machine learning algorithms

Monitoring and controlling the level and sources of dust are crucial in the face of climate change and the development of suitable predictive approaches that directly impact the environment and human health. This study aims to estimate the concentration of PM10 in the city of Ahvaz using various machine learning models. Climate variables and the Aerosol Optical Depth (AOD) index, derived from the MODIS sensor at a wavelength of 476 nanometers, were used as influential variables in estimating PM10 concentration in three scenarios: combining AOD with PM10 (scenario 1), combining climate variables with PM10 (scenario 2), and combining climate variables and AOD with PM10 (scenario 3) .Using six machine learning algorithms, namely Random Forest Regression (RFR), Gradient Boosting Regression (GBR), Artificial Neural Networks (ANN), AdaBoostR with DTR, Support Vector Regression (SVR), and Decision Tree Regression (DTR), the PM10 concentration was estimated in different scenarios, considering accuracy and precision coefficients. The most influential variables in estimating PM10 concentration were determined to be sunshine hours, minimum visibility, maximum wind speed, and the AOD index. The GBR linear regression model, with R2, MAE, RMSE, and IOA coefficients of 0.76, 0.31, 0.49 and 0.93 respectively, was found to be the most suitable model for estimating PM10 concentration in scenario 3. the results showed that incorporating the AOD index alongside climate variables improved the model's performance in estimating PM10 concentration. The proposed final model can be used for daily estimation of PM10 particles.

Water quality assessment is paramount for various sectors, including environmental planning, public health, and industrial operations. With the increasing importance of ensuring safe water sources, especially for drinking and irrigation purposes, modern methodologies like data mining offer valuable tools for predictive analysis and classification of water quality. Knowledge of water quality is considered one of the most important needs in planning, developing, and protecting water resources. Determining the quality of water for different uses, including irrigation and drinking in different areas of life. The use of modern data mining methods can be beneficial for predicting and classifying the quality of provider water. In the current study, the water quality of the Qizil-Uzen River was evaluated at Qara Gunei stations. In this regard, the drinking water quality index (WQI) using the chemical compounds of glass hardness, alkalinity (PH), electrical conductivity, total dissolved substances, calcium, sodium, magnesium, potassium, chlorine, carbonate, bicarbonate and sulfate in the statistical period of 21 years (2000-2020) was estimated. Water quality assessment is paramount for various sectors, including environmental planning, public health, and industrial operations. With the increasing importance of ensuring safe water sources, especially for drinking and irrigation purposes, modern methodologies like data mining offer valuable tools for predictive analysis and classification of water quality.

Due to the relatively large number of variables, principal component analysis and independent component analysis methods were used to reduce dimensions, and then different machine learning algorithms including decision tree, logistic regression, and multi-layer perceptron artificial neural network were used to model the water quality index. By using these methods, the number of parameters needed to calculate the quality index was reduced from 12 to 2. Reducing the dimensions of the data saves the time of sampling, monitoring the samples, and determining the quality of the water and reduces the costs required for modeling to a significant amount. The results showed that among the dimensionality reduction methods, the principal component analysis method can perform better than the independent component analysis method. In the current research, the WQI index was modeled using machine learning algorithms including decision tree, logistic regression, and artificial neural network method. The quality of water in the Qizil-Uzen Qara Gunei river station has been evaluated. Then, to estimate the numerical values of the WQI index, TH, pH, EC, TDS, Ca, Na, Mg, K, Cl, CO3, HCO3, and SO4 parameters of the mentioned station in the statistical period of 21 years (1378-1398) were used. PCA and ICA methods have been used to select different input parameters. Modeling has been done in a Python programming environment. Among the available samples, 75% are considered for training and 25% for testing.

In the present research, to model the water quality index in the first stage, different dimensionality reduction methods such as PCA and ICA were used to reduce the time and cost of implementation. In the second stage, machine learning methods such as decision tree, linear regression, and multilayer perceptron were used. In the method used by Tripathi and his colleagues, by using the principal component analysis method, they reduced the number of parameters needed to calculate the quality index from 28 to 9 and calculated the water quality index with the number of 9 parameters. Examining the two methods of PCA and ICA has reduced the dimensions of the problem from 12 dimensions to 2 dimensions. The results show that the PCA method can help us improve performance with little cost and high accuracy. Because of the PCA dimensions. The comparison of the results of the models was done using different numerical and graphical evaluation criteria, including R2, RMSE, and modified Wilmot coefficient as numerical criteria and Taylor diagram as graphical criteria. Because the PCA algorithm can help reduce noise in data, feature selection, and generate independent and unrelated features from data. The results show that multi-layer perceptron, decision tree, and logistic regression methods accurately perform the water quality index. In this research, for the first time, using the ICA dimension reduction algorithm, while reducing the dimensions of the problem, the water quality index is predicted with an accuracy of over 90%.

Water quality index modeling holds significant relevance in agricultural practices, where access to clean water is crucial for irrigation and crop growth. Surprisingly, only a limited number of studies have explored variable reduction methods in water quality index modeling, with none incorporating the relatively novel Independent Component Analysis (ICA) method for dimensionality reduction. Thus, the current research fills this gap by employing PCA and ICA techniques to reduce the dimensionality of large datasets in water quality index modeling. By utilizing these advanced methods, the study aims to enhance efficiency and accuracy in assessing water quality, thereby offering valuable insights for agricultural water management. Following dimensionality reduction, the dataset is then subjected to modeling using various machine learning algorithms. This approach not only optimizes computational resources but also facilitates a deeper understanding of the complex interrelationships among water quality parameters. Through this pioneering research endeavor, the efficacy of ICA alongside PCA in addressing water quality index modeling challenges is evaluated. By integrating these techniques with machine learning methodologies, the study endeavors to provide actionable intelligence for agricultural stakeholders, aiding in informed decision-making and resource allocation. Moreover, by venturing into unexplored territory with the inclusion of ICA, the research contributes to expanding the methodological toolkit available for water quality assessment. As agriculture faces increasing pressure from climate change and resource scarcity, such innovative approaches hold promise in ensuring sustainable water management practices.

In recent years, the use of machine learning algorithms has been a promising way to improve crop yield predictions, especially when using non-linear relationships. This research was conducted with the aim of evaluating the yield estimation of irrigated and rainfed barley as well as the total yield of barley produced in the provincial centers of Iran using remote sensing data and machine learning methods including XGBoost and SVM. The results showed that by using climatic data, drought indices and plant indices of remote sensing and also XGBoost and SVM algorithms, it is possible to reliably estimate barley yield in different regions of the country with different climates. In general, the RMSE error obtained for both models was acceptable (0.41 and 0.77 t/ha). The R2 determination coefficient values for XGBoost and SVR algorithms in modeling rainfed barley cultivation performance were equal to 0.2 and 0.22 respectively, in irrigated barley performance were equal to 0.52 and 0.55 and for total barley were equal to 0.66 and 0.65, indicating that the rainfed yield modeling were not as suitable as irrigated and total barely yield modeling. The RBF kernel was chosen as the best kernel to use for the SVM algorithm. Also, in this research, while examining the effects of change in train and test data dividing, the parameters of precipitation, temperature, and evapotranspiration were determined as the most important parameters affecting the performance of the barley yield for both algorithms in different evaluated conditions.

Spatiotemporal estimation and monitoring of soil moisture based on remote sensing observations (optical and thermal) is challenging due to its physical nature in high vegetation conditions, necessitating improving and increasing the accuracy of soil moisture estimation in these areas. Therefore, this research aimed to develop a new approach to estimating surface soil moisture in agricultural fields with dense vegetation using machine learning algorithms by incorporating optical and thermal remote sensing data and soil physical properties. For this objective, 16 Landsat-8 satellite images and more than 430 control locations were used during the sugarcane crop’s growth period in 2018-2019 at the Hakim Farabi Sugarcane Agro-Industrial company in the Khuzestan province of Iran. A set of 10 scenarios of various unique combinations of the available input variables were developed and then evaluated by five machine learning algorithms, including multiple linear regression (MLR), decision tree-based algorithms (CART and M5P), and ensemble learning-based algorithms (gradient-boosted regression trees (GBRT) and random forest regression (RFR)). According to the results, the highest correlation between input variables and surface soil moisture was observed in Soil Wetness Index (SWI) and Normalized Soil Moisture Index (NSMI) with R values of 0.79 and 0.69, respectively. Also, the highest accuracy of machine learning algorithms based on R2, RMSE, and MAE results was obtained in GBRT (0.99, 0.011, and 0.006) and RFR (0.99, 0.014, and 0.007), respectively. In general, the findings of this research show the importance of using variables based on Landsat-8 remote sensing data in combination with ensemble learning algorithms that can be independent of any ground measurements.