Advanced Chemometric Techniques for Environmental Pollution Monitoring and Assessment: A Review

Abstract

Chemometrics has emerged as a powerful approach for deciphering complex environmental systems, enabling the identification of pollution sources through the integration of faunal community structures with physicochemical parameters and in situ analytical data. Leveraging advanced technologies—including satellite imaging, drone surveillance, sensor networks, and Internet of Things platforms—chemometric methods facilitate real-time and longitudinal monitoring of both pristine and anthropogenically influenced ecosystems. This review provides a critical and comprehensive overview of the foundational principles underpinning chemometric applications in environmental science. Emphasis is placed on identifying pollution sources, their ecological distribution, and potential impacts on human health. Furthermore, the study highlights the role of chemometrics in interpreting multidimensional datasets, thereby enhancing the accuracy and efficiency of modern environmental monitoring systems across diverse geographic and industrial contexts. A comparative analysis of analytical techniques, target analytes, application domains, and the strengths and limitations of selected in situ and remote sensing-based chemometric approaches is also presented.

1. Introduction

Chemometrics represents a robust and systematic methodology that integrates measurement-based analytical chemistry with advanced statistical and mathematical modeling. It plays a pivotal role in analytical chemistry, particularly in optimizing experimental conditions and extracting meaningful chemical information from complex datasets (Figure 1). By applying both qualitative and quantitative analyses, chemometric tools facilitate the exploration of variable relationships using univariate and multivariate techniques. These methods are instrumental in mining data derived from environmental monitoring processes, employing both supervised and unsupervised pattern recognition strategies.

Supervised learning methods rely on predefined output variables, enabling targeted assessments such as sample classification and quality prediction based on known parameters. This approach enhances the precision of environmental monitoring by reducing data dimensionality and improving interpretability. In contrast, unsupervised methods autonomously identify intrinsic data structures by grouping variables or samples based on similarity, uncovering latent patterns without prior knowledge. PCA, FA, and CA are among the most widely used unsupervised methods, particularly suited for multidimensional environmental datasets [1,2,3,4,5,6].

Traditional statistical techniques often fall short in handling the complexity and variability inherent to environmental systems. In contrast, multivariate statistical and geostatistical approaches offer more comprehensive insights. For example, CA has been applied to evaluate spatial correlations among sampling points while PCA and multivariate regression models have been employed at regional scales to discern environmental attributes and pollutant distributions [7,8,9,10].

Figure 2 illustrates the growing significance of chemometric techniques, as evidenced by a marked increase in publications from 2000 to 2024 using the keywords “chemometrics,” “environmental analysis,” and “pollution assessment.” The literature reveals chemometrics as a highly effective tool for identifying pollution sources by integrating faunal structure dynamics, physicochemical profiling, and in situ toxicity analysis. These approaches have proven particularly useful in analyzing large and complex datasets derived from soil and water systems, including assessments of heavy metal contamination [11,12,13].

Unbiased chemometric techniques, such as PCA, HACA, DA, and MLR, have been extensively used to evaluate the quality of air, water, and soil. HACA typically delineates pollution sources into three clusters: low, moderate, and high intensity. DA, applied via forward and backwards stepwise procedures, has helped isolate key variables responsible for environmental degradation. PCA and FA have further confirmed major pollution contributors, including mobile sources (e.g., vehicular emissions), stationary sources (e.g., fossil fuel combustion and industrial operations), cross-boundary pollution (e.g., transnational wildfires and volcanic emissions), and agricultural activities [14,15,16,17,18,19,20,21,22,23]. Mobile air pollution is predominantly generated by transportation systems—such as cars, buses, and trucks—whereas stationary sources arise from industrial activities, power plants, open burning, and food processing facilities [24,25]. Cross-boundary pollution, meanwhile, encompasses airborne contaminants originating from external events such as wildfires or volcanic activity in neighboring regions.

Modern analytical techniques now generate vast datasets related to environmental pollutants in water, soil, and air. However, sustaining ecosystem health requires actionable insights derived from these data. Ecological resilience is increasingly threatened by anthropogenic pressures—deforestation, biodiversity loss, climate change, and unregulated development—which have already led to the degradation of critical habitats such as coral reefs. Nearly 19% of the world’s coral reefs have been lost, with an additional 35% at risk due to ocean acidification, rising temperatures, pollution, and illegal fishing. This decline jeopardizes the livelihoods of over a billion people, nearly 13% of the global population, who live within 100 km of a reef and depend on its biodiversity for sustenance and economic security [26,27,28,29].

Despite the widespread application of chemometric methods, a comprehensive understanding of their core principles and best practices remains limited in existing literature. This review aims to address this gap by providing a critical evaluation of chemometric tools best suited for environmental applications. It also explores potential pollution sources and their implications for human health. Ultimately, this study advocates for the integration of chemometric frameworks into modern environmental monitoring systems, enabling more responsive, accurate, and scalable assessments in both scientific and industrial domains.

2. Environmental Monitoring and Assessment

Environmental monitoring is fundamental to assessing and managing the quality of natural ecosystems. Chemometric techniques play a crucial role in this domain by enabling the efficient handling of large, complex datasets, uncovering hidden patterns, and facilitating accurate identification of pollution sources. These data-driven approaches are indispensable for interpreting the growing volume of information generated by modern environmental monitoring systems.

Monitoring involves systematic sampling and analysis of environmental media—such as air, water, soil, or integrated ecosystems—to detect changes in contaminant levels over time and space. Effective monitoring requires well-defined sampling strategies and robust analytical protocols that account for both internal variability and external environmental factors. As technologies evolve, the scale and complexity of data increase, necessitating the adoption of advanced statistical and computational tools for data analysis.

At its core, environmental monitoring includes measuring key analytes, identifying pollution sources, and communicating findings. However, comprehensive monitoring is often resource-intensive, involving time-consuming and labor-intensive sampling protocols. These challenges underscore the need for streamlined and automated analytical methods that can deliver reliable insights in real-time or near-real-time settings [30].

2.1. In Situ Monitoring

In situ chemometrics refers to the application of statistical and mathematical models directly at the point of environmental data acquisition. This real-time or on-site analysis approach minimizes sample handling, accelerates decision-making, and improves the accuracy of environmental assessments. Recent advances in in situ monitoring technologies have significantly expanded the capabilities of environmental surveillance, offering fast, cost-effective, and scalable solutions across diverse ecological systems [31,32,33,34,35]. Another innovative technique employed planar microwave sensors for continuous in situ monitoring of water quality in freshwater systems, particularly in mining-impacted areas. These sensors operate by emitting microwave signals and analyzing shifts in resonant frequencies within the GHz range. Changes in resonance indicate alterations in water composition, which correlate with the presence of trace metal pollutants. Field deployments in four contaminated sites in the UK confirmed the system’s sensitivity and responsiveness to metals such as lead, cadmium, arsenic, and mercury. This technology provides a rapid, low-cost, and reliable method for continuous monitoring of freshwater pollution, enabling early detection and timely environmental interventions [32].

One notable application involved the development of an in-situ Raman spectroscopy method for monitoring salt disproportionation in pharmaceuticals. This approach enabled the real-time quantification of polymorphic forms, including two metastable and one stable free base, from the disproportionation of two salt types. Unlike earlier studies that utilized pre-characterized polymorph mixtures, this method analyzed evolving polymorph compositions in situ, providing deeper insight into polymorphic transitions during the disproportionation process [31,35,36,37,38]. Bojko et al. [35] investigated the application of in situ SPME on Mediterranean marine sponges for untargeted exometabolomic profiling. This technique enabled the capture of chemical signatures directly from the sponge–environment interface, revealing rich chemical information that reflects both biological processes and environmental stressors. The work highlighted the effectiveness of SPME in minimally invasive, real-time environmental surveillance. Parallel progress has occurred in pharmaceutical sciences, where in situ spectroscopic techniques are increasingly employed to investigate solid-state transformations, particularly salt disproportionation during drug dissolution and storage. Nie et al. [36] utilized Raman spectroscopy and mapping to study salt-to-free-base conversions within tablet matrices. Their spatially resolved data revealed significant heterogeneity in conversion patterns, underscoring the importance of Raman tools in stability assessment and formulation optimization. Previous research has integrated NIR imaging with Raman mapping to investigate the disproportionation of drug HCl salts during dissolution. Their combined approach enabled real-time visualization of phase changes, providing a deeper understanding of how these transformations affect drug release and solubility [37]. Ewing et al. [38] employed ATR-FTIR imaging alongside Raman mapping to study salt disproportionation under hydrated conditions. Their results emphasized the spatial and temporal evolution of salt forms during drug delivery, providing crucial insights into formulation performance and release kinetics. In the same context, Nie et al. [39] compared backscattering and transmission Raman spectroscopy methods for quantifying solid-state form conversions in pharmaceutical tablets. The authors found transmission Raman spectroscopy to be more accurate for analyzing internal structures due to its better penetration depth and reduced surface bias. This work underlines the importance of selecting suitable Raman modalities for in situ pharmaceutical analyses.

A summary of recent in situ chemometric approaches—including their targeted analytes, analytical methods, environmental applications, advantages, and limitations—is provided in

Table 1. The focus, analytical methods, analytes, application, advantages, and limitations of some in situ chemometric methods are reported in the literature.

Focus	Chemometric Approach	Analytical Method	Analytes	Accuracy/Precision & R2	Application	Advantages	Limitations	Ref.
Monitoring of polymorphic transitions of pharmaceutical compounds	PCA, PLS	In situ Raman spectroscopy and X-ray diffraction	API salts (HCl and maleate)	For HCl salt, RMSECV = 0.056, 0.034, and 0.022 For maleate salt, RMSECV = 0.016 and 0.023	Real-time monitoring for quality control and stability assessment	Detects multiple polymorphs to enhance calibration for complex systems	Requirements of complex calibration	[31]
Real-time monitoring of trace metal contamination	PLS	Microwave spectroscopy and planar sensors	Trace metals (e.g., Pb, Cd, As, and Hg)	R2 > 0.96	Continuous environmental monitoring of water quality	Immediate response, in situ monitoring, and cost-effective	Non-specific metal detection	[32]
In situ SPME for untargeted exometabolome screening	PCA, PLS, DA	GC-MS/LC-MS	Primary (amino acids and fatty acids), secondary (alkaloids and terpenes), and pollutants	80.92–100% p < 0.05	Marine health monitoring, chemical ecology, and natural product discovery	Non-invasive and eco-friendly with minimal contamination	Specific for polar compounds	[35]
Detection of salt disproportionation	PCA, PLS	Raman spectroscopy and X-ray diffraction	Pioglitazone HCl salt	RMSECV = 0.45 R2 = 0.996	Quality control and stability assessment	Detects minor species, provides spatial distribution, and enhances sensitivity	Spectral interference requires advanced techniques	[36]
Disproportionation of the drug HCl salt	NA	NIR and Raman spectroscopy	Avicel and API salt	NA	Formulation stability, drug release, and efficacy	Non-destructive, spatially resolved monitoring, and better process understanding	Requires advanced data analysis	[37]
Disproportionation of drugs	NA	ATR-FTIR and Raman spectroscopy	Avicel and API salt	NA	Particularly for salt-based drug delivery systems, drug solubility, and bioavailability	High spatial and chemical specificity	Requires complex data interpretation and spectral deconvolution	[38]
Solid-state form conversion within intact pharmaceutical tablets	PCA, PLS	Raman spectroscopy	Pioglitazone hydrochloride salt	RMSECV = 0.506, 0.837 R2 = 0.928, 0.981	Process monitoring, quality control, and regulatory	Reducing surface bias, non-destructive, and rapid	Requires careful calibration and uniform sample density	[39]

.

2.2. Remote Monitoring

Remote monitoring refers to the use of advanced technologies to collect, analyze, and interpret data from distant locations without the need for physical presence. This approach is particularly valuable for observing expansive, difficult-to-access, or hazardous environmental systems. It has found widespread applications in diverse sectors, including environmental surveillance, industrial process control, healthcare, agriculture, and security. In the context of environmental monitoring, remote technologies enable continuous assessment of parameters such as air and water quality, deforestation, and indicators of climate change. These systems utilize a suite of tools, including satellites, UAVs, sensor networks, and IoT devices, to facilitate real-time and long-term data acquisition across terrestrial and aquatic environments [40].

Table 2. Brief descriptions and applications of key technologies for remote environmental monitoring.

Technology	Description	Example	Ref.
Satellites	Use optical, thermal, or radar sensors to observe Earth from space	Monitoring deforestation, glacial retreat, and land use changes	[41,42]
Drones (UAVs)	Provide high-resolution aerial images and multispectral data	Mapping crop health, detecting oil spills, and wildfire monitoring	[43,44]
IoT Sensors	Ground-based devices connected via wireless networks	Air/water quality sensors in urban areas or rivers	[45]
WSNs	Networks of sensors transmitting environmental data	Forest fire detection and weather stations	[46]
Buoys and Floating Platforms	For marine environments	Ocean temperature, salinity, pH, and wave height monitoring	[47]

provides an overview of key remote sensing technologies and their typical environmental applications [41,42,43,44,45,46,47].

Satellite-based remote sensing plays a vital role in environmental monitoring by providing extensive spatial and temporal coverage for assessing variables such as air quality, water pollution, land use changes, and vegetation dynamics. The integration of satellite retrievals with chemometric techniques enables advanced, scalable, and cost-effective environmental monitoring solutions. These tools are particularly valuable in regions with limited ground infrastructure and have enormous potential in addressing global climate and sustainability challenges. The data, however, is often high-dimensional and complex. Chemometric techniques, which encompass multivariate statistical methods and machine learning tools, are essential for analyzing these complex datasets, identifying patterns, and making precise predictions [48]. Satellite retrieval refers to the process of deriving geophysical or environmental parameters (e.g., aerosol optical depth, chlorophyll concentration, and land surface temperature) from measurements made by satellite sensors. Popular satellite sensors include MODIS, TROPOMI on S5P, the Landsat series, and the VIIRS. These sensors provide raw radiometric or spectral data, which are processed through retrieval algorithms to extract quantitative environmental indicators. To interpret and model these retrievals, chemometric techniques are widely used. ATBM is highly influential in integrating geometric and radiometric information that corresponds to tensor power iteration and image detection. ANN helps to reduce data dimensionality while preserving variance. It is often used to enhance efficiency and computational complexity in power iterations [49]. GWR is employed to estimate surface-level pollution concentrations, such as PM_2.5, using satellite-derived AOD. The GWR forecasters also included replicated aerosol composition and information on land use [50]. GOSAT has evolved into a widely used method for detecting CH₄, which has been enhanced by the XGBoost algorithm. The model yielded an RMSE value of 0.0251 µg/L and a cross-validation coefficient (R²) of 0.79, along with an MAPE of 0.88%. This model is also capable of apprehending complex nonlinear relationships between satellite column concentrations and their influencing elements [51]. Supervised classification methods such as SVM and RF are applied for vegetation health monitoring, land use categorization, and water quality assessment [52]. For example, Van Donkelaar et al. [50] developed a hybrid geophysical-statistical model combining satellite AOD data with chemical transport models and ground-based observations to generate global estimates of PM_2.5. Gholizadeh et al. [48] demonstrated the application of chemometrics in analyzing water quality by using PCA and ANN to interpret MODIS data for parameters such as chlorophyll-a and turbidity. The satellite types, along with chemometric methods for environmental monitoring, are presented in

Table 3. Satellite types and chemometric methods for environmental monitoring.

Satellite/Sensor	Environmental Parameters Retrieved	Common Chemometric Methods	Applications	Ref.
MODIS	CDOM, SDD, TSS, TP, SSS, DO, BOD, COD	PCA, ANN	Evaluating and quantifying the water quality	[48]
ATBM, UAV	Geometric and radiometric information	ANN	Tensor power iteration and detection process	[49]
MODIS, MISR	AOD	GWR	PM2.5 estimation	[50]
GOSAT	CH4	XGBoost	Atmospheric profiling and greenhouse gas monitoring	[51]
Sentinel 2A	LULC, NDVI	SVM, RF	Water quality assessment, agricultural monitoring, and land cover change detection	[52]

. Despite their effectiveness, these methods encounter challenges such as atmospheric interferences, sensor calibration errors, and the need for adequate ground-truth data for validation. The integration of chemometric techniques with machine learning and data fusion approaches (e.g., combining satellite, UAV, and ground sensor data) is a growing trend to improve accuracy and spatio-temporal coverage.

A wide range of sensor types, as follows, is employed in remote environmental monitoring, often in conjunction with chemometric techniques to enhance data interpretation and predictive modeling:

Optical sensors: Utilizing UV–Vis, NIR, and Raman spectroscopy, these sensors are employed for both qualitative and quantitative analysis of chemical constituents in environmental samples, with a primary focus on monitoring water quality. It is highly recommended to estimate physicochemical variables, such as dissolved organic carbon, nitrate, and turbidity, which also improve measurement absorbance accuracy [53].

Electrochemical sensors: Commonly applied for detecting heavy metals, pesticides, and organic pollutants in soil and water through redox and ion-selective mechanisms. The sensor has been effectively utilized to study the cytotoxicity of heavy metals such as Cd²⁺, Cr⁶⁺, Cu²⁺, Pb²⁺, and Zn²⁺ using eukaryotic cells that are highly sensitive and cost-effective [54].

Biosensors: Engineered for high specificity, these sensors target biological contaminants, including pathogens, toxins, and microbial pollutants. It also offers numerous benefits over orthodox analytical techniques, including portability, sensitivity, ease, affordability, selectivity, and the capability to measure the contaminated water’s toxicity [55].

Gas sensors: Utilized for ambient air monitoring, particularly for gaseous pollutants like NO₂, O₃, and VOCs. The gas sensor based on nanostructures can enhance detection performance by employing nanoparticles and carbon materials as a composite. However, SnO₂ is used on a large scale as an active material for commercial purposes in the detection of various gases. Alternatively, researchers continually explore new sensing materials that exhibit higher selectivity and sensitivity at low operating temperatures, a fast response, high recovery competence, and replicability compared to commercial sensors for detecting gases, while maintaining the elasticity and toughness required for environmental gas sensors [56].

Electronic noses and tongues: These sensor arrays mimic human olfactory and gustatory systems to detect and differentiate complex chemical mixtures in air and water environments [57].

The integration of these sensor platforms with chemometric algorithms significantly enhances the ability to interpret large, multidimensional datasets enabling robust assessments of environmental quality and more accurate identification of sources. As sensor sensitivity and communication technologies advance, remote chemometric monitoring is expected to play an increasingly pivotal role in real-time environmental decision-making and early-warning systems.

Table 4. The focus, analytical methods, analytes, application, advantages, and limitations of some selected remote monitoring chemometric methods reported in the literature.

Focus	Chemometric Approach	Analytical Method	Analytes	Accuracy/Precision & R2	Application	Advantages	Limitations	Ref.
Development of non-destructive, rapid, and accurate method	HVIs, PCA, PLSR	UV–VIS-NIR-SWIR	Chlorophylls, carotenoids, flavonoids, and lignin	R2 > 0.75 p < 0.01	Real-time plant health monitoring and breeding ecophysiological studies	Non-invasive, Rapid, and high throughput	High hyperspectral equipment cost, and complex data analysis	[58]
Estimate the concentrations of heavy metals in soil	PLS, PCA	NIRS	Heavy metals (Cd, Cu, Pb, Ni, Cr, Zn, Mn, and Fe)	RMSEP = 9.63, 11.5% R2 = 0.86, 0.58	Assessing soil contamination	Non-destructive, rapid, cost-effective, and high throughput	Calibration dependency, matrix effects, and limited analyte scope	[59]
Detection of highly polar pesticide residues	OPLS, PCA	LC-MS/MS	50 medium to highly polar pesticides	R2 = 0.49–0.73	Monitor sediment contamination	High sensitivity and selectivity	Requires advanced equipment and technical skill	[60]
The study evaluates the use of plant-based ingredients	PLS, PCA, DA	FTIR, UV–Vis, Raman, GC-MS	Plant-derived ingredients and microbial counts	100, 99.8, 99.6, 96.6 and 93.7% RMSEP-1.1% R2 = 0.97	Useful in the industry for producing reduced fat and natural preservation	Reduces fat content, and enhances safety and shelf life	Limited details on the specific plant ingredients	[61]
Improving water quality assessment	OPLS, PCA	Flow cytometry	50 medium to highly polar pesticides	R2 = 0.49–0.73	Early detection of environmental changes, pollution events, and ecosystem health monitoring	Rapid, real-time tracking, and high-sensitivity	Requires specialized equipment and expertise, high cost, and complex data interpretation	[60]
Assessing groundwater quality	PCA	WQI, SAR, EC, UV–Vis, IC	Ca, Mg, and Cl	WQI = 17–47% EC = 17–64%	Water quality assessment	Identification of patterns in water quality data	Depending on the quality of the data available	[62]
Heavy metal contamination in the groundwater	HCA, PCA	AAS	Fe, Mn, Pb, Cd, Cr, and As	34.21–82.97% p < 0.05	Assessing health risks and guiding safe water	Identifies health-threatening pollutants	It does not cover all possible contaminants.	[63]
Assessing human pharmaceuticals in water	CA, PCA, DA	LC-MS/MS	Nineteen pharmaceuticals	100% R2 > 0.75 p < 0.05	River water quality monitoring	Identifies pharmaceuticals	Focus limited to selected pharmaceuticals	[64]
Evaluating sediment quality	PCA, DA, PLS, ANN	NA	Heavy metals and polycyclic aromatic hydrocarbons	92.3–97.2%	Broad use in environmental analysis	Enhances accuracy, and reduces experimental trials	Requires high-quality, representative data	[65]
Quantifying inorganic arsenic species	DA, ANN	ETAAS	As(III) and As(V)	92–98% 88–91%	Assessment of arsenic contamination	High selectivity and sensitivity	It requires careful pH control and multiple extraction steps.	[66]

summarizes key remote monitoring strategies involving chemometrics, outlining their focus areas, analytical methodologies, target analytes, advantages, and limitations as reported in the recent literature.

3. Types of Chemometric Techniques

Chemometric techniques encompass a suite of mathematical and statistical methods designed to extract relevant insights from complex chemical datasets. These techniques are particularly valuable in fields such as spectroscopy, chromatography, environmental science, and food analysis. Chemometric approaches can be broadly categorized into unsupervised learning, supervised learning, regression analysis, correlation techniques, and time series analysis, depending on the nature and purpose of the studies. Different approaches are typically used with large datasets, dividing them into training or test sets and validation sets. In contrast, when datasets contain fewer samples, resampling algorithms are favored to split the data, assuming that the data consumed by these algorithms, popular for CV, is also representative of future data. The CV divides the data into k-folds where one-fold is used as a validation set and the remaining k-1 folds are used to build the model. These are usually repeated k times to achieve them as a validation set, and the predictive accomplishment is estimated as the average representative and independent datasets (Figure 3).

3.1. Unsupervised Learning Methods

Unsupervised learning methods are primarily employed for pattern recognition, exploratory data analysis, and dimensionality reduction, especially when class labels are unknown. These methods play a central role in extracting hidden structures and trends from complex, high-dimensional datasets in chemometrics.

3.1.1. Cluster Analysis

CA is a fundamental unsupervised chemometric technique used to identify natural groupings or patterns in multivariate datasets. In chemical and environmental sciences, CA is frequently applied to spectral, chromatographic, and environmental data to group samples with similar chemical profiles without requiring predefined class labels.

For instance, El-Rawy et al. [3] employed an integrated chemometric approach that combined PCA and HCA to assess groundwater quality using 180 samples. The SAR classification indicated that 91.67% of the samples were suitable for irrigation, with no adverse effects related to sodium in the groundwater samples, as indicated by an R² value greater than 0.8. The study identified key hydrogeochemical processes, such as evaporation, seawater intrusion, and ion exchange, and grouped sampling sites based on water quality similarities, producing a spatial distribution map that highlights areas unsuitable for irrigation.

In another study, Adejuwon et al. [68] utilized PCA, HCA, and FA to evaluate surface water quality in a lake system. Chemometric analysis enabled the identification of pollution sources, primarily municipal runoff and domestic wastewater, by clustering sampling locations and interpreting factor loadings. The study indicated that higher values of TSS, DO, etc., were observed in the summer season compared to other seasons and showed a strong correlation among them, notably between phosphate and total hardness (r = 0.978) in the wet season and between pH and temperature (r = 0.995) in the dry season.

Similarly, Curcic et al. [69] applied PCA and CA to monitor pesticide residues in surface waters using HPLC. The study effectively tracked the spatial and temporal contamination patterns of organic micropollutants across two river systems with a primary focus on the presence of dimethachlor and its metabolites, such as dimethachlor ethanesulfonic acid and dimethachlor oxalamic acid, in surface water. The accuracy of the method was determined using RMSE and R² values, which ranged from 0.569 to 1.242 and from 0.997 to 0.998, respectively.

An innovative approach by Tinnevelt et al. [60] employed high-resolution flow cytometry data analyzed through a novel chemometric framework, DAMACY. The investigation successfully predicted all the nutrient variables including NO₂, NO₃, etc., with R² values ranging from 0.49 to 0.73, indicating that these variables have a significant influence on phytoplankton cells. This integrated PCA and discriminant analysis to track phytoplankton community shifts in real-time, providing early warning indicators for aquatic ecosystem disturbances.

3.1.2. Artificial Neural Networks

ANNs are powerful machine learning models that are increasingly adopted in chemometrics to address nonlinear, multivariate problems that traditional linear models (e.g., PCA and PLS) may not adequately resolve. ANNs are particularly effective in uncovering hidden structures, reducing data dimensionality, and facilitating predictive modeling in environmental systems (Figure 4).

Nurhayati et al. [71] demonstrated the application of ANN for real-time monitoring of DOC in wastewater using fluorescence and UV absorbance data. The model achieved high predictive accuracy (R² = 0.9079; RMSE = 0.2989 mg/L), positioning ANN as a viable alternative to standard DOC measurement protocols.

ANNs have also been applied to UV–Vis spectroscopic data to estimate nutrient concentrations—such as nitrogen and phosphorus—as well as COD and suspended solids in eutrophic rivers, outperforming traditional regression methods [72]. Further, ANN models have been employed to predict trihalomethane formation in drinking water networks using diverse water quality parameters, achieving high prediction accuracy and enabling proactive water treatment management [73]. Additional studies (Figure 5) have highlighted the effectiveness of ANN in air quality forecasting [74] and in predicting the degradation behavior of pharmaceutical contaminants in aquatic systems, with strong experimental validation [75,76]. The use of AI and ML techniques for the simulation process is becoming increasingly widespread in studying atmospheric properties related to environmental issues, such as pollution, which is primarily concentrated among primary pollutants, including gaseous and PM. Therefore, it is necessary to accumulate atmospheric data based on meteorological variables, which include PM₁₀, T_T, T_T−1, T_Mean, RH_T, RH_T−1, RH_Mean, WS_T, WS_T−1, WS_Mean, PR_T, PR_48H, PR_72H, and day as the consecutive day on which the data was recorded. The model successfully simulated PM₁₀ concentrations between 10 μg/m³ and 50 μg/m³, which is satisfactory, especially considering that these concentrations comprise most of the recorded data in the area [75]. The QSAR model, based on multiple linear regression, plays a significant role in removing pollutants from water. The model utilizes MDs to elucidate the contaminant’s physicochemical properties, including energy difference, electron affinity, halogen atoms, and ring atoms [76]. The study revealed that pH, DOC, and alkalinity could influence the QSAR model to determine the water quality.

A comparative summary of ANN-based models applied in environmental monitoring is provided in

Table 5. Comparative overview of ANN techniques’ performance in some environmental monitoring studies.

Environmental Target	Input Data	ANN Role	Performance Highlights	Ref.
DOC in Wastewater	Fluorescence intensity and UV absorbance	Quantification of DOC	R2 = 0.9079; RMSE = 0.2989 mg/L	[71]
Nutrient and COD levels in Rivers	UV–Vis spectral data	Estimation of N, P, COD, and SS	ANN outperformed regression models	[72]
THMs in Water	Physicochemical water parameters	Prediction of THM levels	High accuracy vs. traditional models	[73]
PM10 in the Air	Meteorological variables	Forecasting air pollution	R2 = 0.81; RMSE = 7.40 µg/m3	[75]
Pharmaceutical Degradation in Water	Molecular descriptors	Predicting optimal degradation techniques	Experimentally validated predictions	[76]

, highlighting their focus, input variables, performance metrics, and application areas.

3.2. Supervised Learning Methods

Supervised learning methods are employed when both input variables and corresponding output classes (labels) are known. These techniques are instrumental in classification tasks, predictive modeling, and pattern recognition, especially when prior knowledge of group membership is available. One of the most widely used supervised chemometric tools in environmental analysis is DA.

Discriminant Analysis

DA is a powerful multivariate technique widely used in environmental monitoring for classifying and differentiating among predefined groups—such as pollution sources, sampling locations, or ecological conditions—based on observed variables, including chemical concentrations or spectral features. DA constructs linear or nonlinear functions that maximize the separation between classes, facilitating both classification and interpretation of key discriminating variables.

Common forms of DA include the following:

Linear Discriminant Analysis: Assumes homogeneity in class covariance matrices; optimal for linearly separable data with equal variance.

Quadratic Discriminant Analysis: This method allows each class to have a distinct covariance structure, providing flexibility when class variances differ.

Partial Least Squares Discriminant Analysis: Integrates dimensionality reduction with classification, particularly useful in handling multicollinearity and noisy datasets. Compared to more complex nonlinear models such as ANNs, DA is computationally efficient, provides interpretable decision boundaries, and yields insights into variable importance. It is often used in combination with other chemometric tools to enhance classification accuracy and interpretability:

PCA–DA: PCA is used to reduce dimensionality before applying DA to classify the resulting components.

PLS–DA: Combines PLS regression and DA to model the relationship between predictor variables and categorical outcomes.

CA with DA: Clustering methods may first identify natural groupings in unlabeled data, which are then validated or refined using DA.

The integration of DA with these complementary techniques enhances its utility in complex environmental monitoring scenarios where spatial, temporal, and multivariate data are prevalent. A comparative summary of DA applications in environmental monitoring, including study focus, classification performance, and target analytes, is presented in

Table 6. Comparative overview of the DA techniques used in some environmental monitoring studies.

Environmental Matrix	Chemometric Approach	Type of DA	Key Discriminating Variables	Accuracy/Precision & R2	Main Application	Ref.
Tea leaf samples	Multielemental analysis + chemometrics	LDA	As, K, La, and Pb	98.9%	Discriminating tea origins based on geochemical fingerprint	[77]
Surface water (lake)	Physicochemical analysis + DA	Not specified	pH, EC, BOD, and TDS	98.5–100%	Assessment and classification of water quality	[68]
Groundwater	Hydrochemical analysis + multivariate statistics	Not explicitly DA-only	Major ion such as Ca2+, Mg2+, Cl−, NO3−, etc.	R2 = 0.62–0.96	Characterizing groundwater facies and pollution sources	[78]
Phytoplankton (flow cytometry)	DAMACY algorithm (DA-based)	LDA-based (with anomaly detection)	Cell size, fluorescence, and scattering	R2 = 0.49–0.73	Real-time monitoring and early pollution detection	[60]
Teff grain	Multielemental ICP-OES + chemometrics	LDA	Fe, Mn, Zn, Ca, etc.	96%	Origin authentication of grains from different zones	[79]

.

3.3. Factorial Methods

Factorial methods are a family of multivariate statistical techniques used to reduce data dimensionality, identify latent structures, and highlight patterns in complex environmental datasets. Several factorial methods are crucial, including exploratory factor analysis, confirmatory factor analysis, correspondence analysis, multiple correspondence analysis, and PCA. Among them, PCA is widely used in environmental chemometrics to simplify complex datasets, identify fundamental patterns, and enhance monitoring techniques. In chemometrics, factorial methods are valuable for analyzing data from water quality assessments, groundwater contamination analyses, soil and agricultural monitoring, and air quality monitoring [78,80,81,82,83,84]. A comparative analysis of factorial method applications in environmental monitoring is summarized in

Table 7. A comparative analysis of factorial method applications in environmental monitoring is summarized.

Environmental Matrix	Main Objective	Factorial Method	Key Findings	Chemometric Contribution	Ref.
Surface water (river)	Assess pollution sources and seasonal changes in water quality	PCA	Identified major pollution sources; separated seasonal trends	PCA revealed key influencing parameters (DO, BOD, NO3−, etc.) and anthropogenic vs. natural impact	[80]
Surface water (wadi)	Evaluate spatial variation in water quality in Wadi Hanifa	PCA	Explained 85% of total variance with 3 PCs; salinity and nutrients were the main drivers	PCA helped classify water quality zones and contamination levels	[81]
Groundwater (plain)	Characterize hydrochemical processes and pollution sources	PCA	Identified geochemical processes: silicate weathering, evaporation, and salinization	PCA simplified hydrochemical data into manageable PCs for interpretation	[78]
Sediment core samples	Identify brine layers and geochemical change points	PCA + Change-Point Analysis	Revealed stratification and historical geochemical transitions	PCA reduced complexity; change-point detection linked transitions to salinity	[82]
Soil sample	Determine the pattern of soil moisture and apparent electrical conductivity	PCA	Provided insights into controlling factors and the major soil water changing aspects responsible for the soil moisture spatial pattern	PCA accounts for 86% of the total dataset’s variance, and all are significant in illustrating the spatial association between the topsoil and its sequential variations in soil moisture.	[83]
Air quality data sample (several monitoring stations)	Air quality changes in terms of air pollution	Variance for functional data (FANOVA).	Significant reduction of NO2 but increased PM10 and P2.5 in the lockdown period	FANOVA analysis was feasible, allowing for the comparison and rejection of the null hypothesis of impartiality for mean functions of all contaminants	[84]

.

Principal Component Analysis

PCA is a foundational unsupervised chemometric technique widely employed for data exploration, dimensionality reduction, and pattern recognition. It transforms a dataset containing potentially correlated variables into a new set of orthogonal variables known as principal components. These components are ranked by the amount of variance they capture, with the first few typically preserving most of the dataset’s variability. This transformation enables the simplification of complex, multivariate data while retaining essential structural information [85].

PCA plays a critical role in chemometrics by uncovering latent patterns, identifying relationships among variables, and reducing data complexity, thereby enhancing interpretation and decision-making. For example, Younes et al. [86] applied PCA to evaluate the efficiency of different aerogel materials in ion removal processes. The analysis helped identify the most influential factors affecting performance, which are pH, porosity, Brunauer–Emmett–Teller surface area, and density, supporting optimal material selection for water treatment applications. Numerous recent studies highlight the versatility of PCA across various scientific domains [87,88,89,90,91]. PCA was used to estimate the specific area size and variation in environmental resonant measurements, including carrying capacity related to resource amount, ecological quality, social economy, and infrastructure composition. It is a beneficial tool for appraising ECC that also reveals the spatial distribution of ECC indices on a large scale. The study shows that the ecological quality carrying capacity is the lowest, followed by the infrastructure composition carrying capacity. The socio-economic carrying capacity, and ultimately the amount of resources, depend on advancing environmental quality and improving resource utilization efficiency in a region [87]. The spatial distribution PCA model is quite satisfactory, as it effectively links the spatial features of soil pollution with data that produces a linear relationship. Applying eigenvector-based PCA, it efficiently evaluates soil pollution in three dimensions, making it easier to identify the potential sources of heavy metals. Agriculture was the major source (65.5%) of soil pollution contributing to the deposition of numerous heavy metals. The other sources, including traffic (17.9%) and natural pollution (11.1%), also contributed to soil pollution [88]. However, multicollinearity between multiple variables can often delay interpretation, which can significantly impact environmental health research. Nevertheless, Supervised Principal Component Analysis can easily solve this issue [89]. The procedure did not account for any potential nonlinear relationship between exposure and response, which may decrease the residual confounding risk by including nonlinear and interaction terms. The SPCA is efficiently adopting the computational and interpretable system to explain collinearity and exposure misclassification, which is inevitable. Predictably, if there were a factual correlation with the health outcome, the consequence evaluations could be biased towards being insignificant due to standard random errors. The PCA analysis is also effective in determining water quality, which is highly significant for human health. PCA modeling revealed that the primary source of anthropogenic pollution is attributed to the five main PCs. In open dumping sites, groundwater is predominantly contaminated with microbial and heavy metals [90]. The PCA and RF algorithm, combined with LCA, is a powerful tool for measuring the environmental effects of numerous solid waste disposal technologies and guiding ecological protection efforts. Multiple methods of LCA–PCA–RF modeling identified the leading cause of environmental loads as human toxicity potential, resulting from the accumulation of large amounts of dioxins and PM_2.5 formed during the carbothermal reduction and oxygen-enhanced side-gusting of inorganic waste disposal [91]. Therefore, encourage the collaborative consumption of solid waste in all regions, both indoors and outdoors, and reinforce environmental management to achieve zero pollution.

Beyond its core applications, PCA offers several advantages that significantly enhance chemometric analyses:

Dimensionality reduction: PCA condenses large datasets into a smaller number of variables (principal components) while preserving the majority of the variance. This simplification improves data visualization and interpretability, particularly for high-dimensional spectroscopic or chromatographic data [92].

Noise reduction: By focusing on components that explain the most variance, PCA helps filter out experimental noise and emphasizes meaningful patterns, improving the reliability of downstream analyses [93].

Outlier detection: PCA is effective for identifying anomalies within datasets, as deviations from the main data structure become more visible when projected onto the principal component space. This capability supports quality control and error detection in analytical workflows [94].

PCA’s adaptability enables its broad application in various chemometric domains, including food authentication, environmental monitoring, and pharmaceutical analysis. In environmental chemistry, PCA helps interpret large datasets to assess pollution levels and identify sources of contamination. In food science, PCA is used to detect adulteration, verify geographical origin, and monitor product quality—for example, distinguishing between apple cultivars, monitoring beer aging, and identifying adulteration in Camellia oils [95]. In pharmaceutical applications, PCA, combined with techniques such as near-infrared spectroscopy and hyperspectral imaging, has enabled the detection of counterfeit drugs and the identification of substandard formulations, contributing to enhanced product integrity and quality assurance [96].

The chemometric approaches that assess the quality of the predictive model are referred to as validation, which requires a diagnostic metric that explains the variance or residual calculation. The diversity of techniques could considerably impact the quality of interference amendment and investigation in numerous applications (Figure 6). The methods vary in their ability to handle diverse types of interference and data intricacies, making the selection of techniques imperative for attaining accurate results. Approaches like PCA are efficient in dimensionality reduction and the identification of variation in primary sources, which helps to isolate interference consequences, whereas PLS regression is more suited for managing overlapping spectral characteristics and quantitative investigation in the presence of interference. Because samples are often complex mixtures, which typically provide intricate data, this is often the case for spectroscopy and chromatographic analysis. However, the chromatographic method’s diversity allows specialists to select techniques tailored to the complexity of the data, confirming that interference is efficiently examined. The various chemometric techniques, including PCA, PLS, DA, RF, and SVM, are not always suitable as a single technique for monitoring and assessing environmental scenarios. The inappropriate use of a technique can mislead the interpretation of results; this issue can be addressed by engaging a diverse set of techniques allowing analysts to cross-validate their outcomes and confirm the robustness of their inferences. However, by employing appropriate chemometric methods, the accuracy and reliability of analytical results can be enhanced efficiently by properly handling and correcting various types of interference.

4. Environmental Application of Chemometric Techniques

4.1. Air Quality

Air pollution remains one of the most pressing environmental challenges globally, arising from the contamination of the atmosphere by physical, chemical, and biological agents in both indoor and outdoor environments. Poor indoor air quality is particularly hazardous, often linked to conditions such as Sick Building Syndrome and other occupational health risks. Beyond human health, air pollution negatively impacts vegetation, livestock, infrastructure, and ecosystems [97,98].

Atmospheric pollution occurs when concentrations of airborne contaminants surpass ambient regulatory thresholds. These pollutants, which vary in type and concentration over time, are typically more pronounced in densely populated urban centers and industrial regions [28]. Due to the dynamic interactions between physical transport mechanisms and chemical transformations, atmospheric pollutants exhibit complex, nonlinear behaviors. ANNs have proven particularly adept at modeling such nonlinear relationships, making them highly effective tools for predicting pollutant behavior and analyzing environmental data [99].

ANNs have demonstrated excellent performance in distinguishing regional patterns of air pollution. In a study analyzing 11 years of air quality data, pollutants such as PM₁₀, NO₂, SO₂, CO, O₃, and the API were found to be within Malaysian guideline limits. Strong positive correlations were observed between PM₁₀, NO₂, SO₂, and API values while CO and O₃ showed negative correlations with the API [100]. Notably, PM₁₀ and SO₂ emerged as the primary contributors influencing API values, primarily originating from industrial activities and vehicular emissions. In contrast, air quality in mining regions exhibited spatial variation based on proximity to water bodies and mining elevation.

Vehicle activity within mining areas is a significant contributor to both gaseous and particulate air pollutants. This emission load is further intensified by environmental factors such as wind speed and surface radiation, which facilitate the dispersion and distribution of contaminants. Statistical tools, such as Pearson correlation analysis, have proven effective in establishing relationships between surface temperature and concentrations of ambient air pollutants, offering insights into pollution dynamics in such areas [101]. Advanced chemometric techniques, including PCA and PMF, have been successfully employed to identify the sources of air pollutants such as CO, NO₂, SO₂, and PMs (Figure 7). The PCA indicated that the type of fuel could have a significant influence on the emission levels, PAHs, and TOC in the dust. The CA identified a correlation between PAH concentrations, dust amounts, and TOC, revealing that the proportionality is related to the size of the molecules [102]. The PCA and FA methods have detected eight pollutants in the LPS and SHPS regions, and eleven pollutants in the MPS region. The MLR models specify that the primary pollutant is PM₁₀, which is predicted to be the result of industrial activities, transportation, and agronomic practices [20]. However, it was found that the SHPS model is the most accurate for predicting total API, as it offers a better correlation between projected and calculated API, with the highest R² values for SHPS, MPS, and LPS being 0.894, 0.878, and 0.837, respectively [103]. The PLS-DA and PCA methods are highly significant in determining pollutants such as PM₁₀, PM_2.5, PM_1.0, and O₃, which affect indoor air quality [104]. The technique is too competent to provide a clear portrait of pollutants’ patterns and trends along with their associated sources. The pollutants that primarily emitted NO₂, CO, and PM₁₀, are attributed to industrial activities, building construction, and motor vehicles [105]. It is assumed that the formation of pollutants, including CH₄, NmHC, THC, O₃, and PM_10, is primarily caused by motor vehicles [106]. These methods detect patterns in pollutant distribution and can pinpoint likely emission sources even in the absence of on-site sampling or field investigations. This capability is particularly valuable for assessing spatial variability in air quality across remote or inaccessible regions (

Table 8. Applications of chemometric methods for air quality monitoring.

Chemometric Techniques	Analytical Method	Pollutants and Levels	Accuracy/Precision & R2	Ref.
DA, HACA, PCA, ANNs	API	Station 1 SO2: Maximum—0.1 ppm, Average—0.015 ppm; NO2: Maximum—0.22 ppm, Average—0.053 ppm; O3: Maximum—0.15 ppm, Average—0.034 ppm; CO: Maximum—10.41 ppm, Average—2.138 ppm; PM10: Maximum—806 µg/m3, Average—88.24 µg/m3; API: Maximum—392, Average—57.651 Station 2 SO2: Maximum—0.06 ppm, Average—0.006 ppm; NO2: Maximum—0.05 ppm, Average—0.022 ppm; O3: Maximum—0.14 ppm, Average—0.045 ppm; CO: Maximum—5.72 ppm, Average—1.393 ppm; PM10: Maximum—640 µg/m3, Average—83.27 µg/m3; API: Maximum—153, Average—51.068 Station 3 SO2: Maximum—0.02 ppm, Average—0.003 ppm; NO2: Maximum—0.12 ppm, Average—0.012 ppm; O3: Maximum—0.08 ppm, Average—0.024 ppm; CO: Maximum—4.13 ppm, Average—0.978 ppm; PM10: Maximum—411 µg/m3, Average—70.616 µg/m3; API: Maximum—188, Average—41.762	87.2% R2 > 0.75 p < 0.05	[99]
PCA, SPC, ANNs	API	SO2: Maximum—0.084 ppm, Average—0.003 ppm; NO2: Maximum—1.325 ppm, Average—0.013 ppm; O3: Maximum—0.149 ppm, Average—0.022 ppm; CO: Maximum—5.658 ppm, Average—0.702 ppm; PM10: Maximum—438.61 µg/m3, Average—51.866 µg/m3; API: Maximum—323, Average—56.431	R2 = 0.9	[100]
PCA, PMF	HQ	PM2.5: 65 ppm; PM10: 150 ppm; CO: 35 ppm; O3: 0.12 ppm; NO3: 0.053 ppm; SO2: 0.014 ppm	R2 = 0.37 p < 0.05	[101]
CA, PCA	GC-FID	Anthracene: Maximum—6420 ppm; Phenanthrene: Maximum—13,880 ppm; Fluorene: 5200 ppm; Acenaphthene: 5791 ppm	99%	[102]
HCA, DA, PCA, MLR	API	O3: Average—0.1 ppm; CO: Average—30 ppm; NO2: Average—0.18 ppm, SO2: Average—0.15 ppm; PM10: Average—120 µg/m3	95.38% p < 0.05	[103]
PCA, PLS-DA, LDA, AHC	Turnkey dust mate detector, and gas meter	O3: 0.467 ppm; CO: 0.781 ppm; CO2: 0.892 ppm; PM1: 0.798 µg/m3; PM2.5: 0.752 µg/m3; PM10: 0.751 µg/m3	89.05% R2 > 0.75	[104]
PCA, SPC	API	CO: CL—0.631 ppm, Upper control limit (UCL)—0.915 ppm, Lower control limit (LCL)—0.347 ppm, Maximum—37 ppm; PM10: CL—47.304 µg/m3, UCL—68.463 µg/m3, LCL—26.146 µg/m3	R2 = 0.49–1	[105]
PCA	UV-fluoresenece, and Teledyne API-FID	Station 1 CO: Maximum—4.85 ppm, Average—1.24 ppm; O3: Maximum—0.12 ppm, Average—0.03 ppm; PM10: Maximum—780 µg/m3, Average—81.24 µg/m3; SO2: Maximum—0.13 ppm, Average—0.01 ppm; NO2: Maximum—0.06 ppm, Average—0.02 ppm; CH4: Maximum—9.75 ppm, Average—2.49 ppm; NmHC: Maximum—5.15 ppm, Average—0.055 ppm; THC: Maximum—10.5 ppm, Average—2.96 ppm; API: Maximum—125.88, Average—57.84 Station 4 CO: Maximum—2.84 ppm, Average—0.86 ppm; O3: Maximum—0.16 ppm, Average—0.04 ppm; PM10: Maximum—202 µg/m3, Average—58.70 µg/m3; SO2: Maximum—0.1 ppm, Average—0.01 ppm; NO2: Maximum—0.06 ppm, Average—0.02 ppm; CH4: Maximum—9.33 ppm, Average—2.91 ppm; NmHC: Maximum—4.81 ppm, Average—0.41 ppm; THC: Maximum—9.6 ppm, Average—3.24 ppm; API: Maximum—158, Average–50.14 Station 7 CO: Maximum—3.82 ppm, Average—0.99 ppm; O3: Maximum—0.12 ppm, Average—0.02 ppm; PM10: Maximum—760 µg/m3, Average—94.66 µg/m3; SO2: Maximum—0.06 ppm, Average—0.01 ppm; NO2: Maximum—0.06 ppm, Average—0.01 ppm; CH4: Maximum—6.4 ppm, Average—2.24 ppm; NmHC: Maximum—6.17 ppm, Average—0.58 ppm; THC: Maximum—8.2 ppm, Average—2.75 ppm; API: Maximum—151, Average—57.32 Station 10 CO: Maximum—3.32 ppm, Average—0.57 ppm; O3: Maximum—0.06 ppm, Average—0.02 ppm; PM10: Maximum—357 µg/m3, Average—55.56 µg/m3; SO2: Maximum—0.04 ppm, Average—0.00 ppm; NO2: Maximum—0.04 ppm, Average—0.01 ppm; CH4: Maximum—6.64 ppm, Average—2.2 ppm; NmHC: Maximum—4.54 ppm, Average—0.4 ppm; THC: Maximum—7.6 ppm, Average—2.54 ppm; API: Maximum—97, Average—38.41	1% R2 > 0.75 p < 0.05	[106]
HACA, DA, PCA, FA, MLR	API	LPS region CO: 0.896 ppm; NO2: 0.939 ppm; SO2: 0.697 ppm; PM10: 0.646 µg/m3; O3: 0.343 ppm; CH4: 0.263 ppm; NO:0.873 ppm; Non-methane hydrocarbon: 0.887 ppm MPS region CO: 0.933 ppm; NO2: 0.733 ppm; SO2: 0.906 ppm; O3: 0.213 ppm; CH4: 0.913 ppm; NO: 0.857 ppm SHPS region CO: 0.801 ppm; NO2: 0.747 ppm; SO2: 0.108 ppm; O3: 0.024 ppm; CH4: 0.263 ppm; NO:0.918 ppm; Non-methane hydrocarbon: 0.218 ppm	91.67–97.22%	[20]

).

Analytical evaluations have further demonstrated the effectiveness of smokeless fuels in reducing airborne pollutants and dust emissions. While solid fuels remain commonly used—particularly for residential heating during colder months, evidence suggests that smokeless alternatives can significantly lower emission levels. These fuels represent a viable pathway toward cleaner coal technologies, especially in regions where solid fuel use is prevalent in household energy consumption.

The adoption of smokeless fuels, combined with improved land use planning and rigorous air quality monitoring, could support the development of more sustainable urban environments. In particular, urban planning strategies that integrate risk assessment for industrial zones and building expansions can contribute to more effective control of air pollution. Additionally, incorporating sustainable transportation solutions—such as biofuel-powered systems—can further reduce the release of toxic emissions, supporting long-term improvements in urban air quality and public health.

4.2. Water Quality

Organic pollutants are among the primary contributors to water pollution, posing significant challenges to both global environmental health and public health. The escalating discharge of untreated wastewater—driven by the expansion of livestock farming, industrial activities, and urban development—continues to deteriorate water quality worldwide. Climate change, particularly global warming, exacerbates this issue by reducing the availability of freshwater from rivers, lakes, and reservoirs. The presence of persistent and toxic contaminants in water further heightens the risk to aquatic ecosystems, natural resources, and overall biodiversity.

The degradation of water resources containing organic pollutants represents a critical environmental crisis. Multivariate statistical methods, particularly chemometric techniques, have proven to be powerful tools for assessing and quantifying water pollution. These approaches are instrumental in evaluating water quality for various uses, safeguarding aquatic life, and managing environmental resources effectively [107]. Chemometric techniques facilitate the reduction in large, complex datasets enabling the extraction of key variables that significantly influence water quality. They also help identify the sources of contamination, whether natural—such as water–rock interactions—or anthropogenic, including agricultural runoff and industrial discharge [108].

To assess the physicochemical properties of water, samples were collected from drainage systems carrying untreated sewage, agricultural runoff, and industrial effluents across urban, peri-urban, and rural areas. Fifteen water quality parameters were analysed using a combination of chemometric models, including PCA, FA, CA, and DA. These methods revealed strong interrelationships among the variables, providing insight into the sources of pollution. PCA successfully identified the dominant factors influencing water quality while FA highlighted agricultural and domestic wastes as the primary contributors to contamination [

Table 9. Applications of chemometric methods to determine water quality.

Chemometric Techniques	Analytical Method	Parameters and Concentration	Accuracy/Precision & R2	Ref.
HCA, PCA	pH meter, conductivity meter, spectrophotometry, and flame photometer	pH: Minimum—4.4, Maximum—7.10, Mean—6.49; EC: Minimum—270, Maximum—1870, Mean—893.56; TDS: Minimum—142, Maximum—1720, Mean—536.88; K+: Minimum—9.8, Maximum—99.6, Mean—56.97; Na+: Minimum—5.1, Maximum—59.52, Mean—31.96; Mg2+: Minimum—3.73, Maximum—9.5, Mean—6.02; Ca2+: Minimum—1.3, Maximum—7.37, Mean—4.71; Cl−: Minimum—57.6, Maximum—1476, Mean—445.01; SO4−: Minimum—0, Maximum—6.1, Mean—2.26; HCO3−: Minimum—100.5, Maximum—609, Mean—253; NO3−: Minimum—0, Maximum—6.09, Mean—1.36	69.9% R2 = 0.849, 0.968 p < 0.05	[107]
PCA, FA, CA, DA	Conductivity meter, titration, UV-spectrophotometry, and TKN	pH: Minimum—6.8, Maximum—8.3, Mean—7.8; COD: Minimum—40, Maximum—120, Mean—81.6; BOD: Minimum—12, Maximum—48, Mean—22.2; Alkalinity: Minimum—222, Maximum—514, Mean—360; TDS: Minimum—104, Maximum—360, Mean—264; TSS: Minimum—7, Maximum—153, Mean—74.2; SO4-S: Minimum—49.4, Maximum—185.3, Mean—89.9; NO3-N: Minimum—0.1, Maximum—4.1, Mean—1.6; NO2-N: Minimum—0.7, Maximum—45.7, Mean—15.8	100% R2 > 0.7	[108]
PCA, HCA	pH meter, conductivity meter, spectrophotometry, TKN, IC, and ICP-MS	pH: 7.33; EC: 1.03; TDS: 728; BOD: 18; COD: 22; K+: 0.56; Na+: 4.67; Mg2+: 1.72; Ca2+: 3.44; Cl−: 3.68; SO4−: 2.08; HCO3−: 4.38; NO3−: 12.55; NH4: 3.31	76–81.1% R2 = 0.3–1 p < 0.05	[109]
PCA, CA	pH meter, conductivity meter, UV–Vis spectrophotometer, and AAS	pH: Minimum—6.83, Maximum—7.63, Mean—7.29; TDS: Minimum—148.5, Maximum—662, Mean—328.3; Fe: Minimum—0, Maximum—0.03, Mean—0.0177; SO4: Minimum—0, Maximum—2, Mean—1; NO3: Minimum—0.6, Maximum—1, Mean—0.833; CaCO3: Minimum—184, Maximum—678.8, Mean—354; Cr: Minimum—0, Maximum—0.05, Mean—0.0267; Zn: Minimum—0.09, Maximum—0.43, Mean—0.2567; CN: Minimum—0, Maximum—0.04, Mean—0.0133; KMnO4: Minimum—2.43, Maximum—3.63, Mean—3.22	18.43–81.57% p < 0.05	[110]
FA, CA	pH meter, conductivity meter, incubation and titration, argentometric titration, and complexometric titration	pH: Minimum—7.16, Maximum—8.34, Mean—7.86; DO: Minimum—6.7, Maximum—8.8, Mean—7.786; TDS: Minimum—304.3, Maximum—452.8, Mean—348.2; BOD: Minimum—3.2, Maximum—5.8, Mean—4.72; Cl−: Minimum—24.61, Maximum—55.85, Mean—36.31; Mg2+: Minimum—18.74, Maximum—119.14, Mean—45.84; Ca2+: Minimum—115.9, Maximum—168.26, Mean—131.05	75.4–83.05% p < 0.05	[111]
PCA	pH meter, conductivity meter, turbidimetry, nephelometric method, titrimetry, and ICP-OES	pH: Minimum—7.42, Maximum—8.59, Mean—8.21; DO: Minimum—4.62, Maximum—8.8, Mean—7.24; CE: Minimum—856, Maximum—2420, Mean—1827.58; Nitrites: Minimum—0.003, Maximum—2.09, Mean—0.447; Cl−: Minimum—134.9, Maximum—724.9, Mean—458.19; NO3−: Minimum—4.22, Maximum—13.64, Mean—9.68; Cu: Minimum—0.036, Maximum—0.539, Mean—0.135; Cd: Minimum—0.088, Maximum—0.378, Mean—0.137; Pb: Minimum—0.069, Maximum—0.307, Mean—0.109; Cr: Minimum—0.0143, Maximum—0.278, Mean—0.073	84–96% R2 = 0.256–0.989	[112]
PCA	pH meter, conductivity meter, and GC-MS	Samples–Crati 13 pH: 8; NH4+: 0.19; N-NO2: 0.06; Al3+: 0.09; As: 0.09; Cr: 0.4; Fe: 36; Hg: 0.3; Ni: 0.5; Pb: 3; B: 5; Se: 0.04	29%, 49%	[113]
CA, FA, DA	pH meter, conductivity meter, incubation and titration, and complexometric titration	pH: 7.30–8.96; BOD: 0.6–9; DO: 4.3–16.4; NO3: 0.02–1.5; NO2: 0.006–0.953; NH4: 0.08–2.8; COD: 22; Mg2+: 4–66; Ca2+: 43–253; Cl−: 25–91; SO4−: 19–185; Pb: 1–13.6; Cd: 1–7	76–100%	[114]

]. Through HCA and DA, the spatial variability of water quality parameters (Figure 8) was simplified into four principal influencing factors [109,110]. However, several parameter concentrations, including TH, BOD, total alkalinity and TDS, exceeded permissible limits—indicating significant pollution from agricultural, domestic, and industrial sources. Routine monitoring and assessment of drinking water are crucial for detecting contamination risks and ensuring water safety [111]. Public awareness campaigns are equally important for promoting an understanding of water quality issues and mitigating the health impacts associated with polluted water sources. Chemometric methods, such as PCA, have also been effectively applied to determine heavy metal concentrations in river water, providing precise and reproducible analyses [112,113,114]. These models serve as valuable tools in independently evaluating water quality and identifying hazardous residues that pose threats to public health. Their predictive capabilities enable the development of practical water management strategies, including intervention plans and early-warning systems. The outcomes of such analyses have facilitated the creation of pollution hazard maps, which provide critical data for environmental authorities and communities to assess risk levels and implement measures to reduce the adverse effects of water contamination on human health and ecosystems.

4.3. Soil Quality

Various agricultural activities, illegal waste dumping, and underground storage tank leakage are primarily responsible for soil pollution, a critical global concern today. It also increased the difficulty related to public health and regulatory compliance due to the results of contaminated groundwater, which directly impacted the ecological systems. Elemental soil analysis is crucial for providing detailed information on contamination resulting from atmospheric accumulation or transport in surface water and groundwater. Multivariate techniques have recently been enhanced to improve the analysis methodology using two-dimensional datasets from soil sampling for evaluation involving CA, DA, FA, and PCA models [115]. These methods can identify site locations with pollutants and categorize the toxic pollutants that affect the environment, as well as their relationship with the influencing variables. The soil profiles indicated the presence of chromium (Cr) and vanadium (V) [115], which is typically due to atmospheric pollutants contributing to pollution, possibly from transportation and agricultural influences. The absolute principal component score approach is applied to reveal the contribution of each identified parameter to the total chemical parameter concentration by the PCA model with a latent factor. The cluster analysis interpreted the similarity of variables among different pattern formations in terms of pollution risk or spatial settings [116]. PCA initially evaluates the dataset structure to identify the most influential factors based on the data structure obtained. It could enhance the soil quality performance of a non-ferrous lead–zinc smelter and help determine the reasons for influencing outcomes from its origin with involved consequences that could affect the region [117]. Chemometric analysis also included PLSR, SPCA, and the Least Absolute Shrinkage and Selection Operator calibration methodology, utilizing UV–Vis-NIR spectra to predict the twenty-two physical, chemical, and biological properties of the soil. It showed the highest Ca and Mg concentrations, as well as the highest β-glucosidase, fungal biomass, and phospholipid fatty acid values, which also enhanced prediction accuracy, particularly in terms of reducing prediction error and bias for innumerable soil properties [118]. Different approaches that integrate with Vis-NIR and SWIR are capable of processing spectral data for quantifying SOM [119]. The study ascertained that the topsoil characteristics are used to establish the optimal wavelength band for developing a model to estimate SOM content from rivers. In addition, PLSR and SVR models were applied to a pre-processing procedure that reduced the discriminant error caused by spectral overlaps, thereby enhancing the model’s performance (Figure 9).

The chemometric approach was able to quantify major pollution resources differently, which may be related to manufacturing activity, soil-specific properties, and aerosol alluviation practices. The corresponding distribution regression patterns are designed to provide a quantitative explanation of each identified source’s contribution to the total chemical parameter concentrations managed for soil quality monitoring and regulatory compliance. Moreover, sampling locations were also performed for spatial analysis to demonstrate differences between sampling regions in similar patterns. This resulted in higher pollution levels in industrial areas and comparatively lower pollution levels in locations near the shoreline and in mountainous regions. These methods are effective for soil characterization (

Table 10. Applications of chemometric methods to determine soil quality.

Chemometric Techniques	Analytical Method	Parameters and Concentration	Accuracy/Precision & R2	Ref.
HCA, PCA	pH meter and EDXRF	NIST SRM-1646a (Estuarine Sediment) K: 0.67 cg/kg; Ca: 0.456 cg/kg; Fe: 1.743 cg/kg; Ti: 0.556; V: 47.62 mg/kg; Cr: 63.9 mg/kg; Ni: 21.8 mg/kg; Cu: 7.9 mg/kg; Zn: 45.57 mg/kg IAEA SOIL-7 (Austria) K: 1.34 cg/kg; Ca: 22.75 cg/kg; Fe: 3.28 cg/kg; Ti: 4583.29; V: 68.84 mg/kg; Cr: 123.89 mg/kg; Ni: <22.9 mg/kg; Cu: 14.3 mg/kg; Zn: 104.21 mg/kg	36.94–85.99%	[115]
HCA, PCA	ICP MS and UV–VIS spectroscopy	pH: minimum—5.45, maximum—8.25, mean—6.91; N total (mg/kg): minimum—794.66, maximum—2856, mean—1564.65; P total (mg/kg): minimum—268.16, maximum—1920.83, mean—744.87; TC (%): minimum—0.88, maximum—4.42, mean—2.28; TOC (%): minimum—0.93, maximum—3.73, mean—1.98; As (mg/kg): minimum—3.59, maximum—16.63, mean—7.55; Cu (mg/kg): minimum—11.76, maximum—97.42, mean—44.86; Cr (mg/kg): minimum—18.1, maximum—230.42, mean—71.87; Ni (mg/kg): minimum—9.44, maximum—85.19, mean—37.63; Cd (mg/kg): minimum—0.22, maximum—0.63, mean—0.4; Zn (mg/kg): minimum—30.58, maximum—115.03, mean—64.03; Pb (mg/kg): minimum—11.41, maximum—37.42, mean—20.78	90–110% R2 = 0.79–0.91	[116]
CA, PCA	pH meter, ICP-OES, and ETAAS	Location: K Zn (mg/kg): 103.74; Cd (mg/kg): 0.17; Pb (mg/kg): 17.66; Cu (mg/kg): 33.55; Hg (mg/kg): 0.033 Location: KRM Zn (mg/kg): 66.33; Cd (mg/kg): 0.38; Pb (mg/kg): 14.33; Cu (mg/kg): 12.07; Hg (mg/kg): 0.036 Location: Kagri Zn (mg/kg): 384.6; Cd (mg/kg): 2.22; Pb (mg/kg): 19.69; Cu (mg/kg): 8; Hg (mg/kg): 0.106 Location: AS Zn (mg/kg): 409.7; Cd (mg/kg): 0.97; Pb (mg/kg): 31.68; Cu (mg/kg): 25.22; Hg (mg/kg): 0.072 Location: PB Zn (mg/kg): 330.4; Cd (mg/kg): 0.81; Pb (mg/kg): 14.73; Cu (mg/kg): 30.76; Hg (mg/kg): 0.043	80% R2 = 0.81–0.93	[117]
PCA, CA	Potentiometry, UV–Vis-NIR, ICP-OES, and GC	Location: F. Sylvatica pH: 6.07 ± 0.24; TOC (mg/g): 171.30 ± 26.80; Total N (mg/g): 10.32 ± 1.85; Total Ca (mg/g): 46.81 ± 11.68; Total K (mg/g): 11.18 ± 4.08; Total Mg (mg/g): 7.10 ± 2.82; Total Mn (mg/g): 1.35 ± 0.56; Total Na (mg/g): 2.76 ± 1.17; Total Fe (mg/g): 24.49 ± 6.70; Total Al (mg/g): 42.46 ± 15.72; Py-Fe (mg/g): 6.18 ± 1.75; Py-Al (mg/g): 18.02 ± 3.97; Ox-Fe (mg/g): 12.65 ± 2.88; Ox-Al (mg/g): 23.93 ± 7.79 Location: Q. Cerris pH: 6.57 ± 0.23; TOC (mg/g): 80.10 ± 9.70; Total N (mg/g): 6.12 ± 0.71; Total Ca (mg/g): 20.89 ± 12.93; Total K (mg/g): 3.84 ± 1.70; Total Mg (mg/g): 9.33 ± 4.62; Total Mn (mg/g): 1.98 ± 1.34; Total Na (mg/g): 0.14 ± 0.06; Total Fe (mg/g): 22.59 ± 4.72; Total Al (mg/g): 22.23 ± 12.20; Py-Fe (mg/g): 1.95 ± 0.42; Py-Al (mg/g): 2.23 ± 0.69; Ox-Fe (mg/g): 8.33 ± 2.31; Ox-Al (mg/g): 6.45 ± 2.37 Location: Q. Ilex pH: 6.87 ± 0.29; TOC (mg/g): 328.80 ± 50.30; Total N (mg/g): 17.65 ± 3.47; Total Ca (mg/g): 140.06 ± 49.95; Total K (mg/g): 6.25 ± 3.04; Total Mg (mg/g): 15.12 ± 7.51; Total Mn (mg/g): 0.96 ± 0.37; Total Na (mg/g): 1.12 ± 0.69; Total Fe (mg/g): 11.14 ± 2.47; Total Al (mg/g): 25.13 ± 19.75; Py-Fe (mg/g): 1.64 ± 0.56; Py-Al (mg/g): 4.86 ± 2.24; Ox-Fe (mg/g): 5.22 ± 3.24; Ox-Al (mg/g): 13.62 ± 10.35		[118]

). However, traditional methods for agrochemical analysis are often time-consuming and costly. They also used hazardous chemical reagents but applied eco-friendly alternatives, such as chemometrics techniques, which are tools that can generate accurate predictions for most soil characteristics.

5. Conclusions

Environmental monitoring and assessment are inherently complex due to the dynamic interplay between natural processes and anthropogenic activities. This complexity often leads to overlapping variables and non-trivial datasets that challenge conventional analysis methods. Chemometric techniques offer a powerful and systematic approach to managing this complexity, enabling the extraction of meaningful information from large, multivariate datasets with high accuracy.

DA, when applied to geospatial datasets, has demonstrated its effectiveness in distinguishing between monitoring locations based on environmental variables. HACA has revealed temporal variations in pollution levels, providing insight into monthly and annual trends often associated with suspended atmospheric particulates such as smoke, dust, and chemical pollutants. PCA has been particularly effective in reducing data dimensionality and identifying key pollution sources—primarily emissions from fossil fuel combustion in transportation and industrial processes.

Moreover, ANNs have demonstrated superior predictive performance and selectivity compared to traditional methods, such as DA, especially in regional discrimination tasks. These findings underscore the importance of selecting chemometric techniques that are specifically tailored to the nature of the environmental problem being addressed.

Ultimately, the integration of appropriate chemometric methods into environmental monitoring frameworks can significantly enhance data interpretation, pollution source identification, and temporal trend analysis. To ensure effective environmental management and pollution mitigation, it is crucial to implement targeted action plans based on robust chemometric analysis.