**1. Introduction**

Groundwater depletion represents a major sustainability challenge for humanity in the 21st Century, given that groundwater is overexploited worldwide, particularly for agriculture [1]. About 38% of global consumptive irrigation water demand is met by groundwater [2]. With being a decade away from Sustainable Development Goals, the need for concrete policies for groundwater management has become urgent, and all water authorities call for methods to assess the effectiveness of policies. Furthermore, climate change is expected to intensify this threat, especially in regions where irrigation sustains agriculture and where population increases. Both influence the need for adaptation strategies, and for methodological approaches to identify and evaluate these same strategies. This is the case in India, where irrigation by groundwater is fully deployed [3].

The awareness that water resource management must account for interactions and feedback between biophysical processes determining movement of water and human behavior in a given socio-economic context has gained significant recognition among scientists in the past few years [4]. Different approaches exist to work on such interactions. Scenario modelling and evaluation is a standard way to explore adaptation strategies in the water resource domain in the context of climate change [5]. These methods are rooted in the more classical Story and Simulation approach [6]. They are often associated with participatory approaches and, in recent decades, several participatory stakeholder frameworks have been designed and implemented in projects that address adaptation strategies for groundwater management [7].

**Citation:** Bergez, J.-E.; Baccar, M.; Sekhar, M.; Ruiz, L. NIRAVARI: A Parsimonious Bio-Decisional Model for Assessing the Sustainability and Vulnerability of Rainfed or Groundwater-Irrigated Farming Systems in Indian Agriculture. *Water* **2022**, *14*, 3211. https://doi.org/ 10.3390/w14203211

Academic Editor: William Frederick Ritter

Received: 31 August 2022 Accepted: 8 October 2022 Published: 12 October 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Simulating decisions on the farm and the technical operations of the system is a good way to test management practices, as it integrates the dynamics of resource sharing and constraints of the farm depending on the weather and the crop's phenological stage [8]. Linking soil-crop models and decision-making models is an appropriate approach for analyzing the adaptation of practices to face new agricultural challenges, such as climate change [9]. However, in order to use modelling and simulation as a tool to provide new insights on the impacts of different water policies on farmers and on groundwater resources, there is a need for simple models which can be utilized to assess scenarios with policymakers, extensionists and farmers [10]. However, given the large diversity of farming systems in the world, there is no "one fits all" model which could adequately represent all types of agriculture. In particular, few models can represent smallholding farms, which account for a large proportion of agriculture in Africa and South Asia, and especially in India [11]. Such a model needs to be parsimonious [12], integrate farmers' decision-making on crops and crop management [13], allow integrating rainfed crops and represent the effect of conjunctive use of surface water and groundwater on the resource and different irrigation techniques [14].

One can find different types of numerical models to manage irrigation for Indian agriculture. Some are based on testing the adequacy of different crop models for the specific conditions in India (see for example Aquacrop and DSSAT-CERES in [15] or [16]). Others are more concerned with more specific biophysical processes, such as salinity [17]. Others are based on linear programming and optimization processes [18], or on artificial intelligence with neural networks and remote sensing data [19] or even the Internet of Things and machine learning [20]. Actually, very few take into account the farmer's decision-making process as proposed by [21,22]. Ref. [23] represents an interesting attempt, but is mainly focused on channel irrigation that is not the only water resource for irrigation.

Our paper presents NIRAVARI, a simple and parsimonious biodecisional model, able to test policy schemes regarding pond, borewells and irrigation equipment. The paper mainly focuses on the farmer's decision-making part of the model, and the complexity of dealing with different water resources to irrigate their farms. The initial ideas of the model were initiated with some policy-makers during a workshop held in Bangalore in March 2019. It was then quite clear that a simple tool was necessary to test a large range of scenarios. Section 1 presents the Indian context, regarding farming system and irrigation management. Section 2 presents the model; i.e., the formalism and the equations. At the end of this section, we present the different scenarios simulated to demonstrate the ability of the model. Section 3 presents the model developed and we elaborate on the outputs of the simulated scenarios. In a final section, we discuss the potential and limits of the model to elaborate and test irrigation strategies at the farm level with stakeholders.

#### **2. Materials and Methods**

#### *2.1. Specificities of Groundwater Irrigated Farms in India*

The dramatic development of groundwater irrigation in the recent decades in India has increased the irrigated area, simultaneously increasing food production to sustain the demand of a growing population. However, this came at the cost of tremendous impacts on energy consumption [24], resource depletion [25] and pollution [26,27], which now calls for an urgent rationalization of groundwater use. However, solutions are not straightforward, and recent detailed studies conducted in a groundwater-irrigated region in southern India, belonging to the Kabini Critical Zone Observatory in Karnataka ([28]; SNO M-tropics, https://mtropics.obs-mip.fr/, accessed on 11 October 2022) have illustrated the complex interactions and feedback between water resource availability and cropping systems, documenting the complex technical functioning of a diversity of farming systems [29,30]. Most Indian farmers are smallholders (less than 1ha on average), and their land is composed of one or several independent pieces of land, called "jeminu", in Karnataka. The size of a jeminu can vary greatly (from 0.1 to several hectares), but typically are about half a hectare, which can be divided into several plots for growing different crops in rotation. When

jeminus are irrigated, they share the same water sources: a borewell and/or, more rarely, a farm pond. In hard rock aquifers, pump yields are small and often insufficient to fully satisfy the needs of water-intensive crops on large surfaces [25], and they vary in time with water table level as hydraulic conductivity decreases sharply with depth [31,32]. In addition, the energy required to operate submersible pumps is provided by the government at no cost but for a limited number of hours per day. Therefore, irrigated cropping systems are still largely dependent on rainfall, and farmers need to elaborate complex strategies involving the combination of different crops in the same jeminu, and allocate the adequate irrigation water to each of them to optimize their crop yield and economic returns. The system can become even more complex, as farm ponds, encouraged by local governments for securing alternate sources of irrigation and/or for enhancing the aquifer recharge [14], are increasingly present, adding further complexity to the system. Model-based decision support systems can therefore be of great help to farmers and decision makers in assessing a range of irrigation strategies.

### *2.2. Overall Description of the System*

The modeled system (Figure 1) is composed of a jeminu (i.e., a farm) divided in *n* beles of the same size (i.e., plots). A jeminu can be rainfed or have a pond (water reserve) and/or a borewell equipped with a submersible pump. The pond can either be used to irrigate the bele or to recharge the groundwater table. The pond is represented as an inversed truncated pyramid. The surface of the pond can be removed from the productive area of the jeminu or not. This is an important feature for very small farms. The pump is used to irrigate the beles, but may also be used to refill the pond. The pump withdraws water from the aquifer and its flow rate depends on the pump characteristics and on the groundwater table depth. On each bele, a different crop can be grown. The crop growth depends on the duration since sowing. Crop management is described in a descriptive manner, i.e., date and action (sowing or harvest). Only irrigation practices follow a more complex decision-making process formalism. Different types of irrigation techniques can be used (e.g., drip, sprinkler or furrow) which differ by the amount of water they provide when irrigation is triggered. The model is run at a daily time step. Exogeneous variables are the climatic values: rainfall (*R*(*t*), mm) and potential evapotranspiration (*E*0(*t*), mm).

**Figure 1.** Schematic representation of a jeminu.

In the following, we describe the decision-making aspect of the model. The main biophysical equations that are more "classical" are given in Appendix A. Crop and water budget processes are based on FAO56 [33]. This decision-making model is based on farmers' interviews carried out in 2015 [29]. This survey was aimed at understanding farmers' farms structure and farmers' assets, and specific attention was given to the irrigation system: whether they are irrigating; whether they are using a pond and/or borewells; irrigation costs; how they pay for the irrigation systems and other equipment. A total of 684 farmers were interviewed with the help of local translators from September 2014 to March 2015, which represents 12.5% of farms in the watershed. The survey consisted of a face-to-face interview lasting 2–3 h. The survey was divided into three parts. The first part focused on household characteristics, farm structure, assets, partnerships and farm objectives. In the second part, we asked farmers about their performances and practices over the past two years. In the last part, in-depth questions were asked about irrigation, borewells and rainfall. Since no yearly records were kept by farmers, information about historical management went no further than the past two years. To develop the decision-making model, we used the approach developed by [34].
