Starting from the 1960s, there has been growth in both food production and the global population [1
]. As the global population will continue to grow in the coming decades, at the same time, food demand will increase while food producers are expected to experience greater competition for land, water, and energy [1
As such, agriculture and farming are, in general, responsible for increasing environmental threats, including degradation of land and freshwater [2
]. The technological innovations allowed a rapid increase in agricultural productivity [3
] during the last fifty years. In fact, the world’s agricultural production grew about three times over this period, while the cultivated land grew 12%. More than 40% of the growth in food production comes from irrigated land, which has doubled its area and can be interpreted as a global signal of the increasing degree of pressure on water resources [3
Major water resources exploited for irrigation are surface and groundwater bodies. For many production areas, groundwater remains the unique source of freshwater when surface water sources are not available [4
]. As a whole, irrigation is currently responsible for groundwater withdrawals of about 2800 km3
per year [5
]. In fact, irrigation represents the most impacting water use on groundwater resources [6
], as it accounts for 70% of global withdrawals and 90% of consumptive water uses [7
]. The irrigation water demand depends primarily on the extension of irrigated land, which ultimately depends on farmers’ decisions. As found in some studies [8
] farmers’ behavior with respect to cropping pattern is driven by economic factors, such as market prices, agricultural subsidies, land, and capital availability.
Focusing on the European Union, the Common Agricultural Policies (CAP) were traditionally introduced as a balancing tool to help national production to compete in both domestic and international markets [9
]. Moreover, agricultural policies enable farming profitability [10
], which produce direct and indirect values in terms of landscape conservation and cultural heritage. On the other hand, the role of agricultural policies is secondarily connected to water resource sustainability and protection [11
]. Specific policies for water resource protection often failed due to their direct and indirect contrast with farmers’ support policies [12
]. Although other policies strictly focused on the diffusion of water-saving technologies, it has not proved to be efficient in controlling irrigation water demand by adjusting the cropping towards more sustainable water resource patterns [14
Broadly speaking, a cropping plan, or pattern, refers to the acreages occupied by all the different crops every year [15
] and can be reported at the farm level where most of decisions are made [16
], or at a higher level in order to address collective issues [17
]. A number of studies have tried to explain the cropping pattern evolutions as a function of market and policy drivers. Econometric models for crop production can also be developed to understand past dynamics of crop productions, evaluate policy effects, and design new policies to enhance economic productivity and environmental conservation [18
]. Two types of models are available to describe cropping plans: cropping plan selection models which are used to support stakeholder decisions (farmers, policy-makers, and other stakeholder), and cropping plan decision models, which explicitly refer to decision-making behavior (e.g., in the field of agricultural economics) and are used to imitate stakeholder decisions to assess large-scale changes [18
]. On a large scale, the collective decisions of farmers are generated by all individual decisions mediating the impact of policy and market changes on land-use [19
] and have been treated in agricultural economics for policy analysis and forecasting [20
]. The model developed in this study belongs to the category of cropping plan decision models applied at the large scale. While the agricultural economy literature is mostly focused on ex-ante policy assessment, the developed model allows the ex-post evaluation of policy measures together with the other economic and physical (i.e., water availability) drivers influencing large scale cropping patterns during a significant time horizon.
The present study concerns the Province of Foggia (Puglia region, Southern Italy), which represents a highly-developed agricultural area and is the largest irrigated area of Puglia (Southern Italy). The irrigation service is provided and managed by the Reclamation and Irrigation Board of Capitanata (CBC), which covers 84% of the utilized agricultural area (UAA) of the province. The CBC adopts volumetric block tariffs, whereby farmers pay according to their actual consumption. However, the surface water resources of the CBC is integrated with on-farm groundwater resources from private pumping wells. In fact, the water derived from surface water (SW) bodies is not enough to fulfil irrigation water requirements, so groundwater (GW) resources are widely exploited to fulfil crop water demands in the study area. As a result, the GW level has been dramatically decreasing during the last decades [21
] and is likely to be further depleted due to climate change [22
]. It is, therefore, crucial to investigate the possible evolution of irrigation requirements which, in turn, depend on cropping pattern changes, to reach a sustainable management of water resources.
Multi-regression models were developed to interpret the inter-annual variability of cropland devoted to processing tomato (an intensive crop with a high irrigation water requirement) and durum wheat (an extensive, rain-fed crop) under the variability of the main drivers related to CAP support, market prices, crop yield, and water availability. Our working hypothesis is that water availability, together with crop economic profitability, may have shaped the evolution of cropping patterns and water resource exploitation. The purpose of the present study is to shed light on drivers of cropping patterns and their impacts on irrigation water requirement.
The article is organized as follows: After this introduction, Section 2
presents the study case; Section 3
reviews some major variables and hypotheses of modelling and, in addition, presents the adopted modelling approach; The parameterization process for two multi-regression models, and the results and discussion are presented, respectively, in Section 4
and Section 5
; The last section draws concluding remarks.
Different changes in the total agricultural area have been observed in the study area, which regarded both rain-fed and irrigated crops. For the latter ones, the observed changes are mainly due to the reduction of tomato area and of vineyard as a function of different influencing drivers. For instance, the processing tomato production has shown a fluctuating trend reflecting European and Italian agricultural policy, as well as climate and economic factors [49
], whereas the reduction of vineyard area was mainly due to the CAP extirpation payments (CE No. 479/2008).
As a whole, the accuracy of model results can be considered good. The variability of cropping area devoted to processing tomato was interpreted by means of the MRM-T, which considers four forcing drivers, which are market price, crop yield, SI, and subsidies. Specific calibrations were performed for coupled and decoupled support schemes. Considering the heterogeneity of the considered drivers, the value of each parameter is representative of both the drivers’ influence and the necessary normalization of the model equation. According to the results, under the coupled support, the crop yield had no influence on the area variability. When water availability for irrigation was regular (i.e., no restrictions in water block tariff), the most influencing driver was the intensity of subsidies (54%), followed by the market price factor (46%) while, with intermediate water availability (i.e., moderate water restrictions), subsidies were the most influencing driver (57%), followed by market prices (40%), and then the SI (3%). Under severe drought conditions, the relative importance of drivers was markedly changed, with subsidies weighing as much as 41%, followed by market price at 33%, and the SI at 26%. On the basis of such results for tomato crops, subsidies were the most influencing driver under the coupled support scheme (average value about 51%), followed by market price variability (average value 40%), while the influence of the SI, as expected, is evident only in cases of drought. In light of these results, the observed reduction of processing tomato area was mainly related to the change in CAP support scheme.
Under the decoupled support scheme, despite the irrigation seasons being regular in the observation period, the SA was performed for the three levels of water availability (Table 2
). The effect of the subsidy amount on the area variability disappeared (
= 0) as subsidies became constant (decoupled from production amount) and were actually paid only to historical areas (i.e., support was not extended outside those zones). Concerning the remaining drivers of crop area variability, under regular water availability, the most influencing turned out to be the crop yield (92%), followed by market price (8%). Under moderate water restriction, once more, the crop yield was the most influencing driver (90%), followed by market price (7%), and then SI
(3%). Then, for severe drought seasons, crop yield was the most influencing driver (59%), followed by SI
(36%) and market price (4%). In conclusion, under the decoupled support scheme, the most influential driver has become crop yield (average value about 80%), while water accessibility becomes important only during drought periods. These results highlighted the effect of water stress on the reduction of crop area. Moreover, simulated reduction of cropping area between periods with coupled and decoupled payment is in line with the findings of other studies [49
In the case of tomato, direct and decoupled payment schemes presumably have incentivised farmers to orient farming decisions to markets. This enhances competitiveness, but in the context of increasing climate variability, it also exposes farmers to yield fluctuation. Although a number of risk management instruments are available to complement farmers’ coping with large income losses, no evidence regarding the CAP 2014–2020 effects is available yet [51
]. Policy-makers should pay great attention to yield fluctuation, including more specific risk management tools within the CAP.
Additionally, during drought periods the reduction of surface water accessibility is likely to produce further negative impacts on groundwater resources. More specifically, while droughts may limit farms supplied only by surface water, farms supplied both by surface and groundwater may take advantage.
The variability of the durum wheat crop area was interpreted by means of the MRM-W model, which considers three forcing drivers, i.e., market price, crop yield and subsidy intensity. In this case, an overall calibration was performed with respect to the study period and the model structure resulted in being almost linear since only the subsidy’s exponent was different from unity (Table 1
). The SA was performed with regard to three levels of crop yield (Table 3
), supposedly directly linked to the climatic conditions. As a whole, subsidies showed to be always the most influencing driver (with an average value of about 92%). These findings are in line with those found in [52
]. In fact, there was a decrease of cropping area simultaneously to the change of the support scheme from “coupled” to “decoupled” and to the implementation of eligible areas. From the environmental standpoint, the observed reduction could increase the exploitation of water resources, due to the increasing interest of farmers towards irrigated crops, being more profitable than durum wheat production.
Generally, it was assumed that after decoupling, the CAP´s influence on farmers’ decision-making processes would be very limited. Results in this research have confirmed these assumptions only in the case of irrigated crops, such as processing tomato, in line with Giannoccaro and Berbel’s [11
] results of the CAP’s slight influence on water use after the decoupling scheme.
Bio-physical and socio-economic drivers were deeply analysed with regard to a wealthy agricultural area where both water-intensive tomato crops and rain-fed cereal crops underwent a substantial areal change.
According to the conceptual maps two distinct multi-regression models were developed to investigate the inter-annual variability of crop land devoted to tomato (an intensive crop with a high water requirement) and durum wheat (an extensive and rain-fed crop). The adopted models allowed the ex-post interpretation of the observed variability of crop area over the study period, also highlighting the different weights of each driver under the changing subsidies’ schemes and accessibility to irrigation water. Concerning the CAP reforms, the decoupled scheme explained the reduction of crop area for both tomato and durum wheat crops. In fact, the role of agricultural subsidies was highlighted for both crops as the main drivers for farming. In detail, the durum wheat area remains strongly influenced by subsidies, as the extension of the cropped area tends to the eligible area. Therefore, a reduction of support could further reduce the rain-fed crop area and increase the interest of farmers toward more profitable irrigated crops. Conversely, under the decoupled support scheme, the tomato crop appeared highly influenced by crop yield, causing an increase of risk exposure for farmers, especially under drought conditions or, more generally, when water supply restrictions are introduced. Consequently, to prevent further depletion of groundwater resources and stabilize farmers’ incomes, under the increase of yield-related risks, more specific risk management tools may be included in future CAP reforms.
In conclusion, the multi-regression modelling approach can help understand the effects of agricultural and water policies on the crop pattern change, thus, on water resources exploitation, by separating the effects of other variability sources. Provided that sufficient observations are available, the adopted approach enabled to effectively weigh the roles of human and physical drivers influencing large-scale cropping patterns and related irrigation needs. The model validation is, nevertheless, subject to certain limitations, such as the availability of crop- and site-specific multi-annual datasets, and the influence of crop production stocks from previous years on farmers’ decisions.