Quantitative assessment of irrigation water and organic/inorganic amendment on biometric growth profiles of Abelmoschus esculentus and Solanum lycopersicum and their varieties

Excessive chemical input in agricultural fields causes a variety of problems, including nutrient loss, soil basification, salinity, decreases in beneficial microflora, and surface/groundwater pollution (Banerjee et al. 2011; Garai et al. 2014; Han et al. 2016). The influence of organic agriculture on natural resources promotes interactions within the agro-ecosystem, which are crucial for both agricultural productivity and environmental protection. The current trend is to investigate the prospect of replacing chemical fertilizers with organic fertilizers which are both environmentally benign and cost-effective (Mondal et al. 2017). Vermicompost is a viable alternative to chemical fertilizer. Currently, 69.9% of India's population lives in villages, where cattle rearing is a common practice (Gupta et al. 2016). Livestock and its byproducts are essential components of agriculture. As a by-product, traditional farmyard manure (FYM) is widely used in agriculture across the world. The use of vermicast, as well as FYM and dried leaves, is a contemporary emerging bioresource strategy in agriculture, if earthworms are introduced in this mixture as a bioreactor which converts normal manure into vermicompost by ingestion and fragmentation. This is accomplished by breaking down the nutrients in the manure. It has the distinct features of a soil conditioner as well as a biocontrol agent (Ievinsh 2011). Vermicompost contains a consortium of microbiota including Pseudomonas oxalaticus (Pathma & Sakthivel 2012), Rhizobium japonicum, and Pseudomonas putida, Azospirillum, Azotobacter, Nitrobacters, Nitromonas, and Ammonifying Bacter (Joshi et al. 2015; Srivastava et al. 2021). Vermicompost is more nutritive than typical FYM and normal chemical fertilizer due to its broad range of consortia and high amount of total nitrogen, total carbon, accessible phosphorus, exchangeable potassium, exchangeable sodium, and magnesium (Joshi et al. 2015).

Apart from fertilizer, irrigational water quality is also important for seed germination and plant development (Rifna et al. 2019; Singh & Singh 2020). Seed germination is a determining structure that influences further plant development, plant structure maintenance, metabolism, germplasm creation, and production capability (Hussain et al. 2018; Rupani et al. 2018). Dry dormant seed absorbs nutrients from irrigation water and develops an embryonic axis during the germination stage (Bewley et al. 1994). Polluted irrigation water has a negative impact on seed germination and seedling vigour index and causes phytotoxicity in plant metabolism as well (Kothari et al. 2022).

The alluvial plains are considered as the cradle of civilization. River ecosystems serve provisioning, regulatory, and cultural services to local residents (Singh & Singh 2020). As anthropogenic pressure increases in the river basin, the river condition deteriorates (Singh et al. 2023). Because of human atrocities, the quality and quantity of the Ganga and its tributaries are steadily deteriorating (Dwivedi et al. 2018). Increasing biological/biochemical oxygen demand, load of organic matter, salinity, heavy metal, pesticide, polyaromatic hydrocarbon, polychlorinated biphenyl, and endocrine disruptors in the Ganga have a detrimental impact on water quality and, indirectly, crops that rely on river water (Singh et al. 2023).

Numerous experiments have been previously conducted on seed germination, phytotoxicity test, germination index (GI), germination rate percent (Kanyatrakul et al. 2020), plumule and radicle length, and vigour index (Rupani et al. 2018; Baruah et al. 2019). Some research has supported the usefulness of vermicompost in seed germination, while others claim an antagonistic effect of increasing vermicompost concentration (Ievinsh 2011; Suthar & Sharma 2013) (Table 1). Several studies have also found that combining vermicompost with chemical fertilizer improves yield production and seed germination (Atteya et al. 2021) (Table 1).

The present research design is an attempt to sustainably explore the integrated and particular amendment of organic/inorganic fertilizer as well as various irrigational water activities on seed germination and allied tests. We hypothesized that the mixture of the recommended dose of NPK and a high amount of vermicompost with irrigation water responds better in terms of viability and germination. The current research was conducted with specific objectives: (i) To evaluate the efficacy of nutrient-laden irrigation water, vermicompost, and conventional chemical fertilizers individually and in combination on seed germination and allied parameters of Abelmoschus esculentus and their varieties Kashi Bhiarav, Kashi Mahima and Solanum lycopersicum along with their two different varieties Vikash and Rakshak. (ii) To figure out how different amendments affect the seed vigour index, phytotoxicity, relative seed germination, and their indices. (iii) To quantitatively estimate the biometric growth profile using ANOVA, multivariate analysis, and neural network. The use of neural networks could enhance the accuracy and efficiency of data analysis and prediction, which could facilitate more effective decision-making for sustainable agriculture.

There are five variables involved in the experiment: (1) seed variety, (2) water quality, (3) soil properties, (4) sampling sites, and (5) amendments.

A seed germination experimental setup has been established in the Waste Management, Resource Recovery, & Ecotoxicology (WRE) laboratory, Department of Environment and Sustainable Development, Institute of Environment and Sustainable Development, Banaras Hindu University, Varanasi, Uttar Pradesh. A. esculentus with its two varieties Kashi Bhiarav and Kashi Mahima), and S. lycopersicum with its two varieties, Arka Vikas and Arka Rakshak, have been used for experimentation. Seeds were generated through heterosis breeding (F1 hybrid). A. esculentus and S. lycopersicum are economical and popular vegetable crops among the global population, especially in low-income countries. They have ample nutritional potential and medicinal benefits such as detoxification, an antifatigue effect, etc., increasing the demand for A.esculentus and S. lycopersicum among common people (Elkhalifa et al. 2021). Because of its nutraceutical significance and easy cultivation requirements, we have selected a particular plant for the current experiment. Both chosen hybrid varieties of A. esculentus were prepared by IIVR, Varanasi. S. lycopersicum varieties were also hybrids, Arka Vikas and Arka Rakshak. The ICAR-Indian Institute of Horticultural Research has developed both.

Water samples were collected from the two sites, i.e., river Ganga and groundwater which was 1.3 km away from Ganga at Varanasi in May 2021. The river samples were collected in triplicate from the depth of 0.5–0.7 m from the surface of the river and were transferred into a clean High-Density Poly Ethylene bottle and brought to the WRE laboratory in an ice box to perform all the quick assessment of water. A controlled sample was collected from the hand pump as well as the tap. Analysis of physio-chemical properties of collected irrigation water from two selected sites including pH, electrical conductivity (μS), salinity (ppt), total dissolved solid (ppt), and temperature (°C) has been measured using a water parameter analyzer (Model 371 Systronics kit), and phosphate (mg/L), nitrate (mg/L) and chemical oxygen demand (COD) (mg/L). All the experiments were determined using the APHA 2017 standard protocol. All the collected water samples from two different sites were digested in the presence of nitric acid and used for micronutrient (Cr, Cd, Zn, Cu, Mn, Fe, Ni) and macronutrient potassium analysis by an atomic absorption spectrophotometer (Perkin Elmer, USA).

Two kilograms of soil mixture has been prepared using groundwater/Ganga water-irrigated soil and water as well as inorganic/organic fertilizer. Three different amendments of fertilizer, i.e., recommended dose of inorganic fertilizer (urea (N) + diammonium phosphate (P)+ muriate of potash (K)) applied at 120-60-60 kg/ha for S. lycopersicum varieties (ICAR-IIHR), 120-60-60 kg/ha for A. esculentus (IIVR, Varanasi), 3 ton/ha vermicompost and combination of both NPK and vermicompost thoroughly mixed with irrigated field soil (groundwater and Ganga River water-irrigated soil). Calculation of fertilizers has been done on the basis of 2 kg soil preparation for the lab scale experiment. After mixing amendments in soil, it was kept for 7 days for soil stabilization and moisture was maintained with their respective irrigation water, i.e. groundwater and Ganga water (Figure 2).

The physicochemical parameters of all the soil samples were examined after being air dried, pulverized, and sieved through a 2 mm screen. For pH, electrical conductivity, and salinity measurement, a 1:5 (w/v) suspension has been prepared and agitated vigorously for about 2 h. After shaking, the suspension was left overnight. The next day, the supernatant was carefully poured into another glassware and was used for measurement with the calibrated multiparameter analyzer (Model 371 Systronics kit). Other soil nutrients, such as soil organic carbon (SOC), were determined using titration methods by Walkley and Black (Allison 1973), and total phosphorus (TP) was determined with the triacid digested sample using stannous chloride and ammonium molybdate (Jackson 1958; Allen et al. 1974). All the acid-digested samples were further used for macronutrient analysis (Na, K, Ca) with a flame photometer (Model 128 Systronics) and micronutrient analysis (Cd, Cr, Cu, Mn, Zn, Ni) analyzer with the atomic absorption spectrophotometer (PerkinElmer, USA) (Supplementary material, Table S4). Heavy metal concentration in groundwater, river water, inorganic/organic, or a combination of both fertilizer extractants is used in seed germination (Bhiarav, Mahima, Vikas, Rakshak) for the present experiment.

A complete mixture of fertilizer, soil, and irrigation water has been used for the preparation of the germination extract. Preparation of the extract involved distilled water and soil in appropriate proportion (1:5, w/v) using the Zucconi and de Bertoldi method (1987; Hussain et al. 2018; Gusain & Suthar 2020). In order to separate the phases, the extracts were vigorously stirred for 3 h at room temperature and then centrifuged for 15 min at 3,000 rpm. The supernatant was filtered using a Whatman No. 1 filter paper. Six seeds of both A.esculentus (A. esculentus) varieties and eight seeds of both S .lycopersicum varieties were placed in Petri dishes with a 90 × 15 mm diameter, a filter paper lining, and 5 ml of each extract (two extra seeds of S.lycopersicum varieties were placed because of their small size). A similar arrangement with 5 ml of distilled water served as control and all the plates were prepared in triplicate. Before placing Solanum and Abelmoschus seeds in a Petri plate for germination, they were washed thoroughly with 0.1% mercury chloride and then the seeds were soaked in their respective suspension for 6 h (Kannan & Upreti 2008). A total of 84 petri plates have been placed, including a combination of fertilizer and irrigation water and individual irrigation water for both sites, along with controlled plates of distilled water for both varieties. Plates were incubated at 23 ± 2 °C in 12 h light (364 lumens per square meter) and 12 h dark. The number of germinated seeds is measured every 24 h to assess the germination rate. The germination rate was recorded, till no new seeds germinated for the next 3 days. Calculation of cumulative germination percentage depends on the emergence of the radicle at least 1 mm.

The datasets were statistically analyzed using SPSS version 26 (Illinois, USA) software. The ANOVA and DMRT (Duncan's Multiple Range Test) were used to compare the significant differences among amendments for studied parameters (Supplementary material). Sigma Plot software (version 15) was used for plotting graphs. Principal component analysis (PCA) (to reduce dimensionality) and correlation was carried out using open source software R (version 4.0) to evaluate responses concerning different fertilization treatments. Deep neural network (networks composed of several layers, usually two or more, that include input, output, and at least one hidden layer in between) analysis was performed using R.

In this experiment, high-quality analytical grade (A.R.) chemicals purchased from SRL, Himedia, and Merck were used without the need for further purification. To ensure quality control, standard operating protocols were employed to calibrate, estimate, and recover known spiked specimens, followed by the triplicate analysis of all the samples. Additionally, the accuracy and precision of the metal estimates were verified through repeated analyses according to the National Institute of Standards and Technology's Standard Reference Material (NIST SRM 1570), and the reported results were within 2% of the certified value.

There are two different samples, Ganga River water sample and groundwater (Table 2), used for the present experiment. pH was observed in groundwater (7.34) and the Ganga water sample (7.25). The temperature was found to be 32.1 and 28.9 °C, respectively, in groundwater and Ganga in May 2021. Salinity and total dissolved solids (TDS) have been found at 1.27 ppt, 1.5ppt, and 0.55ppt, 1.27ppt in both the sampled groundwater (gw) and Ganga River water, respectively. The value of electrical conductivity was found to be 0.49 μS in gw and 2.37 μS in Ganga water. Electrical conductivity (EC) is a helpful indication of TDS because the conduction of current in an electrolyte solution is mostly reliant on the concentration of ionic species (Hayashi et al. 2003). COD has been reported at 7.93 and 8.76 mg/L, nitrate 0.39 and 1.40 mg/L, phosphate 0.009 and 0.021 mg/L, potassium 40.48 and 144.36 mg/L, respectively in groundwater and Ganga in May 2021. Zinc and manganese have been observed at 0 and 0 mg/L at both sites. Cadmium has been observed at 0.05 and 0.05 mg/L, chromium was reported at 0.02 and 0.05 mg/L, nickel 0.35 and 0.36 mg/L, copper 0.003 and 0.007 mg/L, iron 0 and 10.96 mg/L in groundwater and Ganga water respectively. In addition, leachate from dumping sites may potentially cause surface water and groundwater contamination (Gupta et al. 2019). However, some of the contaminants are persistent owing to their non-biodegradability and lengthy biological half-life, e.g., heavy metals (Maurya et al. 2019). Pollution from home and industrial trash is high in Kanpur, Allahabad, and Varanasi (Singh et al. 2023).

The pH values range from 6.9 to 7.5 across the different samples NPKVCrw (7.5) > rw (7.4)> NPKVCgw (7.38) > VCgw (7.1)> VC (6.9) (Figure 4). A neutral pH is generally conducive to nutrient availability and microbial activity. Minimal pH was reported in VC only because of the presence of humic acid and fulvic acid in vermicompost (Molina et al. 2013). The electrical conductivity values vary significantly, ranging from 414 to 460 μS. Higher electrical conductivity often suggests higher ion concentrations in the solution, which can be related to nutrient levels or other dissolved substances (Möller & Müller 2012). The wide range of EC values might indicate variations in the composition of the samples, potentially due to differences in soil or water sources. The potassium (K) concentrations range from 12.96 to 23.18 mg/L. Maximum potassium was found in 23.18 mg/L in the combination of inorganic and organic in Ganga River (NPKVCrw) (Figure 3). A similar trend was also found in previous studies (Pandey et al. 2019; Mahajan et al. 2023). These variations could be attributed to differences in soil types, land use practices, or other factors affecting nutrient availability. K is an essential macronutrient for plant growth, and the concentrations observed in the samples could impact the fertility of the soil. The total kjeltron nitrogen ranges from 0.45 (ground water + irrigated soil)< 0.65(river water + irrigated soil)< 0.82(vermicompost)< 0.97 (vermicompost + ground water) < 1.02 (vermicompost + river water) < 1.05 (NPK + vermicompost + irrigated soil + groundwater) < 1.1997 (NPK + vermicompost + irrigated soil + river water). Vermicompost is rich in organic matter, which helps improve soil structure, water-holding capacity, and microbial activity. When combined with inorganic fertilizers, it creates a synergistic effect, enhancing nutrient availability to plants (Sharma & Garg 2018). Vermicompost contains beneficial microorganisms that can increase nutrient mineralization in the soil. These microorganisms break down organic matter and convert it into plant-available forms of nutrients (Lim et al. 2015). C/P ratios vary from 1.16 to 7.67. Balanced C/P ratios are important for efficient nutrient uptake by plants. The concentration of phosphate varies between 0.45 and 0.97 mg/L. The maximum concentration of phosphate is found in combination with organic and inorganic fertilizers with irrigated soil and river water because vermicompost can enhance nutrient retention in the soil, reducing the risk of nutrient loss through leaching or runoff and concentration of phosphate found more in river water (Table 2). This helps maintain higher nutrient concentrations. The concentrations of heavy metals such as chromium (Cr), cadmium (Cd), nickel (Ni), and zinc (Zn) have been found to be 0.03–0.97 mg/L. The heavy metal concentration reported in the germination extract is due to contaminated water and soil matrix (Table 2). These metals can have toxic effects on plants, animals, and the environment at elevated concentrations. The low values observed in the table indicate that these samples may not pose an immediate concern in terms of heavy metal contamination (Figure 3).

Based on the germination percentage in heavy metal-contaminated water and soil extract amended plates, the impact of soil properties on seed germination was ranked as silt loam > sandy loam > loam for Ni and Zn, and silt loam > loam > sandy loam for Cd (Zhao et al. 2021). These results support previous studies that emphasized the role of heavy metal type, concentration, and oxidation state, as well as soil and water properties, in heavy metal uptake, mobility, and translocation from water to soil as well as soil to plant tissues or cells (Sytar et al. 2013; Sutradhar et al. 2014; Emamverdian et al. 2015; El Rasafi et al. 2016; Seneviratne et al. 2019).

Based on the provided information, it seems that a neural network model was used to predict the Germination Rate Index (GRI) of plants based on various environmental and soil factors, such as seedling length, carotenoid levels, Total Chlorophyll levels, SOC, TP levels, TKN levels, pH, electrical conductivity, salinity, and heavy metal concentrations (Cr, Ni, Zn, Mn, Cd, Cu).

The structure of neural network (Figure 7), which is a type of machine learning model used for making predictions or classifications based on input data. The input layer of the network has 15 covariates, which are variables used to predict the dependent variable. These covariates include measurements related to seedling length, carotenoid content, chlorophyll content, SOC, TP, TKN, pH, electrical conductivity, salinity, and several heavy metals (chromium, nickel, zinc, manganese, cadmium and copper). The number of units in the input layer is also 15, which corresponds to the number of covariates. The covariates are standardized, which means they are scaled to have zero mean and unit variance, so that they are all on the same scale and have equal influence in the model. The network has one hidden layer, which contains two units. The activation function used in this layer is the hyperbolic tangent function, which transforms the inputs to produce outputs between −1 and 1. The purpose of the hidden layer is to learn and extract relevant features from the input data that are useful for making predictions. The output layer of the network has two units, corresponding to the two possible categories of the dependent variable, which is labelled as ‘GRI’. The activation function used in this layer is the softmax function, which transforms the outputs into probabilities that sum to 1. The error function used in this network is the cross-entropy function, which measures the difference between the predicted probabilities and the true probabilities and is used to optimize the weights and biases of the network during training. Most heavy metals in soils can be harmful to germination rate index and seedling growth due to their high concentrations. This is because they can interfere with various biochemical processes, such as enzyme and antioxidant production, protein mobilization, and photosynthesis. The negative effects can include growth retardation, chlorophyll destruction, biochemical activity disorders, mutations, and reproductive disorders in plants. The extent of growth inhibition of root length for sunflower seeds can vary depending on soil properties and the type and concentration of heavy metals. Given the complexity and harmful effects of heavy metals on root growth through various biochemical processes, it is inadequate to rely solely on root length to explain the phytotoxicity of heavy metals in different soils and its extract (Gall & Rajakaruna 2013; Emamverdian et al. 2015; Seneviratne et al. 2017; Kothari et al. 2022).

The given data set summary provides information about the performance of a trained neural network model. The model has been trained to predict the dependent variable ‘gri’ using 15 input covariates. The training phase of the model resulted in a Cross-Entropy Error of 21.687 and a percent incorrect prediction of 20.4%. The Cross-Entropy Error is a measure of how well the model is able to predict the correct class for each input sample, with lower values indicating better performance. The percent incorrect predictions is the percentage of samples in the training set that were incorrectly classified by the model. During the training process, a stopping criterion was implemented, and the model was trained until there was no further drop in error for a single consecutive step. This aids in mitigating overfitting and ensures that the model is not excessively trained, which could result in poor generalization performance. The model was additionally assessed on an independent testing dataset to evaluate its performance on this particular dataset. The cross-entropy error during testing was 3.856, which is considerably lower than the training error. This suggests that the model has a strong ability to generalize to new data. The testing dataset had a 3.3% error rate, which was much lower than the training error, suggesting strong generalization performance. Overall, the model demonstrated strong performance on both the training and testing datasets, with few mistakes and a low percentage of inaccurate predictions.

The classification results for the dependent variable ‘GRI’ are presented in Table 3. The table provides information on the observed and predicted classifications for the training and testing samples, as well as the percentage of correct predictions for each sample.

In the training sample, all observations of ‘GRI_Fair’ were correctly predicted, while only 8.3% of the ‘GRI_Good’ observations were correctly predicted. The overall percentage of correct predictions for the training sample was 79.6%. In the testing sample, all observations of ‘GRI_Fair’ were correctly predicted, while none of the ‘GRI_Good’ observations were correctly predicted. The overall percentage of correct predictions for the testing sample was 96.7%. These results suggest that the model is better at predicting the ‘GRI_Fair’ category compared to the ‘GRI_Good’ category. However, the overall percentage of correct predictions is relatively high for both the training and testing samples, indicating that the model is performing well in predicting the dependent variable ‘GRI.’ Further analysis may be needed to improve the model's performance for predicting the ‘GRI_Good’ category.

Area Under the Curve (AUC) is a measure of the ability of a classification model to distinguish between different classes. In this study, the AUC was calculated for the dependent variable ‘GRI’ with two levels, GRI_Fair and GRI_Good (Table 4). The AUC value for GRI_Fair was 0.883, indicating that the model had good discrimination ability in distinguishing samples with a GRI_Fair rating from those with a GRI_Good rating. Similarly, the AUC value for GRI_Good was also 0.883, suggesting that the model performed well in identifying samples with a GRI_Good rating. Overall, the high AUC values indicate that the classification model used in this study was effective in distinguishing between the two levels of the dependent variable ‘GRI’.

Table 5 presents the importance of independent variables in predicting the dependent variable. The normalized importance of each independent variable is also provided. The independent variables include seedling length, carotenoid, total chlorophyll, SOC, T.P., TKN, pH, E.C., salinity, Cr, Ni, Zn, Mn, Cd, and Cu. The results indicate that T. Chlorophyll is the most important predictor of the dependent variable, with a normalized importance of 100.0%. Carotenoids, Cd, and Ni also appear relatively important, with normalized importance values of 75.5, 86.5, and 49.8%, respectively. The other variables have lower importance values ranging from 9.5 to 42.1%. The findings suggest that T. Chlorophyll, carotenoid, Cd, and Ni are the most influential variables in predicting the dependent variable and should be considered in future analyses.

Finding all the analysis of the biometric growth profile of vegetable plants and their varieties A. esculentus (Kashi Bhiarav, Kashi Mahima) and S. lycopersicum (Arka Vikas, Arka Rakshak) in two different irrigation water sources (groundwater and river water) individually and in combination with the suggested dosage of NPK and vermicompost and vermicompost individual suggested that every plant variety has its own specificity and preferences. Among all the implementations of different seed germination extracts, river water provided a better result in the Bhiarav variety but in Mahima, vermicompost with ground water performs better. In the case of S. lycopersicum Vikas and Rakshak variety performs better in slow releasing mixture of vermicompost and ground water but sensitive to less nutritive extract ground water individually. Instead of directly amending any inorganic/organic fertilizer in eco sensitive riparian zone petri plate germination the test result gives the evident result of different plants and varietal response with amendment. The output of neural network analysis indicates that the model's predictive accuracy is higher for the ‘GRI_Fair’ category in comparison to the ‘GRI_Good’ category. Despite this, the model achieves a relatively high level of accurate predictions for both the training and testing datasets, suggesting satisfactory performance in predicting the dependent variable ‘GRI.’ Nevertheless, additional analysis may be required to enhance the model's ability to predict the ‘GRI_Good’ category.

Further research is required to investigate the impact of different irrigation and fertilization practices on soil health and microbial communities. Soil health is crucial for maintaining long-term productivity and sustainability of agricultural systems. Understanding the interactions between plant growth, soil health, and microbial communities can help in developing more integrated and holistic approaches to crop management.

The authors are thankful to the Dean & Head, DESD (Department of Environment and Sustainable Development) and Director, Institute of Environment and Sustainable Development, Banaras Hindu University, for providing needed facilities. Rajeev Pratap Singh is grateful to the authorities of Banaras Hindu University, Varanasi for providing support under the Council of Scientific and Industrial Research (CSIR) scheme under scheme No. P025/0880

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.