spectral data. However, most existing mineral identification models are supervised learning models trained solely on labeled spectra, which limits the effective use of abundant unlabeled data and constrains accuracy improvements. To address this, this paper proposes a hyperspectral alteration mineral identification model (HSSRNet) for drill-core hyperspectral data based on self-supervised contrastive learning combined with residual networks. Large-scale unlabeled spectra are first employed for pre-training via contrastive learning and spectral data augmentation, and the pre-trained encoder is subsequently fine-tuned on limited labeled samples to perform high-precision alteration mineral identification. In total, 8386 spectral samples collected from seven drillholes surrounding the Daxigou gold deposit in Chengde, Hebei Province, China, were used to train and evaluate the model. Experimental results demonstrated that HSSRNet effectively leveraged a large number of unlabeled spectra, achieving an overall accuracy of 94.47
The presence of an aquifer significantly affects the strength and bearing stability of rock masses and orebodies, posing serious challenges for geological exploration and mining. To evaluate the damage effect of water immersion on magnetite ore, an ultrasonic metal analyzer ZBL-U5100 was used to measure the wave velocity of magnetite ore before and after water immersion. Then, impact tests were carried out on the specimens using a split Hopkinson pressure bar with immersion time and strain rate as variables. The results showed that water immersion has a negative effect on the dynamic compressive strength of magnetite ore, weakening its brittleness and load-bearing capacity. The dynamic compressive strength and peak strain of the specimen decrease with increasing immersion time, the fractal dimension, and degree of fragmentation increase, and the range of the particle size distribution becomes wider. The magnetite ore with immersion time of 14 d has the highest percentage of absorbed energy and energy dissipation density, which reaches 53.71
Solution gas–oil ratio ( R_s ) is a critical pressure–volume–temperature (PVT) parameter in reservoir engineering, essential for accurate techno-economic evaluations. Traditional laboratory measurements of R_s are costly and time-consuming, while existing empirical correlations often lack generalizability, having been developed for specific crude oil systems. Machine learning (ML) models offer improved accuracy by learning from diverse datasets, but their black-box nature limits interpretability and practical deployment. This study presents a novel white-box long short-term memory (LSTM) model for predicting R_s , trained on 594 PVT data points from oilfields across multiple geographic regions. The model achieved high predictive accuracy with mean absolute error of 0.0905, root mean square error of 0.0175, and coefficient of determination (R2) of 0.9826 on the test set. A 95 R_s while retaining much of the predictive capability of the full LSTM model. To the best of our knowledge, this is the first effort to translate LSTM-learned relationships into an explicit correlation for predicting Rs. The resulting formulation offers a practical balance between predictive accuracy and computational simplicity.
The phenomenon of coal–oil coexistence brings many uncertain risk factors to the mining areas where coal and oil coexist. To conduct a more in-depth exploration of the intrinsic relationship between the active groups of Chaijiagou coal mine (CJG) coal–oil symbiosis (CJG-COS) and small-molecule gas during low-temperature oxidation (LTO), the effects of different crude oil mass fractions at 30–200 °C on the variation patterns of marker gas and active functional groups of CJG-COS sample were investigated through temperature programmed and in situ FTIR. The formation of marker gas and the change of functional groups were quantitatively analyzed for their correlation by gray relational analysis. The results showed that crude oil will significantly increase gas production in the initial stage, then the gas production increase rate slows down. Moreover, the larger the crude oil proportion, the more significant the gas production lag effect. The aliphatic hydrocarbon content of the sample increased by 45.8
The nature of targets extracted through data-driven mineral prospectivity mapping (DD-MPM) is dependent on geodata annotation. However, geodata annotation is noisy because mineral systems are complex systems and the definition of positivity changes with annotator, jurisdiction and time. Consequently, historically annotated data (HAD), which are the exhaustive collection of labeled samples (sites) that have accumulated over the history of exploration, is a noisy and subjective source of ground truth. However, the spatial sensitivity and selectivity of DD-MPM products to annotation is quantitatively unknown. Although HAD are always used at large scales of MPM, the cost-to-benefit analysis of improving and augmenting them is unclear because of inherent limitations of DD-MPM, including: (1) scientific confounds that ironically bias targeting toward the known as opposed to truly greenfield; (2) research design challenges that create circular logic during validation; and (3) geoscience and implementation uncertainty that make DD-MPM products heuristic. We conducted two experiments to: (1) verify the distribution of noisy positive samples in HAD and test whether a minority of samples carry a majority of information; and (2) assess the effectiveness of mineral system constraints by annotating maximally separable mineral systems of a commodity and relating their target overlap with class separability. Using a HAD database containing all copper samples in Canada, we observed that: (1) about 14
Regolith-hosted rare earth element (REE) deposits in South China provide a critical source of heavy rare earth elements (HREEs), yet the geochemical signals that discriminate protolith sources and track REE enrichment remain obscured by intense weathering. Here, the Renju REE deposit in Guangdong Province was investigated using an integrated approach combining zircon U–Pb dating, trace element geochemistry, and unsupervised machine learning. This paper reports three findings. (1) Zircon ages and ternary degree of saprolitization trends reveal that the Renju regolith is a composite of weathering products from Late Cretaceous rhyolite ( 97 Ma, 0–41 m) and Jurassic quartz diorite ( 189 Ma, 44–60 m). Pearson correlation analysis validates that weakly mobile trace elements such as Ti, Th, and V serve as effective indicators for tracing this protolith heterogeneity. (2) Cerium (Ce) is decoupled from all other REEs in the oxidized surface horizon (0–10 m). The pronounced positive Ce anomaly (δCe > 2) in the surface (A horizon) is geochemically complementary to negative Ce anomalies in deeper (B horizon) ion-adsorption-type REE ore bodies, suggesting that the surface Ce anomaly can serve as an indicator for underlying REE mineralization. (3) Gallium (Ga) shows co-migration behavior with light rare earth elements (LREEs), while strontium (Sr) and barium (Ba) correlate with HREEs. Consequently, Ga, Sr, and Ba can serve as practical geochemical pathfinders for LREE and HREE enrichment. This paper demonstrates that accurate identification of protolith heterogeneity is a critical prerequisite for understanding the formation mechanisms of regolith-hosted REE deposits in South China. Moreover, unsupervised machine learning can extract interpretable geochemical signals from high-dimensional weathering datasets, and provides transferable trace element fingerprints for cost-effective exploration of regolith-hosted REE deposits.
Accurate estimation of coalbed methane content is constrained by anomalous diffusion characteristics induced by complex pore networks, while classical Fick's models fail to describe this process. Therefore, the space–time fractional-order derivative was introduced in this study based on the classical Fick diffusion model. The fractal dimension was introduced into the model, and a gas fractal-based sub-diffusion model was constructed. To enhance the model’s engineering applicability, the study derived an approximate analytical solution based on the √(t) model by focusing on the mechanism that induced anomalous mean square displacement of methane molecules (Qt/Q∞ < 0.5). By comparing the fitting data of four models against desorption data, it was found that the fitting effect of the gas fractal-based sub-diffusion model was good (R2 > 0.97), and there was no infinite series, which greatly enhanced the practicability of the model. Moreover, gas diffusion in such anomalous space–time exhibited sub-diffusive behavior, with a rapidly decaying diffusion coefficient and heavy-tailed desorption curves. Coal samples with different degrees of metamorphism exhibited exponential and power-law decay. The study also found that the time fractional derivative (ξ) effectively captures the heavy-tail phenomenon. When ξ → 0, the tailing effect in gas diffusion becomes more pronounced. When ξ equals 1, the diffusion process can be idealized as classical Fick diffusion.
The definition of ore and waste blocks is typically made using run-of-mine cut-off grades, calculated with Lane’s formula. This approach implicitly considers metallurgical recovery and yield; however, it overlooks one important ore property: the quality of the product concentrate. This problem can be overcome by leveraging extensive geometallurgical knowledge and this paper presents an example of the gains that can be achieved by using this information. The study was conducted in a phosphate mine, where the current definition of ore/waste blocks is based on P2O5 and CaO grades. The general solution proposed here is to incorporate geometallurgical information in the cut-off’s decision-making process by applying hierarchical indicator kriging to assign a concentrate category to each block and to apply a neural network to forecast each block yield. This information (yield and block category) is then used to define a block as ore, marginal ore, or waste. Using the proposed approach, the percentage of blocks classified as ore in the bebedourite domain was 61.6
With the continued decline in shallow coal reserves, exploiting deeper coal is now increasingly important for maintaining stable energy supply. China’s deep coal seams contain substantial coalbed methane (CBM) resources, yet the accompanying high geothermal temperatures and strong in-situ stresses markedly reduce gas drainage efficiency. Methane desorption and diffusion in coal, which govern gas migration, are highly sensitive to variations in temperature and pressure. In deep reservoirs, these two processes interact and form a coupled system that modifies methane-release behavior. To investigate this coupled response, this study performed orthogonally-designed desorption experiments under combined temperature–pressure conditions and evaluated the corresponding changes in desorption performance and diffusion coefficients. On this basis, a temperature-dependent correction model for the diffusion coefficient is established. The results demonstrated that increasing temperature decreases the total desorbed amount while accelerating the early desorption rate, whereas higher pressure enhances overall desorption capacity and alleviates temperature-induced inhibition. Compared with raw coal, tectonic coal responds more noticeably to pressure and may display an offsetting interaction between temperature and pressure under specific conditions. Temperature exerts the dominant influence on diffusion-coefficient evolution, and pressure plays a secondary role. These findings clarify the mechanism by which temperature–pressure coupling governs methane release and transport in coal, offering a theoretical basis for enhancing CBM recovery from deep coal seams.
In the fields of 3D geological modeling and geoscience big data analysis, efficiently representing volumetric data characterized by massive scale, non-uniform spatial distribution, and complex geological features remains a core bottleneck constraining the performance of geophysical inversion, geostatistics, and spatial data analysis. To address the challenge of representing large-scale geological volumetric data, this paper proposes and implements a novel hierarchical sparse voxel data structure and its implementation framework—VoxOct. The core innovation of the VoxOct framework lies in its integration of the efficient indexing capability of sparse voxel grid with the flexibility of adaptive spatial partitioning offered by octrees, resulting in a hybrid sparse voxel octree (SVO) structure. VoxOct is a hybrid data structure based on octree spatial subdivision and N‑tree hierarchical compression, designed to achieve the representation, storage, and computation of large‑scale sparse geological volumetric data through the adaptive spatial division. In terms of design principles, VoxOct follows a topology‑attribute‑separated construction approach, which can drastically reduce storage space for invalid or redundant data through pruning optimization strategies. At the algorithmic level, VoxOct designs a set of efficient construction, dynamic traversal, and rapid update algorithms for large‑scale geological volumes. The framework implements an efficient sparse storage scheme and memory management strategy, supports native access to explicit octree nodes, and optimizes fast traversal and dynamic refinement algorithms for massive grids. Through an efficient sparse octree indexing mechanism, the framework enables high‑performance processing of billion‑scale voxel models on general‑purpose computing devices and facilitates dynamic updates to the octree grid model. Experimental results demonstrate that VoxOct reduces runtime memory access overhead compared with pointer-based octrees and requires less storage space than out-of-core octrees, which strikes a balance between the two for large and sparse geological volume data with high-resolution attribute distributions. Currently, the core algorithms of the framework have been integrated into commercial geological application software.
Mineral prospectivity mapping increasingly relies on machine learning to integrate geochemical and geological information, yet purely data-driven models often perform unreliably under sparse sampling, strong class imbalance, and complex geological controls, while traditional knowledge-driven approaches remain difficult to formalize and scale. This paper presents a neuro–symbolic prospectivity framework that incorporates geological knowledge extracted from deposit-model literature using large language models and embeds it into a data-driven learning pipeline. The framework integrates multi-element geochemical data, spatial geological features, and embedding-based geological priors derived from 87 ore deposit models, where the priors encode semantic relationships between mapped geological environments and mineral system concepts and act as soft guidance rather than deterministic constraints. Model performance is evaluated using spatial cross-validation and area-based targeting metrics that reflect exploration decision-making under limited spatial footprints. A case study of the CUMO porphyry Cu–Mo district (Boise County, Idaho, USA) shows that the proposed framework improves early-stage prospectivity targeting, particularly for joint Cu–Mo anomalies and under strong area constraints, while producing spatial prospectivity patterns that are consistent with mapped lithological and structural controls. The results demonstrate how LLM-assisted neuro–symbolic modeling can bridge geological reasoning and machine learning, supporting interpretable and decision-relevant mineral exploration workflows.
Underground coal fires can cause environmental damage and resource wastage. Accurately predicting the combustion state is critical for assessing their hazard levels and implementing control measures. Underground coal fires produce radon gas, which serves as an effective indicator for predicting the coal fire. However, traditional radon measurement is time-consuming and costly. This study proposes a machine learning (ML)-based prediction method that correlates surface radon concentration with atmospheric environmental factors. Four ML models—ridge regression, random forest, gradient boosting, and extra trees—were applied to predict surface radon concentration anomalies caused by underground coal fires. In total, 216 samples were collected from the Haizhou open-pit mine fire area, including four key variables: ambient temperature, ambient pressure, relative humidity (RH), and wind velocity. The results demonstrated that the gradient boosting model exhibited exceptional fit and robust generalizability, achieving R2 of 0.981, RMSE of 170.29, and MAE of 124.69 on the testing dataset. Sensitivity analysis revealed that RH was the most influential factors. The gradient boosting model proved to exhibit high stability and accuracy in predicting radon concentrations over different time durations. The application of this model in radon concentration prediction provides significant support for forecasting underground coal fires and expands new directions for the application of ML in environmental fields.
Due to the environmental degradation caused by population growth, land-use changes, and climate change, the assessment of watershed health has taken on greater importance in recent decades. Though various methods have been proposed for watershed zoning and prioritization, their performance has not been directly compared. Therefore, this study employed the pressure–state–response (PSR) framework to assess watershed health under current and projected land-use and climate change scenarios (SSP5-8.5) and compares its performance with results derived from the Complex Proportional Assessment (COPRAS) and VlseKriterijumska Optimizacija I Kompromisno Resenje (VIKOR) models. Landsat satellite imagery from 2003, 2013, and 2023 was used to generate historical land-use maps, while future scenarios for 2033 and 2043 were forecast using the Land Change Modeler. Climatic, environmental, and land-use variables were integrated into the PSR framework to quantify watershed health under current and projected conditions. The pressure, state, response, and health indices for the baseline period were 0.58, 0.55, 0.40, and 0.49, respectively, while those under climate change conditions in 2033 and 2043 were 0.57, 0.67, 0.40, and 0.52, respectively. Under the influence of land-use changes, the pressure, state, response, and health indices were 0.59, 0.56, 0.40, and 0.49, respectively, in 2033 and 0.59, 0.57, 0.59, and 0.57, respectively, in 2043. Finally, the combined effect of climate change and land-use produced pressure, state, response, and health indices of 0.58, 0.67, 0.40, and 0.52, respectively, in 2033 and 0.58, 0.68, 0.59, and 0.61, respectively, in 2043. Examining changes in health compared to the baseline period (2023) under the influence of climate and land-use changes revealed that the health index would increase by 5.66 and 19.53
