Abstract
Farmers in dryland regions are highly vulnerable to rainfall variability. This vulnerability is unequal, as it is mediated by biophysical and social factors. Implementing policies for climate resilience requires identification of farmers who are most vulnerable to extreme events like dry spells. We develop a novel approach by conceptualizing dry spell vulnerability at the farm scale in terms of monsoon crop water deficit. Using inputs of weather, terrain, soil properties, land-use-land-cover, crop properties, and cadastral maps, our tool models an hourly soil water balance at 30 m × 30 m resolution and maps the crop water deficit under rainfed conditions. This is a good indicator of the relative sensitivity of farmers to dry spells and allows prioritization of interventions within the focus region. Our tool, developed and deployed within the Maharashtra State Project on Climate-Resilient Agriculture, is iteratively calibrated and refined. We present the result of one such iteration where 72% of cases were found to have an agreement between the modelled output and farmers' perception of dry spell-induced crop water stress. Our work demonstrates how vulnerability to climate hazards may be mapped at micro-scales to assist policy makers in targeting interventions in ecologically fragile regions with high rainfall variability.
HIGHLIGHTS
Vulnerability of rainfed farmers to dry spells is unequal.
Mapping farm-level vulnerability is important to target interventions for climate-resilient agriculture in ecologically fragile regions.
We conceptualize farm-level dry spell vulnerability as a function of biophysical and social attributes and extend the concept to practice in the form of a tool that is embedded in the state process.
INTRODUCTION
The Indian summer monsoon provides close to 85% of the annual rainfall to the Indian subcontinent. It crucially supports food production and livelihood security for a large population. Climate change impact is causing a shift in the rainfall pattern (Singh et al., 2014) seen in the form of delayed onset of monsoons, long dry spells, and destructive floods, resulting in large crop failures and economic losses. The agricultural sector, dominated by marginal and smallholders, is highly vulnerable to monsoon variability. The monsoon (or kharif) crop is the most important crop, especially in dryland regions where most smallholders rely heavily on it for their livelihoods and household nutritional security.
Farmer vulnerability to monsoon variability is unequal (Kuchimanchi et al., 2019; Dagdeviren et al., 2021), and is mediated by access to natural, social, and economic resources (Duncan et al., 2017; Singh, 2020). Farmers who have farms with poor soil quality or unfavourable locations within the watershed, and have limited or no access to irrigation are highly vulnerable to dry spell-induced crop losses (Kumar et al., 2018). Depletion of soil moisture during dry spells, especially during key crop growth stages have been shown to have a significant impact on yield (Panigrahi et al., 2005). A systems study of dryland farmers in Northeastern Tanzania has shown how climate-related poverty traps are created as a result of crop losses from frequent dry spells and depletion of soil moisture (Enfors, 2013). Securing irrigation and retaining soil moisture is emphasized as an important strategy to reduce yield gaps caused by dry spells in Africa (Rockström & Falkenmark, 2000; Rockström et al., 2010) and India (Manivasagam & Nagarajan, 2017; Sikka et al., 2018). This has also been extended to watershed-scale supplementary irrigation planning (Imbulana & Manoharan, 2020; Hessari & Oweis, 2021). The ability to identify and map relative vulnerability is therefore desirable in order to target investments for life-saving irrigation access to farmers who are most vulnerable to dry spell-induced crop losses. This is true not only for India but also for dryland rainfed regions all over the world which face increasingly variable rainfall seasons. However, there is a gap in methodology on how dry spell vulnerability may be conceptualized and computed at the unit of a farm and, at the same time, mapped for all farms at an aggregate scale such as a village to support policy implementation.
Vulnerability has been defined and used in many different contexts and disciplines. In the context of climate change, the third assessment report of the Intergovernmental Panel on Climate Change (IPCC) provides a commonly accepted definition that frames the vulnerability of a system as a function of its exposure to a climate hazard, its sensitivity to the hazard, and its adaptive capacity. Brooks (2003) refers to this as the biophysical vulnerability, which relates to the combined effect of the physical impact of the hazard (e.g., loss in soil moisture or crop stress) and its eventual outcome (e.g., economic loss). One of the determinants of biophysical vulnerability is the social vulnerability of the system (Brooks, 2003). This is understood as a system's structural property determined by deep-rooted factors such as poverty, inequality, power relations, and access to resources (Ramprasad, 2018; Nyantakyi-frimpong, 2020), which mediate the impact of climate events resulting in outcomes such as economic damage or loss of life.
Studies on climate vulnerability assessments use varied approaches and differ in their objective, scale, and how actionable they are. The purpose of the assessments may be to identify current and potential hotspots to identify entry points for climate adaptation intervention (Kienberger et al., 2016) or to track changes in vulnerability through monitoring of adaptation efforts. The scale of mapping varies from national to village scale, and the unit of mapping varies accordingly. Studies vary in their focus on different components of vulnerability, i.e., exposure to climate hazard, sensitivity, or adaptive capacity. For example, Kadiyala et al. (2021) and Kuchimanchi et al. (2021) have mapped block-level exposure to variability in climatic elements in the Indian state of Telangana, but do not focus on differences in sensitivity or adaptive capacity. de Sherbinin et al. (2019) provide a comprehensive review of different studies for mapping social vulnerability. Social mapping is useful to understand who is structurally the most vulnerable and how this variation in vulnerability results in unequal outcomes of a climate hazard (Kuchimanchi et al., 2019; Swami & Parthasarathy, 2020). However, these studies do not capture the biophysical drivers of sensitivity to climate hazards. Also, de Sherbinin et al. (2019) find that while most works emphasize the importance of vulnerability mapping to adaptation planning, not many engage with policymakers to support decision-making. A variety of indicators and composite indices-based vulnerability assessments have been made by scholars to estimate and compare both social and biophysical vulnerability to droughts and other climate hazards at local or regional level (Lindoso et al., 2014; Keshavarz, 2016; Swami et al., 2018). Kienberger et al. (2016) identified and mapped a variety of geo-spatial indicators to spatially assess vulnerability of agricultural and pastoral livelihood systems to climate change in two provinces of Mauritania using a combination of methods including stakeholder engagement, modelling, composite index development, and visual mapping. Remote sensing-based approaches have also been used to capture spatial and temporal aspects for the computation of drought vulnerability indicators at national and sub-national scales (Meza et al., 2021).
In our review of current vulnerability assessments, we do not find one that is developed to identify differential vulnerability to dry spells at the intra-village scale, incorporating biophysical and social drivers. Moreover, none are developed for the purpose of empowering communities for demanding greater accountability from state interventions which may otherwise be appropriated by local political interests. Our paper addresses this gap by not only making a contribution to the scientific literature through our conceptualization of dry spell vulnerability as a measurable and comparable entity but also contributing to the practice of science, by developing a tool to map farmer vulnerability to dry spell hazards at the farm scale in the context of a state implemented project.
Our objective is to conceptualize and map farm-level vulnerability to dry spells to target interventions in regions that face high rainfall variability. This entails conceptualizing dry spell vulnerability, using an appropriate model to capture it, and developing a tool to map it. An iterative cycle of model calibration and tool enhancements is followed to improve the social acceptance of the final outcome. In this paper, we present the model and the results of one such cycle of field application to illustrate the novelty of the approach and the challenges to be overcome. Finally, we reflect on our transdisciplinary approach to inform decision-making for climate-resilient agriculture.
State response for climate-resilient agriculture
India has a long history of watershed programmes in drought-prone regions which have evolved over time with a gradual change in their goals and implementation methods. Although a watershed-based approach to land management started in the early 1960s, planning of scattered soil and water conservation efforts pre-dated this. The Drought-Prone Areas Programme (DPAP) was one of the early programmes launched in 1972–73 focusing on soil and water conservation in fragile, drought-prone regions. Following the severe drought of 1987, the concept of integrated watershed development was institutionalized in the form of the National Watershed Development Programme for Rainfed Areas in 1990–91. There was also a subsequent shift from the programmes being driven by state technical agencies to greater participation of non-governmental organizations.
The state of Maharashtra has been at the forefront of watershed development programmes. Drought-proofing works were already being promoted under the Employment Guarantee Scheme in the 1970s. Also in the 1970s, the Pani Panchayat movement in Maharashtra offered many success stories of people-centred watershed development which became role models for the participatory watershed programmes that emerged in the 1990s. Over the next couple of decades, watershed development guidelines called ‘Hariyali guidelines’ were developed and refined, eventually leading to the Integrated Watershed Management Programme (IWMP) in 2008. Jalyukta shivar abhiyaan (JSA), launched in 2014, was a major state-led watershed programme that promised to make all villages in Maharashtra ‘drought-free’ by the end of 2019. The programme focused on conventional water conservation activities for augmenting water availability, and at the same time introduced steps towards demand-side management. Simultaneously, Magel tyala shettale was launched in 2016, which offered subsidies to farmers for constructing private farm-ponds. Programmes such as the JSA typically expended between INR 5 million to INR 30 million in a village. Although there have been success stories (Garg et al., 2020), there have also been criticisms of the programmes by experts, civil society organizations and practitioners regarding the conduct and lack of scientific soundness of the programmes (DTE, 2019; Prasad et al., 2022).
The Project on Climate-Resilient Agriculture (PoCRA) was launched in the state of Maharashtra in 2017. Funded by the World Bank, PoCRA is the first state project on climate-resilient agriculture in the country, being implemented in over 5,000 villages in 15 districts, largely in Marathwada and Vidarbha regions. It aims to address farm-level vulnerability not only to inter-seasonal climate variability but also to intra-season variability. The programme especially focuses on farmer vulnerability to dry spells that occur during the rainfall season. Our work is developed in the capacity of a knowledge partner to the Project Management Unit of PoCRA. One of our mandates is to conceptualize climate vulnerability as an actionable condition that may be mapped and compared within a village and to develop a tool to guide the targeting of state investments. The tool is to support not only the programme implementation agency but also the villagers themselves by bringing greater accountability and transparency in the targeting of state interventions.
METHODS
In this section, we begin by first discussing our conceptualization of dry spell vulnerability based on the IPCC definition of climate vulnerability. We then describe the method used to compute and map farm-level sensitivity to dry spells followed by our approach for field-level application of the developed mapping tool.
Conceptualizing dry spell vulnerability
Based on the commonly accepted IPCC definition of vulnerability, we conceptualize monsoon dry spell vulnerability for a farmer as a combination of three components: exposure to the climate hazard, sensitivity to the exposure, and adaptive capacity of farmers. Specifically, we consider (a) exposure in the form of the occurrence of dry spells within the rainfall season, (b) the crop's sensitivity to dry spells in terms of the water stress experienced by it given the farm biophysical attributes, and (c) adaptive capacity in terms of access to irrigation for the farm plot under consideration. Since the purpose of this exercise is to map relative vulnerability within a village, it may be assumed that exposure to dry spells is not a differentiating factor, as all farm plots at a village scale are likely to experience the same dry spell events. The key differentiating factor is the sensitivity of the plots to a dry spell event and the farmers' adaptive capacities. We conceptualize sensitivity to dry spells in terms of the water deficit faced by the crop on a given farm plot during a dry spell event. This deficit is a function of the available soil moisture relative to the crop water need and depends upon the farm's biophysical properties and farm practices for moisture retention. The sensitivity varies within a village because of the variability in soil moisture retention across plots, which itself is due to the diversity in soil type within a village (Prasad & Sohoni, 2020). For example, farms with thin sandy soils experience greater crop stress than those with deep clayey soils. Hence, the need for life-saving protective irrigation (Jurriens et al., 1996) during dry spells (both quantity and frequency) varies depending upon the biophysical attributes of the farm plot for the same choice of crop. To map this sensitivity, we compute the monsoon crop water deficit (in mm of water) for the predominant kharif crop at the unit of a farm plot. Computation of this is an important requirement from our model and our key contribution in incorporating the biophysical drivers of vulnerability. Finally, the adaptive capacity of the farmer is seen as a combination of the capacity to irrigate and socio-economic factors such as access to capital, knowledge, social networks, etc. which determine access to irrigation (Kumar & Saleth, 2018). We use two proxies to capture the adaptive capacity: (a) cropping intensity of the farm as indicated by the land-use map, i.e., if only one rainfed crop is cultivated on the farm or if the farmer can farm in multiple seasons indicating that they have access to irrigation, and (b) socio-economic factors such as landholding size, gender, and caste. Together, these are used to capture farmer vulnerability to monsoon dry spell hazards.
Estimating dry spell sensitivity
The agricultural calendar has three main seasons – kharif (the monsoon crop, from June to September), rabi (the post-monsoon winter crop, from October to February), and summer (March–June). For most farmers, kharif is the main cropping season for food and income security. Typical kharif crops in the regions targeted under PoCRA are soybean, lentils such as mung, urad, or long kharif crops (6–8 month duration) such as tur (pigeon pea) and cotton. Rabi crops require irrigation and are usually cultivated only on a part of the cultivated area, and by those who have access to irrigation. Unlike other parts of Maharashtra, sugarcane cultivation in the project focus villages is limited due to no access to canal irrigation and unreliability of groundwater availability.
The primary source of water for kharif crops is the soil moisture replenished by monsoon rainfall. Variability in rainfall in the form of long dry spells causes soil moisture stress during key crop growth stages and results in crop water deficits. For the same level of exposure, the sensitivity of farm plots to a dry spell (in terms of crop stress) may be unequal. In this section, we first discuss our method to compute dry spell sensitivity at the unit of a farm, and then discuss how this information is aggregated to produce village-level maps.
Computing sensitivity to dry spell at the farm plot unit
Method developed for mapping dry spell vulnerability. An hourly soil water balance is modelled for each point at 30 m × 30 m resolution within the aggregate village boundary and the resultant crop water deficit over the rainfed season is mapped for each point. This creates the dry spell sensitivity map, which when combined with context-specific indicators of farmer adaptive capacity helps to identify farmers who are most vulnerable to dry spells.
Method developed for mapping dry spell vulnerability. An hourly soil water balance is modelled for each point at 30 m × 30 m resolution within the aggregate village boundary and the resultant crop water deficit over the rainfed season is mapped for each point. This creates the dry spell sensitivity map, which when combined with context-specific indicators of farmer adaptive capacity helps to identify farmers who are most vulnerable to dry spells.
Crop water deficit for soybean crop during a 33-day monsoon dry spell modelled for two different farm plots in Sangamner, Maharashtra, for the year 2016. (a) Clayey loam soil of 1m thickness and (b) gravelly sandy loam soil of 40cm thickness. The shaded area shows the kharif crop water deficit which is computed as the difference between crop evapotranspiration and actual evapotranspiration. For the purpose of illustration, the sowing date is assumed to be the same in both cases.
Crop water deficit for soybean crop during a 33-day monsoon dry spell modelled for two different farm plots in Sangamner, Maharashtra, for the year 2016. (a) Clayey loam soil of 1m thickness and (b) gravelly sandy loam soil of 40cm thickness. The shaded area shows the kharif crop water deficit which is computed as the difference between crop evapotranspiration and actual evapotranspiration. For the purpose of illustration, the sowing date is assumed to be the same in both cases.
Collating and aggregating sensitivity over the village map
The above section discussed the procedure developed to compute monsoon crop water deficit at a single point. To compare this across all farm plots in a village, we incorporate the model into a tool that maps this value for all points (i.e., ‘pixels’ at the resolution of 30 m × 30 m) in the cropped area within the village boundary. This is done using a Python plugin for QGIS, an open-source GIS platform. The plugin accesses GIS input layers, viz., the village boundary, land-use map, digital elevation map, soil depth, and texture map to produce a grid of crop water deficit values (see Figure 1). The crop water deficit values are classified into different categories (e.g., 0–25 mm, 25–50 mm, 50–75 mm, etc.) for visual comparison. The village cadastral map is overlaid on this to obtain a farm-plot level mapping of sensitivity to dry spells. This indicates which farm plots have a greater protective irrigation need than others to be resilient to monsoon dry spells. In the results section, we illustrate this for two selected villages. When this dry spell sensitivity map is seen in combination with farmers’ adaptive capacity, it allows us to capture relative farmer vulnerability to dry spells. We illustrate an application of this in the Supplementary file.
Field application: study area
The outputs of the model, and hence the vulnerability maps, depend on sound data inputs and calibration, especially soil and crop parameters. Thus, an important task is to combine the calibration of the model with the verification of the key input data sets through field application in sampled villages. We made a comparison of the simulated impact of dry spells in preceding years obtained by the model against the farmers' experiences. This was recorded through farmer surveys in more than 20 villages as part of the PoCRA project in iterative cycles. Here, we present the results of one such exercise conducted in two villages.
For each village, we developed a model-computed dry spell sensitivity map. We then conducted farm surveys by sampling farm plots to cover a wide geography and hence different soil attributes of the village. Farmers were asked for details such as: farm biophysical properties (soil quality, depth), kharif crops sown, dry spell details (if experienced, how long and when), if the crop required irrigation, and if they were able to irrigate. We also asked the farmers their perception about the potential yield loss if they were unable to irrigate. In the next section, we elaborate on the findings.
RESULTS
In this section, we present the results of applying the dry spell sensitivity mapping tool to the two selected villages for the monsoon of 2021. We then compare the model output against farmers' own perception of the water deficit faced by their crops during the dry spells.
Dry spell analysis
Daily monsoon rainfall pattern and dry spells for the year 2021: (a) Mangrul village, Nanded district and (b) Adgaon village, Yavatmal district (rainfall data source: Skymet).
Daily monsoon rainfall pattern and dry spells for the year 2021: (a) Mangrul village, Nanded district and (b) Adgaon village, Yavatmal district (rainfall data source: Skymet).
Dry spell sensitivity map
Mangrul village maps: (a) satellite map of village from Google Earth, drainage, (b) elevation, (c) land-use and land-cover (LULC), (d) cadastral map, (e) soil depth, (f) soil texture and output sensitivity map, and (g) cadastral level crop water deficit (in mm) for soybean crop in monsoon of 2021. Maps (a)–(f) are input data, which result in the output map (g). LULC, soil depth and soil texture maps based on shape files sourced from Maharashtra Remote Sensing Application Center (MRSAC), cadastral map sourced from Maha-bhulekh.
Mangrul village maps: (a) satellite map of village from Google Earth, drainage, (b) elevation, (c) land-use and land-cover (LULC), (d) cadastral map, (e) soil depth, (f) soil texture and output sensitivity map, and (g) cadastral level crop water deficit (in mm) for soybean crop in monsoon of 2021. Maps (a)–(f) are input data, which result in the output map (g). LULC, soil depth and soil texture maps based on shape files sourced from Maharashtra Remote Sensing Application Center (MRSAC), cadastral map sourced from Maha-bhulekh.
Adgaon village maps: (a) satellite map of village from Google Earth, drainage, (b) elevation, (c) land-use and land-cover (LULC), (d) cadastral map, (e) soil depth, (f) soil texture and output sensitivity map, and (g) cadastral level crop water deficit (in mm) for soybean crop in monsoon of 2021. Maps (a)–(f) are input data, which result in the output map (g). LULC, soil depth and soil texture maps based on shape files sourced from Maharashtra Remote Sensing Application Center (MRSAC), cadastral map sourced from Maha-bhulekh.
Adgaon village maps: (a) satellite map of village from Google Earth, drainage, (b) elevation, (c) land-use and land-cover (LULC), (d) cadastral map, (e) soil depth, (f) soil texture and output sensitivity map, and (g) cadastral level crop water deficit (in mm) for soybean crop in monsoon of 2021. Maps (a)–(f) are input data, which result in the output map (g). LULC, soil depth and soil texture maps based on shape files sourced from Maharashtra Remote Sensing Application Center (MRSAC), cadastral map sourced from Maha-bhulekh.
Field interview findings
Comparison of soybean crop water deficit computed for surveyed plots using two different input data for soil type (state MRSAC data vs. primary field data).
Comparison of soybean crop water deficit computed for surveyed plots using two different input data for soil type (state MRSAC data vs. primary field data).
Kharif dry spell vulnerability in 2021 monsoon for Mangrul (a) interviewed farm plots, (b and d) dry spell sensitivity map in terms of irrigation requirement for (b) soybean and (d) cotton based on farmers' soil description (assumes irrigation not needed when crop water deficit is ≤15mm, and irrigation needed when deficit is more than 15mm), (c and e) farmers' description of whether irrigation was needed in 2021 monsoon for (c) soybean and (e) cotton. (0 = no need; 1 = needed). Cadastral map sourced from the government of Maharashtra Maha-bhulekh.
Kharif dry spell vulnerability in 2021 monsoon for Mangrul (a) interviewed farm plots, (b and d) dry spell sensitivity map in terms of irrigation requirement for (b) soybean and (d) cotton based on farmers' soil description (assumes irrigation not needed when crop water deficit is ≤15mm, and irrigation needed when deficit is more than 15mm), (c and e) farmers' description of whether irrigation was needed in 2021 monsoon for (c) soybean and (e) cotton. (0 = no need; 1 = needed). Cadastral map sourced from the government of Maharashtra Maha-bhulekh.
Kharif dry spell vulnerability in 2021 monsoon for Adgaon (a) interviewed farm plots, (b and d) dry spell sensitivity map in terms of irrigation requirement for (b) soybean and (d) cotton based on farmers' soil descriptions (assumes irrigation not needed when crop water deficit is ≤15mm, and irrigation needed when deficit is more than 15mm), (c and e) farmers' descriptions of whether irrigation was needed in 2021 monsoon for (c) soybean and (e) cotton. (0 = no need; 1 = needed). Cadastral map sourced from the government of Maharashtra Maha-bhulekh.
Kharif dry spell vulnerability in 2021 monsoon for Adgaon (a) interviewed farm plots, (b and d) dry spell sensitivity map in terms of irrigation requirement for (b) soybean and (d) cotton based on farmers' soil descriptions (assumes irrigation not needed when crop water deficit is ≤15mm, and irrigation needed when deficit is more than 15mm), (c and e) farmers' descriptions of whether irrigation was needed in 2021 monsoon for (c) soybean and (e) cotton. (0 = no need; 1 = needed). Cadastral map sourced from the government of Maharashtra Maha-bhulekh.
As an example, let us consider plot number 129 in Mangrul (Figure 8(a)). There are multiple farmers within the plot and two of them were interviewed. The first farmer cultivates 1.25 acres on this plot and also owns three other plots (numbers 106, 113, and 119) with a total of 7.65 acres across the four plots. The second farmer cultivates 3.7 acres on plot number 129. In the kharif season of 2021, the first farmer cultivated soybean while the second farmer cultivated both soybean and cotton. For farm plot 129, the MRSAC data indicated shallow soil depth (between 10 and 25 cm deep). Due to this, the computed deficit in Figure 5(g) is high (75–100 mm for soybean). However, both farmers described the soil depth in their farms to be more than 2 feet (>60 cm). As a result, we recomputed the crop water stress with corrected soil depth for all surveyed plots. With the correction, the model output shows that soybean would face a water deficit of about 30 mm. This agrees with the farmers’ views who both indicated that their soybean crop needed to be irrigated. The first farmer was able to provide this by accessing a well, while the second was unable to do so. This was also reflected in the differences in their crop yields for soybean. In the case of the cotton crop on the same plot, the model estimates no need for irrigation due to the larger root zone for cotton from which moisture is accessible to the crop. However, in this, there was a disagreement with the farmer who believed that the crop needed irrigation (although he was unable to irrigate) while the model estimated that it was not needed.
Computation method for farm-level water balance tool.
Water balance component . | Computation method . | Reference . |
---|---|---|
Precipitation | Input from meteorological data | |
Reference Evapotranspiration | Hargreaves model | FAO Paper 56 (Allen et al., (1998)) |
Surface Run-off | SCS Curve number adjusted for slope | SWAT manual (Neitsch et al., (2011)) |
Crop evapotranspiration (ETc) | FAO methodology for crop ET under standard conditions | FAO Paper 56 (Allen et al., (1998)) |
Actual crop evapotranspiration (AET) | FAO methodology for crop ET under soil water stress | FAO Paper 56 (Allen et al., (1998)) |
GW recharge | SWAT methodology | SWAT manual (Neitsch et al., (2011)) |
Soil moisture | Mass balance | |
Monsoon crop water-deficit | Difference between potential Crop Evapotranspiration and Actual evapotranspiration aggregated over the monsoon period |
Water balance component . | Computation method . | Reference . |
---|---|---|
Precipitation | Input from meteorological data | |
Reference Evapotranspiration | Hargreaves model | FAO Paper 56 (Allen et al., (1998)) |
Surface Run-off | SCS Curve number adjusted for slope | SWAT manual (Neitsch et al., (2011)) |
Crop evapotranspiration (ETc) | FAO methodology for crop ET under standard conditions | FAO Paper 56 (Allen et al., (1998)) |
Actual crop evapotranspiration (AET) | FAO methodology for crop ET under soil water stress | FAO Paper 56 (Allen et al., (1998)) |
GW recharge | SWAT methodology | SWAT manual (Neitsch et al., (2011)) |
Soil moisture | Mass balance | |
Monsoon crop water-deficit | Difference between potential Crop Evapotranspiration and Actual evapotranspiration aggregated over the monsoon period |
In this way, a plot-wise comparison was made for each crop and farmer. This comparison is tabulated in Table 2, where the rows are the bins for the deficits obtained from the corrected soil parameters, and column A is the frequency of that bin. Columns B and C record the farmers' perceptions for comparison. We find that for 72% (i.e., 56 of the 77 farm-crop pairs) of the data points there is an agreement between the model's output and the farmers' perceptions about their need for irrigation. It was seen that there are several nuances in farmers' perceptions of when a crop needs irrigation. Some farmers do not feel that there is a need for irrigation unless there is significant crop water stress, i.e., they seek protective irrigation for the crop to avoid a heavy loss of yield. And there are others who believe that irrigation is needed at the earliest signs of stress, i.e., they aim for productive irrigation to maximize crop yields. It is therefore expected that due to the qualitative nature of the questions posed to the farmers, there would not be a perfect match between farmers' perceptions and the model output. However, the model and farmers' perceptions match for plots with high crop water deficit (75 mm and higher) where protective irrigation would typically be required to reduce significant yield loss. Since 2021 was a high rainfall year, there were few plots with high crop water deficits. Another factor that may have contributed to mismatches between the model output and farmer perspective is poor recall of the dry spell impact by the farmers as they were heavily impacted by the destructive wet spell at the harvesting stage.
Comparison of modelled monsoon crop water deficit using corrected soil type with survey output for soybean and cotton farmers, year 2021.
Crop water deficit model output (mm) . | (A) Number of farm plots . | (B) Farmer perspective: irrigation was not needed . | (C) Farmer perspective: irrigation was needed . |
---|---|---|---|
0 | 17 | 10 | 7 |
0–15 | 21 | 14 | 7 |
15–30 | 27 | 6 | 21 |
30–45 | 4 | 1 | 3 |
45–60 | 0 | 0 | 0 |
60–75 | 0 | 0 | 0 |
75–100 | 2 | 0 | 2 |
100–125 | 6 | 0 | 6 |
Total | 77 | 31 | 46 |
Crop water deficit model output (mm) . | (A) Number of farm plots . | (B) Farmer perspective: irrigation was not needed . | (C) Farmer perspective: irrigation was needed . |
---|---|---|---|
0 | 17 | 10 | 7 |
0–15 | 21 | 14 | 7 |
15–30 | 27 | 6 | 21 |
30–45 | 4 | 1 | 3 |
45–60 | 0 | 0 | 0 |
60–75 | 0 | 0 | 0 |
75–100 | 2 | 0 | 2 |
100–125 | 6 | 0 | 6 |
Total | 77 | 31 | 46 |
It is assumed that a modelled deficit of <15 mm requires no irrigation. This is compared with farmers' perception of whether irrigation was needed by the crop.
DISCUSSION
Vulnerability is a complex phenomenon, which needs better understanding and conceptualization. Currently, in practice, it is measured through social identities such as caste, class, gender, and their intersections which are shown to generally map to different levels of access to material and social assets. When it comes to climate vulnerability, these factors influence the adaptive capacity of farmers. However, crop sensitivity to dry spells is a largely biophysical phenomenon, which has so far not been captured in climate vulnerability literature or in state interventions to address farmer vulnerability. Our key contribution is our conceptualization of dry spell sensitivity in terms of crop water deficit at the unit of a farm and a methodology to map its distribution to allow comparison between farms at an aggregate scale such as a village or cluster of villages. A related quantity that has been considered a good indicator of agricultural droughts is the soil moisture content. However, soil moisture alone is not a good indicator of dry spell sensitivity as it needs to be seen in combination with other variables such as precipitation patterns and type of crop grown. For example, Gao et al. (2016) used a distributed hydrological model to capture and map soil moisture for their study area in North China Plain and extended this to map drought vulnerability. This required mapping of four distinct factors: rainfall distribution, available run-off, soil water storage capacity, and crop type, and creating a vulnerability score by weighing them equally. Our method of mapping crop water deficit is an improvement over such an approach as it integrates these related factors into a single physical quantity that captures the stress experienced by the crop during a dry spell. Secondly, crop water deficit is an actionable attribute that suggests the extent of irrigation needed to avoid crop loss. In this way, our proposed map not only presents who is most vulnerable to dry spells but also relates it to the action needed to address it. Preston et al. (2011), in their review of approaches to map climate vulnerability, call this an important criterion for placing vulnerability mapping on a robust footing. Thirdly, our approach to mapping vulnerability makes it possible to validate or refute the output, unlike vulnerability mapping approaches which are broader representations of vulnerability and are thus limited in their ability to influence policy and action (de Sherbinin et al., 2019).
With increasing variability in rainfall patterns across the world, especially in rainfed regions of Asia and Africa, identifying and targeting support to farmers facing dry spell vulnerability will be increasingly important. Our method is easily replicable in other geographies. Coupling dry spell sensitivity with appropriate regional proxies for adaptive capacity (such as access to irrigation, and knowledge) leads to a robust way to measure different levels of vulnerability to enable action. In the Supplementary material, we illustrate how our tool assists prioritization and planning of a specific type of investment, i.e., the open dug-well, within the PoCRA project by combining dry spell sensitivity maps with indicators of adaptive capacity. Supplementary material, Fig. S1 depicts the Well Beneficiary module used to create a village scale vulnerability map based on which subsidies for dug wells are prioritized. A sample of the generated output report is shared in Supplementary material, Fig. S2. There are, however, many challenges associated with the approach which we reflect on in the following section.
As seen in our results, the quality of input datasets presents many challenges. For example, significant inconsistency was found between the soil characteristics as indicated by the state-provided MRSAC maps and the soil characteristics described by the farmers. As a result, the project partnered with the National Bureau of Soil Survey and Land-Use Planning (NBSS&LUP) to improve the accuracy of soil maps. Besides this, other challenges in input data sets were encountered, which are tabulated in Table 3. Similar challenges in vulnerability mapping have also been called out by other scholars. Given the inter-sectoral nature of vulnerability assessments, Kienberger et al. (2016) in their spatial vulnerability mapping in Mauritania emphasize the importance of the availability of data at good spatial resolution and policies for data sharing across sectors.
In addition to efforts on improving input data quality, the model requires many refinements, for example, to improve the calibration and to incorporate the effect of key farm-preparation practices. Although the tool is based on a standard crop water–soil model, there is a need to calibrate it to context-specific practices, e.g., irrigation methods, intercropping practices, etc. for better results. For example, Gao et al. (2016) report an error margin of 10% in their mapping of modelled soil moisture when compared to observed values, despite the availability of detailed input data. Some of our ongoing efforts for model improvements are tabulated in Table 4. These demonstrate the challenges in the practical application of science for enhancing climate resilience in agriculture. It requires long-term engagement, extensive fieldwork, and continuous engagement with a large number of stakeholders.
Status of input data sets and needs for further improvement.
Input dataset . | Current source of data . | Scale and resolution . | Limitation . | Corrective measures needed/taken . |
---|---|---|---|---|
Soil texture and soil depth shape files | Maharashtra Remote Sensing Application Centre (MRSAC) | Scale: 1:50000 | Poor accuracy of soil texture and depth | MoU with National Bureau of Soil Survey (NBSS) initiated for generation of more accurate (higher resolution) soil maps |
Soil properties (e.g., bulk density, saturated hydraulic conductivity, field capacity etc.) for each texture class | FAO and SPAW (Soil–Plant–Air–Water) model | NA | No state-level database available for local soil properties, no studies undertaken | Extensive field surveys and lab tests required to generate local state-level soil databases capturing organic matter, salinity etc. that impact soil properties |
Crop properties such as: root depth, depletion factor, crop duration, growth stages, Kc value | FAO | NA | Kc values may differ from region to region and variety to variety. No state-level data available | Estimation of Kc for few major crops through lysimeter experiments at State Agriculture Universities to be initiated in PoCRA. More local studies required for documenting different crop varieties and their properties |
Rainfall and weather data | Skymet automated weather stations | Spatial: Roughly 1 station per 13,000 ha. Temporal – hourly | Problem of some weather stations going down during monsoon season due to power outage or network issues | Significant improvement in data resolution and frequency is already achieved – from yearly district level rainfall data (1998), then daily block level data (1999), followed by daily revenue circle level rainfall data (from 2013). Since 2018, hourly weather data is recorded by Skymet in collaboration with GoM through weather stations in all revenue circles |
Curve number | USDA | Available for different land covers and HSGs (Classified as per US conditions) | Researchers have shown differences in curve numbers when computed locally. But no state-wide data available on curve numbers. | Need several local studies for generating local database for Curve Numbers |
Digital elevation map | SRTM | 30 m × 30 m | Current resolution and quality are sufficient for generating slope and drainage maps. | 1 m × 1 m resolution DEM maps available. May be used for better delineation of stream proximity regions in the future |
LULC data | MRSAC | Scale: 1:50000 Temporal: once in 5 years | Latest available data is from 2015 to 16. Yearly changes in LULC due to rainfall are important but not currently recorded/published | LULC map must be updated at least every year but preferably quarterly to capture seasonal cropping |
Cadastral data | Maha-Bhulekh and MRSAC | Temporal frequency: Maps updated every 5 years | Changes and divisions of land-parcels not reflected yearly. Quality: several duplicates, overlaps in cadastre polygons, survey plot numbers encountered | Data needs frequent updates, resolution of errors |
Input dataset . | Current source of data . | Scale and resolution . | Limitation . | Corrective measures needed/taken . |
---|---|---|---|---|
Soil texture and soil depth shape files | Maharashtra Remote Sensing Application Centre (MRSAC) | Scale: 1:50000 | Poor accuracy of soil texture and depth | MoU with National Bureau of Soil Survey (NBSS) initiated for generation of more accurate (higher resolution) soil maps |
Soil properties (e.g., bulk density, saturated hydraulic conductivity, field capacity etc.) for each texture class | FAO and SPAW (Soil–Plant–Air–Water) model | NA | No state-level database available for local soil properties, no studies undertaken | Extensive field surveys and lab tests required to generate local state-level soil databases capturing organic matter, salinity etc. that impact soil properties |
Crop properties such as: root depth, depletion factor, crop duration, growth stages, Kc value | FAO | NA | Kc values may differ from region to region and variety to variety. No state-level data available | Estimation of Kc for few major crops through lysimeter experiments at State Agriculture Universities to be initiated in PoCRA. More local studies required for documenting different crop varieties and their properties |
Rainfall and weather data | Skymet automated weather stations | Spatial: Roughly 1 station per 13,000 ha. Temporal – hourly | Problem of some weather stations going down during monsoon season due to power outage or network issues | Significant improvement in data resolution and frequency is already achieved – from yearly district level rainfall data (1998), then daily block level data (1999), followed by daily revenue circle level rainfall data (from 2013). Since 2018, hourly weather data is recorded by Skymet in collaboration with GoM through weather stations in all revenue circles |
Curve number | USDA | Available for different land covers and HSGs (Classified as per US conditions) | Researchers have shown differences in curve numbers when computed locally. But no state-wide data available on curve numbers. | Need several local studies for generating local database for Curve Numbers |
Digital elevation map | SRTM | 30 m × 30 m | Current resolution and quality are sufficient for generating slope and drainage maps. | 1 m × 1 m resolution DEM maps available. May be used for better delineation of stream proximity regions in the future |
LULC data | MRSAC | Scale: 1:50000 Temporal: once in 5 years | Latest available data is from 2015 to 16. Yearly changes in LULC due to rainfall are important but not currently recorded/published | LULC map must be updated at least every year but preferably quarterly to capture seasonal cropping |
Cadastral data | Maha-Bhulekh and MRSAC | Temporal frequency: Maps updated every 5 years | Changes and divisions of land-parcels not reflected yearly. Quality: several duplicates, overlaps in cadastre polygons, survey plot numbers encountered | Data needs frequent updates, resolution of errors |
Ongoing and planned model improvements.
Problem addressed . | Model improvement . | Remark . | Status . |
---|---|---|---|
Model run-off higher than field observation | 1. Change from daily to hourly time-step | Complete. Report published here: https://www.cse.iitb.ac.in/∼pocra/Report_a.pdf | |
2. Shift from using static monthly ET0 values to weather-based hourly values. | Linked with hourly Skymet data | Complete | |
3. Calibrating reference evapotranspiration (ET0) computation | Comparison of ET0 values from three different empirical equations (Hargreaves, Jensen-Haise and FAO-56 Penman Monteith) against measured data from WALMI to select the appropriate one | Complete. Report published here: https://www.cse.iitb.ac.in/∼pocra/MoU%20II%20Phase%20II/PET_Method.pdf | |
Different farm practices and their impact on recharge and soil moisture retention | 1. Furrows: Incorporating ponding effect | Ponding constant introduced in the model | Complete. Report published here: https://www.cse.iitb.ac.in/∼pocra/Report_a.pdf |
2. Impact of intercropping | To be addressed in the future | ||
Improvement in groundwater recharge estimates in high rainfall situations | 1. Improvement in groundwater recharge estimation | Incorporating hydrogeological parameters such as aquifer thickness and specific yield for capping GW recharge and accounting excess as surface or base flows | Conceptual prototype recommended and accepted by GoM. The changes can be implemented over the whole state once GSDA publishes/shares better resolution hydrogeological data |
Problem addressed . | Model improvement . | Remark . | Status . |
---|---|---|---|
Model run-off higher than field observation | 1. Change from daily to hourly time-step | Complete. Report published here: https://www.cse.iitb.ac.in/∼pocra/Report_a.pdf | |
2. Shift from using static monthly ET0 values to weather-based hourly values. | Linked with hourly Skymet data | Complete | |
3. Calibrating reference evapotranspiration (ET0) computation | Comparison of ET0 values from three different empirical equations (Hargreaves, Jensen-Haise and FAO-56 Penman Monteith) against measured data from WALMI to select the appropriate one | Complete. Report published here: https://www.cse.iitb.ac.in/∼pocra/MoU%20II%20Phase%20II/PET_Method.pdf | |
Different farm practices and their impact on recharge and soil moisture retention | 1. Furrows: Incorporating ponding effect | Ponding constant introduced in the model | Complete. Report published here: https://www.cse.iitb.ac.in/∼pocra/Report_a.pdf |
2. Impact of intercropping | To be addressed in the future | ||
Improvement in groundwater recharge estimates in high rainfall situations | 1. Improvement in groundwater recharge estimation | Incorporating hydrogeological parameters such as aquifer thickness and specific yield for capping GW recharge and accounting excess as surface or base flows | Conceptual prototype recommended and accepted by GoM. The changes can be implemented over the whole state once GSDA publishes/shares better resolution hydrogeological data |
Our experience is consistent with de Sherbinin et al. (2019) who caution that developing vulnerability maps should not become an end in itself. The centrality of stakeholder engagement in vulnerability mapping cannot be sufficiently emphasized. At the same time, conducting transdisciplinary research with different stakeholders who have different expectations has its own challenges. For example, the government agencies want measurable output within a given timeframe, the scientists look for scientific robustness of the output and the communities expect an outcome that they deem to be relatable and understandable. The desired approaches of each stakeholder are also different. State agencies look for more centralized, secondary data-based processes to plan interventions. The communities have their institutions and structures of power and politics embedded within them which have strong interests in planning interventions. It is these considerations that researchers must negotiate when working together with policymakers and programme implementors.
CONCLUSION
All over the world, rainfed agriculture faces growing risk from climate change. Farmers suffer from dry spell-induced crop failures. Implementing policies to enhance climate resilience of farmers within the dryland regions requires ways of identifying and mapping relative vulnerability in order to target investments to the most vulnerable. Vulnerability mapping has been used at coarser scales for problem orientation and policy decisions on which hotspots to focus on. Here, we present a methodology that allows vulnerability mapping at the unit of a farm, which is useful to support targeting of interventions for climate resilience within a project focus area, such as a cluster of villages. This is important since there is wide variation in vulnerability within any focus region.
In India, the Government of Maharashtra's PoCRA is the first state programme in the country that has the mandate to support farmers identified as most vulnerable. Our scientific inputs and tools for mapping vulnerability have been developed and deployed in the programme through a transdisciplinary partnership with the implementing agency. Our scientific contribution is in the conceptualization of dry spell vulnerability as a combination of crop water deficit, which is a function of farm biophysical attributes, and the farmers' adaptive capacity, which is driven by social attributes. We extend this concept to practice in the form of a tool that is embedded in the state process to target interventions and bring more transparency to the process. The development of our tool has been through an iterative effort of field validation, reporting, and model improvements. The paper presents the results of one such cycle of field application. Our work demonstrates how vulnerability to climate hazards may be mapped at micro-scales to assist policy-makers and practitioners in targeting interventions. Our methods can be easily adapted to plan for climate resilience in other ecologically fragile regions which experience high variability in rainfall and other weather parameters. Our tool and the process in which it is embedded is a significant step in current intervention planning for climate-resilient agriculture, yet there continue to be several areas of improvement. Our key learning is that enhancing climate resilience in agricultural production requires infusion of new science within state processes through long-term partnerships with different agencies and communities.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the support of the PoCRA Project Management Unit (PMU): Shri Rastogi (IAS, Project Director till 2021), Shri Kolekar (Agronomist), Ms Indra Mallo (IAS, Project Director), the staff at PMU and relevant staff of the Department of Agriculture, Government of Maharashtra. We also thank the innumerable farmers who provided key inputs. We would like to acknowledge the inputs of all the students who have been associated with the project – Gopal Chavan, Swapnil Patil, current staff member Asim R. P. (who contributed to the Well Beneficiary module), and previous staff members: Dr Rahul Gokhale, Shubhada Sali, and Vidyadhar Konde.
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.
CONFLICT OF INTEREST
The authors declare there is no conflict.