The CLD contains 30 variables and 49 arrows. The stakeholders identified 7 key driving components - groundwater level, working well count, financial resources, rainfed crop, forest area, less water intensive ZBNF crop, and soil fertility or soil productivity. Working wells are those which are yielding water; a count of such wells is called working well count. Rainfed crops are those which are watered by rains and not irrigated. Forest area is the land that is covered by forest. Financial resources are the sources of money for the farmer including farm income, bank loans, government schemes, sale of personal property, and debt. Groundwater is the water available underground, extracted by digging wells. Soil fertility or soil productivity is the ability of the soil to supply nutrients to the crop and the absence of toxic substances. In this work, we have used this term loosely, without specific definitions, only to communicate to the stakeholders, and we have taken the quality of the soil prior to use of artificial fertilizers as 1.0 and thereafter depreciated it using a crop-based heuristic obtained by participatory methods. ZBNF, is a method of farming where the cost of farming is zero because the manure, seed treatment, and soil treatment are done by natural and traditional methods from products of the farmer's cattle. This means that farmers need not purchase fertilizers and pesticides in order to ensure the healthy growth of crops (Duddigan et al., 2022).
The CLD in Fig. 4 contains three reinforcing or positive feedback loops R1-R3 and five balancing or negative feedback loops B1-B5. Loop R1 shows the interdependence between working well count, income from irrigated/ horticultural crops, and financial resources. An increase in working well count leads to an growth in the irrigated area yielding more crops, hence growth in income from irrigated/ horticultural crops. Eventually, this leads to growth in financial resources, which enable the farmer to dig more tube wells, increasing the working well count. In the balancing loop, B1 counteracts this reinforcing loop R1. An increase in crop area results in an increase in expenditure on irrigated/horticultural crops, thereby leading to a decrease in financial resources. In Loop R2, an increase in the less water intensive ZBNF crop area increases Soil Fertility/Soil productivity which increases the income from ZBNF crops.
In the balancing loop B3, Groundwater Resources are increased by an increase in Recharge, and decreased by increase in Discharge. It is also connected with Working Well Count. Any increase in the Groundwater Resources increases the Working Well Count, and any decrease in the Groundwater Resources decreases the Working Well Count. There are three variables identified as causes for recharge: Rainfall, Percolation, and Catchment Area.
Both the Crop area and Forest area are connected by balancing loop B5, while there is an increase in one leading to a decrease in the other. Loop R3 shows the interdependence between crop area, income from rainfed crops, and financial resources. An increase in crop area leads to an increase in the income from rainfed crops. This in turn, leads to growth in financial resources. Loop B4 counteracts R4, where increase in crop area results in increase in expenditure on rainfed crops thereby leading to a decrease in financial resources.
Next, we built an SFD as shown in Fig. 5, explained below:
-
Groundwater level stock: Rainfall over the catchment area and its percolation into the ground determine inflows ('recharge') into this stock. Outflows ('discharge') from it are a product of working wells count and how much they each draw/yield.
-
Working well count stock: Increases as their cost becomes affordable and the farmer sees the potential income from irrigated crops. Their count decreases as groundwater resources deplete.
-
Financial resources stock: Income/expenditure from/to three kinds of crops determine inflows/outflows of this stock: (i) rainfed crops, (ii) irrigated/horticultural crops, and (iii) ZBNF crops.
-
Rainfed crop area and Forest area stocks: Forest area is converted into cropland as land clearance cost becomes affordable and potential income becomes attractive.
-
Less water-intensive ZBNF crop stock: Urban consumers are becoming aware of the health hazards of consuming products yielded from chemical farming and are willing to pay a premium for ZBNF products. Such higher sale prices make it possible for farmers to meet high labor costs associated with ZBNF practices and still earn profits, leading to an increase in ZBNF adoption. Further, unless mechanization takes place, higher labor requirements and wages will lead to a decrease in ZBNF adoption.
-
Soil fertility stock: Soil fertility/soil productivity increases by less water-intensive ZBNF crops and decreases by irrigated/horticultural crops using chemical fertilizers.
From this SFD, we developed equations (Appendix B) to simulate the model. While almost all stakeholders could understand, contribute to and own the CLD, most of them found the SFD more difficult to understand.
The information about wells and water level, yield, income, expense, soil productivity and other practices was collected through interviews, brainstorming, surveys, and workshops to run the model. In addition, government reports and scientific papers provided rain data and other technical information (Appendix C and D).
As a participatory approach, communicating the results to the stakeholders was a crucial criterion for selecting the platform for simulation. The output formats of open-source System Dynamics simulators like Modsim and Vensim were not suitable for our stakeholders. Therefore we chose to design and develop the code the algorithm of which is in Appendix E.
4.1 Model Validation
The data from 1970 to 2020 was used to build and validate the model by comparing the simulated output with the observed data. The plots in Fig. 6 show various parameters of interest plotted against time in years where Fig. 6 (a) is data validated, and the rest of the graphs were considered face valid by local stakeholders as well as stakeholders from NGOs and academic institutes. In Fig. 6(a), the simulated results correlate with observed data. Especially, the working wells count peaked around 1995 but fell around 2000 and remained low until 2015, when there was heavy rainfall. In Fig. 6(b), the simulated water dynamics were inferred by stakeholders from the water depth of their wells. As can be seen in Fig. 6(b), the groundwater depth shows a rapid fall around 2000, both due to lack of rainfall and the growing number of deep tube wells with which the discharge rate is very high. Shown in Fig. 6(c) is the farming area dynamics. More tube wells account for the growth of irrigated crop area due to mechanized withdrawal (as opposed to manual withdrawal from open wells). Some of the rainfed areas were also converted to irrigate until 1995. Further, the dry spell and exhausted groundwater of 1995–2015 resulted in almost all the farming areas going to fallow.
For understanding the income-related graphs in Fig. 6 (d) and (e), a little background is required on the nature of the irrigated crops and the rainfed crops that depend entirely on uncertain rainfall. Typically, farmers do not use chemical fertilizers in rainfed crops for two reasons:
-
Every application of fertilizer is followed immediately by plenty of water, or else the crop may die. Since rainfall is not in his control, fertilizers are not used.
-
The yield is totally dependent on rainfall, which might be very low if there is no rainfall. Hence the farmer doesn't invest in fertilizers.
Also, the changing trends in agriculture as described in section 3.2, contributed to the increase in the expenditure on farming. Specifically, they can be recapitulated here as follows:
-
family labor vs. paid labor – high perceived income with family labor; urban employment, and migration of youth created the need for paid labor
-
manual labor vs. mechanization – labor migrated due to lack of work when mechanical implements were used resulting in increased labor costs.
-
natural fertilizers vs chemical fertilizers – initial rise in yield absorbed the higher costs of fertilizers and showed an increase in income (the small peaks in Income in Fig. 6 (d)) but gradually, soil fertility/soil productivity decreased, leading to more use of fertilizers.
These trends can be seen in the farm income graphs, plotted using an approximated scaled value considering the income of the 1970s as the base. The income plot of Fig. 6 (d) shows small peaks around 1975, 1998–2000 and again towards the end of 2020. This is due to the farmer going for mechanization when the labor costs become a significant portion of the expenditure. The decrease in groundwater resources was substantial enough to make most wells stop yielding any water by about 2002, resulting in a sharp decrease in the financial resources of the farmers shown in Fig. 6 (e).
The plot of Fig. 6 (a) validated using data which is collected from stakeholders and Fig. 6 (b) to (e) have been validated using face validation. All the plots show that the simulated values match closely with the stakeholder perception or observed data.
4.2 Simulation - Prediction and What-if scenario analysis
The model was used to analyze five scenarios designed to visualize the future upto 2050, for different agriculture options
-
Scenario 1: Business As Usual (BAU) where the current trend of drawing groundwater through tube wells and use of fertilizers continues.
-
Scenario 2: Less Water-Intensive Crops (LWIC) where crops like vegetables and millets are grown to avoid overexploitation of groundwater.
-
Scenario 3: ZBNF where no chemical fertilizers are used, but only traditionally prepared manure and pesticides from available organic matter. This brings down fertilizer cost and restores the natural soil quality in the long run.
-
Scenario 4: GroundWater Regulation (GWR) where all stakeholders in the area decide to fix tube well depth up 300 ft. Deeper wells are closed.
-
Scenario 5: LWIC and ZBNF together
The average annual rainfall data of the past thirty years in the area, including a dry spell, was replicated for the simulation of the next thirty years, providing a realistic basis for scenario analysis.
The working wells count and water discharge rate are shown for the five scenarios in Fig. 7 (a) and (b). BAU leads to a drastic fall in the number of working wells. With GWR, deeper wells are not allowed and shallow wells don't yield water. ZBNF by itself is not useful and LWIC is the only useful scenario. In Fig. 7 (b), only LWIC are helpful in reducing the discharge, whereas BAU, ZBNF and GWR are scenarios of high water stress. Figure 7 (c) shows the total farming area, which will go down sharply in BAU, ZBNF, and GWR when there is no rainfall or groundwater, but will be a softer change in LWIC or LWIC + ZBNF confirming that any effort in reducing water discharge will sustain farming by keeping water available. The stakeholder's belief that only lack of rainfall leads to water scarcity is questioned by showing that even in drought, it is possible to conserve water.
The soil productivity inferred from the amount of chemical fertilizer added to maintain previous yield, is shown in Fig. 7 (d). It falls steadily except in ZBNF and LWIC + ZBNF scenarios. The simulation has been done starting with current soil condition ratio as about 0.6 and the productivity ratio of 1.0 is restored after a few years of using ZBNF. In Fig. 7 (e), the yield per unit area is shown. The aim of the farmer is to maintain a certain yield, which is accomplished by adding more fertilizer in BAU, GWR and LWIC scenarios, due to declining soil productivity. This yield gives the farm profit per unit area shown in Fig. 7 (f). Profit fluctuates sharply with lack of rainfall or lack of groundwater availability in all scenarios, but the effect is softer in LWIC + ZBNF scenario. But the overall profit level is quite meager.
The most important observation of the scenarios is that BAU, LWIC and GWR do not bring significant changes in profit or the groundwater situation. They are affected by rainfall, as in the past. In Fig. 7 (e) and (f), it appears that doing only ZBNF or ZBNF + LWIC give similar results, hence ZBNF only may be sufficient to maintain profitability. These two graphs are per unit area. Hence, from Fig. 7 (c), the farming area is to be considered, where it is clear that ZBNF only is similar to BAU. Farming area sharply reduces without groundwater or rainfall in both BAU and ZBNF. Summing up, the scenario ZBNF + LWIC is the only option where farming area can be maintained due to water availability, and profitability can be maintained due to reduction in chemical fertilizer cost. This can be seen Fig. 7 (g), where the profit from the total farming area in the case study village is plotted for each scenario. While the ZBNF + LWIC profit is the only one that never goes below 3000 units, all other scenarios go further down, even close to zero.
In the past, mechanisms like drip irrigation, sprinklers, and farm ponds subsidized by the government extended the availability of groundwater by 2 to 3 months. This benefited the crucial stage of the crop, thereby avoiding the labor and cost of fetching water from distant places by water tankers. Though the thrust of ZBNF is on chemical-free farming and LWIC, the water conservation methods employed therein, convinced the stakeholders that ZBNF has the potential for sustainable and profitable agriculture.
ZBNF crops present the only hope for the revival of agriculture as urban consumers trend toward chemical-free products, willing to pay more.
- If ZBNF products don’t fetch more revenue, then rainfed and irrigated crop areas will continue to stagnate at low levels.
- If ZBNF produce prices continue to rise as in the last 1–2 years, supported by the Government’s ongoing massive offline and online marketing efforts, the ZBNF crop area will increase substantially.
- Together with this, if the Government's ongoing mechanization efforts (e.g., of routine operations such as manure preparation) succeed in reducing the labor needed for cultivating ZBNF crops, then the ZBNF crop area will increase even more, and the overall financial health of the village economy will improve substantially.
- If however, the marketing and mechanization efforts do not succeed much, and higher ZBNF produce prices to keep pace barely with increasing labor wages, the ZBNF crop area is very unlikely to take off.
4.3 Sensitivity Analysis
The parameters for our model were provided by stakeholders (mostly empirical), or taken from scientific literature and government data, which may not be exactly specific to the study area. To check whether our model is robust and dependable, in spite of uncertainty in the parameters, a sensitivity analysis was carried out. It also helps to identify the parameters with the highest influence and to suggest better data collection methods. It provides confidence in the "what-if'' scenarios for generating awareness and future action plans.
We selected eleven input parameters - percolation, porosity, threshold depth, irrigated extent per well, irrigated farm yield, rainfed farm yield, yield per well, groundwater withdrawal rate, minimum rainfall, threshold rainfall, and maximum water depth and individually offset them by ± 10% of their values in the model. In each run, the value of the parameter being tested was offset keeping other parameters unchanged.
Farm profit and Groundwater level were taken as variables of interest in the model behavior. The results of sensitivity analysis for these two variables for the year 2030 are shown in Fig. 8. According to Kotir et al. (2016), 5 to 14% change represents low sensitivity, 15 to 34% is moderate, and 35% or more is high sensitivity. In Fig. 8 (a), the Groundwater level is moderately sensitive to the variable Irrigated extent per well and yield per well, which are both related to groundwater usage. The other parameters are within low sensitivity range. As shown in Fig. 8 (b), none of the parameters have affected the Farm profit or Groundwater level to more than ± 7%, so they are in the low category.
4.4 Evaluation of the Work
While this framework has been demonstrated to be useful to understand the problems and plan and implement sustainable solutions, the difficulties associated with semi-literate stakeholder understanding and usage of the SFD is an issue that needs more attention. A post-project survey helped to summarize the feedback on the framework. A total of 13 questions, on a scale of 1–10, covering the intent, methodology, facilitation, content, effectiveness, involvement and ownership were given. Figure 9 shows the average of the responses received. All questions have almost 80% acceptance and some have even higher value. Overall, the framework has been successful in systematic analysis and action plan for a complex problem using participatory methods.