مجله مهندسی و مدیریت آبخیز
سال هفدهم شماره 4 (زمستان 1404)

In integrated watershed management, assessing the status and dynamics of watershed health is essential as a fundamental tool for identifying and implementing effective management responses. Diagnosing the issues affecting water and soil resources, along with identifying the causes of various problems in watersheds, constitutes a critical step toward sustainable natural resource management. This understanding not only enables the analysis of threatening factors but also provides a basis for identifying appropriate solutions to protect and restore watershed resources. The DPSIR framework (Drivers, Pressures, State, Impact, and Response) serves as a comprehensive analytical model, capable of explaining the causal relationships among various factors and effectively evaluating watershed health. The aim of this research is to use the DPSIR framework for a comprehensive analysis of watershed health status and to identify key factors contributing to its decline, with a focus on the watershed area draining to the Ilam Dam, in order to propose effective and sustainable management solutions.
 
To accurately identify the factors related to each component of the DPSIR framework, a systematic review of the literature and previous studies was conducted through library research to gather theoretical background and scientific resources. Field visits to the watershed were then undertaken to assess the current conditions and directly observe the natural and human factors affecting the region. Additionally, brainstorming sessions and semi-structured interviews with experts and local stakeholders, including residents of the watershed area, were carried out to collect comprehensive information on the issues and influencing factors. Based on the preliminary findings, a questionnaire was designed and its validity was confirmed by a panel of experts. Cronbach's alpha was used to measure reliability, and the results indicated an acceptable level of reliability. The survey, using a Likert scale, was administered to 20 experts and 20 local residents of the watershed. To analyze the data and prioritize factors from the participants' perspectives, Friedman’s test was used to determine the relative importance of each factor within the DPSIR framework.
 
The findings show that in the studied watershed, five driving forces have led to 34 distinct pressures on watershed resources, which in turn have caused 11 unfavorable states. These states have also resulted in 20 unintended impacts. Additionally, 32 management responses were proposed to improve the current situation. The relationships among the factors within each of the main components of the DPSIR framework were examined and prioritized based on the views of both experts and stakeholders. According to the results, the alignment of shared priorities within the top 40% of the most important factors was as follows: 50% for driving forces, 69% for pressures, 80% for states, 75% for impacts, and 84% for responses.
 
Planning and policymaking aimed at achieving sustainable economic, social, and environmental development require access to accurate and comprehensive information about the conditions and dynamics of watersheds. The results of this study indicate that identifying and implementing appropriate management strategies can play a decisive role in improving the health of natural resources and watershed ecosystems. This research was conducted with the aim of identifying and prioritizing management responses to improve the environmental conditions of the study area. The analyses, carried out using the DPSIR approach as an effective cause–effect analytical framework, provided valuable tools for identifying the main problems and challenges of the watershed. Moreover, the study identified the key factors and pressures impacting natural resources. Therefore, the proposed management responses can serve as practical solutions for improving current conditions and preventing future problems in the sustainable management of natural resources and ecosystems in the targeted area. These management responses play a significant role not only from a scientific perspective but also in the practical implementation of comprehensive watershed management programs, offering actionable guidance for decision-makers to enhance the state of natural resources and ecosystems.

Accurate streamflow prediction is essential for water resources management and flood control. Due to the complex and nonlinear behavior of streamflow, traditional models are often inadequate. Machine learning and deep learning algorithms offer more robust solutions; however, their accuracy can be affected by sudden climatic fluctuations. Consequently, employing hybrid methods is necessary to improve prediction accuracy. The literature review reveals that, despite the high capabilities of machine learning models, a research gap still exists in managing multi-scale fluctuations in streamflow data. This underscores the necessity of using hybrid approaches to enhance prediction accuracy. The innovation of this study is a hybrid framework that simultaneously models both long-term patterns and short-term fluctuations by integrating wavelet analysis, used to decompose the streamflow signal, with a powerful deep learning model. 

In this study, to predict the streamflow of the Kurkursar River in Nowshahr, hydrological data including daily precipitation and river discharge over a 20-year period at a daily resolution were utilized. The input variables included daily precipitation (Pt​) and streamflow with time lags of one, two, and three days (Qt−1​, Qt−2​, Qt−3​). Before the modeling process, data preprocessing was performed, which included reconstructing missing data, removing anomalous data (outliers), and normalizing the values to improve data quality and enhance their reliability in hydrological analyses. The hydrological data from the watershed were divided into three subsets: training (70%), validation (15%), and testing (15%). Four streamflow prediction scenarios were selected based on Pearson correlation coefficient analysis to identify sensitive variables and determine the model inputs. The river streamflow modeling process was carried out using two algorithms: Random Forest (RF) and the deep learning Long Short-Term Memory (LSTM) recurrent neural network. Furthermore, to enhance the accuracy and improve the generalizability of the models, various wavelet transform methods, including Daubechies 4 (Db4), Haar, and Mexican Hat wavelets, were used to extract multi-scale features and combine them with the input data for the RF and LSTM models. This hybrid approach facilitated the identification of complex spatio-temporal patterns in the hydrological time series. After the final evaluation of the prediction models' performance, the Daubechies 4 (Db4) wavelet transform was employed to optimize their coefficients and structural parameters. Performance evaluation metrics, including the Coefficient of Determination (R²), Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Percent Bias (PBIAS), Mean Absolute Percentage Error (MAPE), and Kling-Gupta Efficiency (KGE), were used to assess the accuracy of the models' predictions. Ultimately, the optimal models were selected based on a comparative analysis of these quantitative criteria. Additionally, for data analysis and visual presentation of the results, various plots were used, including scatter plots, time series of observed and predicted data, and error distributions such as error histograms, normal density curves, cumulative distribution functions of errors, and quantile-quantile (Q-Q) plots. 

The results showed that in streamflow prediction, previous time steps (different lags) were the most important variables for predicting all subsequent horizons. The final results regarding the model scenarios indicated that the first scenario (S1), which only used the precipitation variable, was the weakest performer in all cases. Furthermore, the sixth scenario (S6), which utilized all available variables (Pt​,Qt−1​,Qt−2​,Qt−3​), had the best performance in the training and testing phases for both standalone and hybrid models. The research findings indicated that the hybrid Random Forest-Wavelet (RF-Wavelet) model had the best performance in both the training (R²=0.907, RMSE=0.0192) and testing (R²=0.942, RMSE=0.0106) phases. Additionally, the standalone Long Short-Term Memory (LSTM) deep learning model had the weakest performance in the training (R²=0.499, RMSE=1.6) and testing (R²=0.579, RMSE=1.149) phases. The findings also showed that the Daubechies 4 wavelet , when combined with the Random Forest model, was able to reduce the error of the standalone RF model by approximately 55%. Additionally, the wavelet, when combined with the LSTM model, was able to increase the prediction accuracy by approximately 39%. Furthermore, a comparison of the wavelet-hybrid models showed that the RF-Wavelet model reduced the error by approximately 23% compared to the hybrid LSTM-Wavelet model. 

In this research, various wavelet transform models, including Daubechies 4, Haar, and Mexican Hat, were utilized for integration with RF and LSTM algorithms. Quantitative and qualitative analyses showed that the Daubechies 4 wavelet transform had significant superiority in improving streamflow prediction accuracy compared to other wavelet types within both RF and LSTM model frameworks. Therefore, this type of wavelet transform was selected and used as the primary basis for integration with these two prediction models. Examination of the error distribution pattern in the training data indicates a major concentration of error values in regions adjacent to zero. The distribution of errors was observed to be approximately symmetrical and showed considerable consistency with a normal distribution. This pattern signifies the model's satisfactory accuracy in the training and data-fitting process. Ultimately, the present study focused on the development of data-driven models to determine the optimal combination of predictor variables for modeling and predicting river streamflow. This research demonstrated that integrating the Daubechies 4 wavelet transform with the Random Forest (RF) model served as the optimal and superior approach for predicting hydrological streamflow in the present case study. The aforementioned hybrid model, in addition to significantly enhancing performance compared to standalone models by reducing prediction error by up to 55%, showed notable superiority over complex deep learning models, including LSTM and its associated hybrid combinations. This achievement highlights the importance of extracting multi-scale time-frequency features using the wavelet transform and emphasizes its pivotal role in improving the accuracy and generalizability of hydrological streamflow predictions, even in comparison to advanced architectures of deep temporal models.

Estimates in Iran indicate an annual loss of about one billion cubic meters of soil from the country's land resources. Although quantifying the economic value of this volume of soil resource loss is difficult, the fragile ecological balance of ecosystems in the country's land resources suggests that this volume of resource loss will cause irreparable and irreversible damage to the production capacity and functioning of these resources. A part of this soil loss is compensated for and replaced by soil formation processes, and as long as the rate of erosion does not exceed the rate of soil formation, it is considered a natural and inevitable process. Knowing the rate of natural replacement is necessary for monitoring changes in the quality and quantity of this natural resource and for understanding the process of its deterioration or recovery. In addition, soil erosion beyond its place of occurrence causes economic, biodiversity, and natural landscape damages. The tolerable limit for this type of damage also requires its own scientific standard. On the other hand, substantial amounts of the country's financial resources are spent annually on watershed management measures. However, there is no quantitative regional standard for the design of these measures, nor a practical guideline for evaluating their effectiveness of these measures. This standard, along with the rate of soil regeneration and renewability, is known worldwide as tolerable soil erosion. 

In this article, the development of the concepts of tolerable erosion worldwide has been monitored, and the developed techniques and methods have been investigated for their potential generalization and application to the conditions of Iran. In this study, 109 domestic and international research articles were reviewed, and a summary is provided of the evolution of the concept of tolerable soil erosion, the factors influencing the T-value, and the methods for its calculation. Additionally, a summary of the research conducted, their results related to tolerable erosion, and various methodologies in this field are presented. Recommendations, research needs, and optimal strategies for estimating tolerable erosion in the country's conditions are also presented. 

The concept of tolerable soil erosion used in soil conservation programs is not suitable for maintaining the long-term indefinite productivity of agricultural lands over the long term. This is because these values are based on incorrect assumptions about the rates of topsoil formation and mineral weathering processes. The concept is built on two assumptions: first, that soil scientists can reliably and accurately assess the maximum tolerable erosion rates; and second, that policymakers can objectively weigh and balance these assessments against various interests or needs. Both assumptions should be challenged. Short-term political considerations might require that public policies allow soil resources to be degraded gradually and continuously to the point where they are no longer usable for agriculture. However, sustained support for such policies must clearly take into account the quantity and quality of available information on soil formation rates under agricultural conditions. Another point is that assessing soil erosion damages beyond the site of erosion to facilities, infrastructures, and biological resources also requires monitoring indicators off-site for evaluation. Tolerable soil erosion values are crucial and should not be determined based on incorrect and unscientific assumptions. Overall, after reviewing the proposed methods for determining tolerable erosion, it can be concluded that the method suggested by Macedo could be appropriate for Iran's conditions, given that the parameters examined in this method are easily accessible and, considering the data scarcity in Iran, this method is well-suited to these conditions. 

Given that the concept of tolerable erosion based on soil fertility and soil formation rates is insufficient, and that the off-site effects of soil erosion should also be considered, more extensive and scientific research in this field is necessary. Since the concept of tolerable soil loss is useful for planning soil conservation strategies, its quantitative estimation is essential and should be performed taking into account extreme rainfall events rather than average conditions. Therefore, research on the statistical distribution of maximum annual soil losses in different regions of the world is necessary.

Changes in river discharge fluctuations, whether increases or decreases, can lead to irreversible damage to both human and natural environments. It is now well established that variations in the phases of teleconnection patterns can cause significant increases or decreases in river discharge across different regions of the world, depending on their influence on precipitation cycles, evapotranspiration, and drought occurrence. 

For this study, data on 28 major teleconnection indices affecting Iran’s climate were obtained from the NOAA website. In addition, river discharge records from selected hydrometric stations located upstream of the dams on the Karun River (Telezang, Lordegan, Armand, Pataveh, and Kata stations) were collected from the Ministry of Energy for a 30-year period (1993–2022). Following a preliminary assessment of the discharge data, missing values were reconstructed using the multiple regression method to ensure data consistency and reliability. This method was chosen for its ability to preserve the general trend of the dataset while minimizing disturbance to the data. To analyze the trends and magnitude of seasonal and annual discharge variations, the non-parametric Mann–Kendall test and Theil–Sen slope estimator were applied. Furthermore, the relationship between teleconnection indices and river discharge was examined using the Pearson correlation coefficient across three temporal scales: monthly, seasonal, and annual. These correlations were assessed both simultaneously and with time lags of one to three months. Given the large number of indices, only those with statistically significant correlations with discharge were included in the final analyses. 

The results indicated that, overall, the mean discharge exhibited a continuous decreasing trend during the study period. Specifically, the mean discharge in all seasons showed a declining pattern, and at the annual scale, a significant decrease of 11.3 m³/s per year was observed, accompanied by a negative Mann-Kendall statistic (Z = –4.1). The correlations between teleconnection patterns and discharge in the study basin were further explored. The findings revealed that, at different temporal scales (monthly, seasonal, and annual), several teleconnection indices—including GLBTS, WHWP, SOI, Solar Flux, TSA, TNA, NINO4, AMO, AMM, MEI v2, PDO, NINO1+2, AAO, Warm Pool, PNA, EPO, WP, TNH, NCP, RMM1, and RMM2-exhibited statistically significant correlations (at the 0.05 and 0.01 confidence levels) with the discharge of the Karun River headwaters, either simultaneously or with lags of one to three months. 

Overall, the mean discharge recorded at the selected upstream stations of the Karun River dams demonstrated a decreasing trend. Moreover, the study confirmed the existence of significant simultaneous and lagged correlations between the fluctuations of several teleconnection indices and river discharge. Consequently, it can be concluded that the identified teleconnection patterns exert considerable influence on discharge variations in the basin. Given the predictability of these teleconnection indices and their significant correlations with discharge, future fluctuations of the Karun River discharge can potentially be forecasted at monthly, seasonal, and annual scales based on the findings of this research.

Over the past decades, excessive and unplanned exploitation of watershed resources (soil, water, and vegetation) has led to a decline in their overall health. One approach that can play a preventive role in this regard is to focus on alternative livelihoods, especially in the ecotourism and tourism sectors. Therefore, it is necessary to identify and assess geosites and geomorphosites across different regions. The aim of the present study is to evaluate the tourism value of geomorphotourism landscapes in the Hableh-Rud basin using the Pralong model. 

To evaluate these geosites, detailed information about the studied area was obtained through field and library studies. The Hableh-Rud watershed, with an area of 1,265,977 hectares, is located between 52°39' to 53°08' east longitude and 35°26' to 35°57' north latitude in Tehran and Semnan provinces. From among the geomorphic sites in the Hableh-Rud region, three sites were selected, and a detailed profile was prepared for each geomorphosite. The sites include Rudafshan Cave in the west of Damavand city, Khumdeh Spring, and Vashi Gorge. In this research, based on the Pralong model, a survey of experts knowledgeable about the study area and visiting tourists (30 people) was conducted using a simple random sampling method. The Pralong model specifically evaluates the tourism quality of geomorphosites and their potential for use. In this model, the tourism value of a site is calculated as the average of its aesthetic, scientific, cultural, and functional-economic values. Specific criteria and scales have been defined to determine the value of each aspect of the tourism quality of geomorphic sites. Accordingly, the tourism score of a site is the average of these four criteria. 

After analyzing the data, it was determined that, among the three selected sites, the Vashi Gorge geosite has the highest geomorphotourism value, with a tourism score of 0.54 and an average use potential value of 0.68. Factors that have increased the value and importance of the Vashi Gorge include its uniqueness at the national level, pleasant and cool climate, distinctive geomorphological forms and unique engravings on the stone wall of the gorge, as well as the scenery of the waterfall and the presence of a very beautiful meadow surrounding it. This gorge, with its raging river and cool water, has attracted many visitors and tourists. The Rudafshan Cave geosites ranked second with a tourism score of (0.41). The reasons for this ranking can be attributed to the concentration of geomorphological phenomena in a small area (inside the cave), its location next to Rudafshan village, and natural attractions (Delichai River, the pristine nature of the village with green and tall trees). The Khumdeh mineral spring with a score of (0.31) ranked third. The reasons include the unique character of the spring in the region, its therapeutic properties, its proximity to the flowing Hableh-Rud River, the beautiful surrounding nature, and easy access via an asphalted road. The evaluations show that the tourism values of the geomorphological landforms in the Vashi Gorge region are mainly due to the high scores for physical beauty, cultural value, and scientific value of this landform. The interrelationships between these values should be considered. In general, across all landforms, the aesthetic score and the scientific score are similar. 

According to the results, the highest tourism quality and value belong to the Vashi Gorge, and the lowest belongs to the Khumdeh mineral spring. However, except for the Vashi Gorge, the scores for scientific and aesthetic quality, as well as use potential, in the other landforms are very low. It is necessary to provide these landforms with the facilities needed by tourists to ensure optimal utilization of these areas. The tourism value of these sites depends on the high quality of their aesthetic appeal and their cultural and scientific qualities. Factors such as difficult access, the lack of amenities and proper services, and insufficient attention to geotourism have contributed to the relatively low total calculated scores.

Natural ecosystems play a vital role in maintaining environmental processes and life-support systems, but due to land-use changes and improper management, essential ecosystem services for human well-being have declined. This reduction has led to problems in areas such as energy, climate, and ecosystem services. With increasing population and human needs, greater pressure has been placed on ecosystems, highlighting the global importance of ecosystem services worldwide. The InVEST model, developed by the Natural Capital Project, serves as a tool for assessing and mapping these services. The objective of this research is to spatially analyze the hydrological ecosystem services using the InVEST model in the Neyshabur Watershed. 

In this study, two sub-models—the Seasonal Water Yield and Sediment Delivery Ratio (SDR) models in InVEST—were used to assess water yield and estimate soil erosion in the study watershed. To run the model, inputs including Digital Elevation Model (DEM) maps, watershed and sub-watershed maps, precipitation, reference evapotranspiration, soil hydrological groups, land use, soil erodibility, rainfall erosivity, a biophysical data table, and the number of precipitation events were required. For calibration and validation of the model in the study area, data from three hydrometric stations in the watershed for the period 1982 to 1993 were used. Then, by adjusting the model’s sensitive parameters, including β (local topographic and soil parameter), γ (parameter related to infiltration rate in cells), Z (number of rainfall events), CN values, k_b, and IC₀ (parameters for the relationship between hydrological connectivity and sediment delivery ratio), the model was calibrated within the allowed range. Calibration was performed using the discharge and sediment data recorded at the Bar, Taghun, and Zarandeh stations for the period 1982 to 1989, and validation was carried out for the period 1990 to 1993. To determine the hydrological ecosystem services, monthly temperature and precipitation maps from WorldClim for the period 1970 to 2000 with an appropriate resolution were used, and the model was run based on these data. 

Among the parameters α, β, γ, CN, Z, k_b, and IC₀, the model showed the highest sensitivity to the parameters Z (number of rainfall events), CN values, k_b, and IC₀. The spatial variation pattern of quickflow in the Neyshabur watershed indicates that quickflow is primarily influenced by precipitation. The northern, northeastern, and eastern parts of the watershed, due to steep slopes and intense rainfall, experience higher quickflow, while the southwestern, southern, and central areas of the watershed have lower quickflow. A strong correlation of 0.83 between precipitation and quickflow confirms the significant impact of precipitation on surface runoff. The average annual quickflow in the Neyshabur watershed is 34.3 mm. The highest quickflow occurs in sub-watershed 1, and the lowest in sub-watershed 4. Soils with low permeability in sub-watershed 1 lead to increased quickflow, while more permeable soils in sub-watershed 4 reduce it. Regarding water yield, the annual average water yield of the watershed is 43.4 mm. The highest water yield is reported in sub-watershed 1, while the lowest is in sub-watershed 5. These differences are mainly attributed to the climatic conditions and land-use types. The average erosion rate in the Neyshabur watershed is 0.6 t ha⁻¹ yr⁻¹. The highest erosion occurs in sub-watershed 3, and the lowest in sub-watershed 4. Areas with steep slopes and intense rainfall are more prone to erosion. The most important finding regarding soil retention is that the soil retention rate in all sub-watersheds exceeds the erosion rate. Sub-watershed 3 has the highest soil retention, while sub-watershed 4, with the lowest soil retention, requires more protection. Forest, shrubland, and scrubland land uses have the greatest capacity to retain soil, while saline and marshy lands and residential areas play the least role in soil retention. 

Determining the hydrological ecosystem services of watersheds is crucial for better and more targeted watershed management. This study was conducted to spatially analyze these services in the Neyshabur watershed in Razavi Khorasan Province. In the northern and northeastern parts of the watershed, higher water yield is observed, which could be considered for water resource exploitation and flood control measures in these areas. Regarding soil retention, the results indicate that sub-watershed 3, with the highest soil retention, has high potential for preventing erosion. In contrast, sub-watershed 4, with the lowest soil retention, has limited potential to prevent erosion. Overall, the findings highlight the importance of evaluating the spatial pattern of hydrological ecosystem services in watershed management.

Land subsidence, a significant geological hazard, poses widespread risks to infrastructure, agriculture, and the environment. This phenomenon may result from factors such as excessive groundwater extraction, mining activities, oil/gas extraction, or natural causes like sediment compaction and tectonic movements. Accurate monitoring of land subsidence using advanced technologies is essential for mitigating its adverse effects. Interferometric Synthetic Aperture Radar (InSAR), particularly the Persistent Scatterer InSAR (PSInSAR) technique, is one of the most advanced methods for monitoring ground deformation. With millimeter-level precision, it enables the detection of subtle land movements and serves as a critical tool for long-term subsidence monitoring over large areas. In this study, Sentinel-1 satellite data (ascending, descending, and combined modes) and the PSInSAR technique were utilized to assess and map land subsidence risk in major watersheds of Isfahan Province, including Isfahan-Borkhar, Najafabad, Northern Mahyar, Southern Mahyar, and Kuhpayeh-Sejzi. By providing detailed insights into the extent and severity of subsidence, this research identifies spatial and temporal patterns, offering crucial information for policymaking and risk management. 

The study area encompasses extensive watersheds that include the metropolis of Isfahan, as well as significant agricultural lands and residential areas. The Sentinel-1 radar data spanning from 2014 to 2023 were used, including ascending and descending imagery, to resolve displacement ambiguities caused by the directional nature of movement. Initial data processing involved co-registration of radar images to align pixels accurately and generate interferograms for phase change extraction. Persistent scatterers (PS) were identified using the Amplitude Dispersion Index (ADI) and phase stability analysis. Atmospheric and orbital errors were corrected using statistical models and inversion techniques to eliminate biases. Temporal analysis of ground displacement was conducted to calculate deformation trends, with data georeferenced for spatial interpretation. Validation was carried out by comparing results with ground-based data and independent sources. Final outputs included cumulative subsidence maps, annual subsidence rates, and risk zoning maps highlighting areas prone to land subsidence. 

The findings reveal that subsidence in the study area ranged from negligible levels to 55 cm over the nine-year observation period. Annual subsidence rates in parts of the Isfahan-Borkhar and Southern Mahyar watersheds reached 60 mm per year. Combining ascending and descending data improved accuracy and enabled the separation of vertical and east-west horizontal displacement components. The highest cumulative subsidence was observed in urban and agricultural zones of the Isfahan-Borkhar watershed and in clayey sediment areas within the Southern Mahyar watershed. Risk zoning maps indicate that the Isfahan-Borkhar and Southern Mahyar watersheds have the largest areas classified as high-risk. The other watersheds predominantly exhibit moderate to low-risk zones. The maps demonstrate a strong correlation between severe subsidence and land use (urban and agricultural areas) as well as geological features (clayey sediments and alluvial deposits). 

The application of PSInSAR for monitoring land subsidence in Isfahan Province provides valuable insights into the patterns and trends of this phenomenon. The results highlight severe and ongoing subsidence in the Isfahan-Borkhar and Southern Mahyar watersheds, necessitating urgent planning and management measures to mitigate the associated risks. The link between subsidence and anthropogenic factors, such as excessive groundwater extraction, and geological characteristics underscores the need for integrated planning for sustainable water resource management and land use. The risk zoning maps presented in this study serve as essential tools for policymakers and urban planners to optimize risk management strategies. Future research should focus on continuous monitoring and the development of predictive subsidence models to address this issue effectively.

The complexity of aquatic systems, their spatiotemporal variations, the high cost and time-consuming nature of traditional testing methods, and the need for continuous monitoring all contribute to the difficulty of monitoring and evaluating water quality. Therefore, artificial intelligence, machine learning, and deep learning approaches are useful for predicting water quality given the diverse parameters involved and are more cost-effective compared to traditional testing methods. There is no uniform algorithm that performs optimally for predicting water quality, and different algorithms exhibit superior performance in different contexts. The quality of rivers in the Zagros mountainous region has decreased due to the erosion of karst formations, their passage through residential areas, changes in land use, drought, and climate change. Some of these rivers, such as the Karkheh, Karun, and Dez, serve as the source of drinking water for a population of over 10 million people. Therefore, changes in water quality pose increased risks and threats to the drinking water supply of these settlements. Assessing the quality of water resources and developing a suitable model will play an effective role in managing the required water supply. The present study aims to identify the best machine learning algorithm for estimating the water quality parameters of the Cham Anjir watershed while facilitating ongoing monitoring of the resource. 

Data from 321 samples of discharge and water quality parameters at the Cham Anjir hydrometric station during the period 1969–2021 were used to select a suitable artificial intelligence model and to assessthe quality of surface water resources in the Cham Anjir watershed. These samples included physical and chemical indicators: TDS, TH, SAR, Na, Mg, Ca, and Cl. Machine learning algorithms were employed to model the water quality of the Cham Anjir watershed. Through iterative testing of different models, the Support Vector Machine (SVM) and Classification and Regression Tree (CART) models were selected as the best performers. A correlation matrix was used to evaluate relationships among the variables, and based on these correlations, monthly discharge and monthly water quality indices—including TDS, TH, SAR, Na, Mg, Ca, Cl, EC, %Na, pH, HCO₃, and temporary hardness—were analyzed. A total of 321 monthly samples of the two key indices, TDS and TH, were studied over the statistical period of 1969–2021. To evaluate the accuracy of the machine learning algorithms in water quality modeling, the following performance indices were used: coefficient of determination (R²), mean squared error (MSE), and mean absolute error (MAE).

The p-value from trend tests confirmed the existence of an increasing trend at the 95% confidence level for the two variables TDS and TH (surface water quality parameters of the Cham Anjir watershed) since 1985. The average TDS in the first period (1969–1984) was 286.6 mg/L, and in the second period (1985–2021) it was 422.08 mg/L. The average TH in the first period (1969–1984) was 181.5 mg/L, and in the second period (1985–2021) it was 278.6 mg/L.The results of the correlation matrix indicated that TDS has a strong correlation with EC and TH. TH has a positive correlation with TDS, EC, HCO₃, Ca, Cl, and Mg , and an inverse correlation with pH. The machine learning algorithms SVM and CART successfully captured the increasing trend for the two parameters of total hardness and total dissolved solids in the discharge of the Cham Anjir watershed. The SVM algorithm with the linear kernel exhibited the best performance. In addition, the root mean square error (RMSE) validation index also showed lower values for the SVM algorithm compared to the CART algorithm. Therefore, SVM is more accurate in predicting water quality indicators. 

Machine learning algorithms are effective for water quality modeling. The results indicated an increasing trend in TDS and TH concentrations in the Cham Anjir watershed since 1985. SVM performed better than CART in predicting TDS and TH. The findings suggest that decreasing river flow, increasing water consumption, and karst geology influence TH and TDS concentrations in the quality of this watershed. Water quality monitoring is essential for water resource management. For future estimates of water quality in this context, the SVM algorithm is recommended.