Catalonia (NE Spain) is a region located on the north-western Mediterranean coast with a contrasting terrain that generates a great diversity of microclimates. The Pyrenees and the smaller mountain ranges that extend parallel to the coastline have determined distinct bioregions with a characteristic climate, natural vegetation and traditional crops. The total area of Catalonia is 31,895 km2, of which 26% is agricultural land, 64% is forest, and the rest is urban regions and other land uses. Traditionally, the coastal strip has been an area of dryland agriculture based on crops with a low water demand (hereafter ‘Littoral subregion’). Inland, there is a very dry area where irrigation is predominant (‘Interior subregion’), and then finally there is the mountain climate zone (‘Pyrenees subregion’) dedicated to pastures, forestry, and crops with higher water demands (forage and some fruit and vegetable species) (Fig. 1).

The impact of climate change on this complex landscape is already underway (Peñuelas & Boada, 2003; Pascual et al., 2016). Numerous studies predict a change to warmer conditions and an increase in water deficit for the North-western Mediterranean coast (Quiroga & Iglesias, 2009; Tanasijevic et al., 2014; Donatelli et al., 2015), and the future performances of specific crops have been modelled with a broad regional scope (Garrido et al., 2011; Ferrise et al., 2013; Saadi et al., 2015). However, local studies are needed to promote the optimal adaptation strategies for the agricultural sector of each territory in a changing environment, a process that should be championed and monitored by local administrations and local farmers organizations (Ronchail et al., 2014).

In this article, we aimed to quantify the impacts of climate change predictions on traditional rainfed species and identify the regions in which crop replacement or irrigation developments will be required. To measure these impacts, we used Papadakis’ agroclimatic classification with a high spatial resolution in Catalonia. The modifications that this agroclimatic zoning will experience according to regional projections of climate change were calculated. The impacts of these modifications on five important Mediterranean crops, which added up to 44.9% of the arable land of Catalonia in 2018, were analysed. Alfalfa (Medicago sativa; 4.5% of total arable land) and barley (Hordeum vulgare; 23.6%) served as models for rainfed herbaceous crops traditionally produced in the more humid or in dry Mediterranean areas, respectively. Almond (Prunus dulcis; 4.6%) and olive (Olea europaea; 12.0%) have been the most common rainfed woody crops for centuries in the study area. However, modern commercial agriculture requires constant high productivities, and so irrigation has been implemented where possible also for these four crops. Finally, orange (Citrus × sinensis; 0.2%) was studied as a crop with a great expansion potential in irrigated conditions. Oranges are successfully produced in the southern warm areas of Catalonia, and are a traditional local product in Valencia, which is immediately to the south. Orange is therefore likely to expand its agroclimatic viability northwards as winters expectedly become milder.

FigureFigure 1.Situation of the study area, showing the three traditional agricultural and climatic subregions. Black dots indicate the location of meteorological stations

Retrieval of public geographic and climate data

Papadakis’ classification is based on a calculation cascade starting with four monthly values (48 variables in total): monthly average of daily minimum temperature (tmin, ºC) and maximum temperature (Tmax, ºC), average monthly precipitation (P, mm/month), and average absolute minimum temperature of every month (tminabs). The first three types of values (36 variables) are publicly available from the Digital Climate Atlas of Catalonia (DCAC) as data layers in raster format with a resolution of 180 m, derived from meteorological records (Ninyerola et al., 2000). For every location of this grid, altitude (Alt, meters above the sea level), distance to the sea (DistSea, km), the latitude expressed as a cosine (CosLat), UTM coordinates, and monthly data of average solar radiation (Rad, 10 kJ·m-2·day-1) were obtained in order to calculate tminabs following the methodology of the publicly available temperature data layers. All data were extracted with QGIS v3.10.

Generation of monthly absolute minimum temperature layers from meteorological records

The Catalan Meteorological Service provided tminabs data of 151 meteorological stations spanning a minimum of 12 years (2007-18). The locations of these stations are indica-ted in Fig. 1. For these 12-year series the average tminabs of every month were calculated, and multiple linear regressions were performed to obtain average tminabs as a function of four independent variables (Alt, CosLat, DistSea, and Rad).

where z (s) is tminabs at location s, β is the intercept, fk(s) are the four independent variables and αk are the regres-sion coefficients of the independent variables.

Subsequently, average monthly tminabs values were obtained throughout the study area using the Universal Kriging (UK) interpolation method:

where λi are kriging weights and ε(si) is the observed resi-dual at measurement locations si. Further details are readily available in the literature (Benavides et al., 2007; Hengl et al., 2007; Bostan et al., 2012). For each month, the UK in-terpolation was performed using only the variables that pro-ved significant in the multiple regression (Rad was found to be not significant for some of the months, see Results). Therefore, the value p is 3 or 4 depending on the number of significant variables, while q=151 meteorological stations.

For both kriging and regression, models were vali-dated by initially using 66.6% of the data (100 stations) and contrasting the results with the remaining 33.4%. Root-mean-square errors (RMSE) were calculated by comparing predictions for the meteorological station sites with the actual observed values (20 random iterations).

where Oi=observed value based on meteorological station records; Pi=predicted value by either kriging or linear re-gression. To obtain the final layers of monthly tminabs, data from all 151 stations were used. All computations were developed with the statistical package R Studio v1.2.50001.

Table 1.Constants used to transform temperature and precipitation values among baseline and future scenarios (temperature variation in °C and precipitation change in %) as reported in the Third Report on Climate Change in Catalonia (Calbó et al., 2016).

Baseline (1971-2000) and future (2031-2050) scenarios under study

The Third Report on Climate Change in Catalonia (TRCCC) offers estimates of temperature and precipitation changes for the 2012-2020 and 2031-50 time periods (relative to the period 1971-2000, hereafter “baseline scenario”). These estimations were based on a regionalization of the RCP4.5 emissivity scenario (Calbó et al., 2016). In the TRCCC, the expected percent variation of precipitation and average temperature increase were calculated in a subregion- and season-dependent manner (Table 1). In the present study, data layers of monthly precipitations and temperatures were obtained for the period 2031-2050 (hereafter termed “future scenario”) based on these TRCCC projections:

where Tx=Tmax or tmin; Kt2031=temperature variation constant (as ºC) for the 2031-2050 scenario, for the corresponding month and subregion (Table 1); Kp2031=precipitation variation constant (as %) for the 2031-2050 scenario, for the corresponding month and subregion (Table 1).

The limits of the three subregions in which Catalonia was divided are depicted in Fig. 1.

On the other hand, public datasets of Tmax, tmin, and P are based on meteorological records that cover the baseline scenario, but no tminabs datasets are available for this period. Since the tminabs layers calculated in this study are based on data from years 2007-2018, they were considered to correspond to the 2012-2021 scenario, so the variation constants between the baseline and the 2012-2021 scenarios were subtracted to generate simulated tminabs data for the baseline scenario.

where Kt2012=temperature variation constant (as ºC) for the 2012-2021 scenario, for the corresponding month and subregion (Table 1).

FigureFigure 2.Flow diagram of the programming developed in spreadsheets to obtain the cultivation suitability of five crops. A) Manipulation of the available starting data to generate values of the four variables for three baseline and future scenarios through the correctors shown in Table 1. B) Calculation of agroclimatic categories according to Papadakis and resulting suitability of cultivation, performed for the baseline and future scenarios. ETP0=potential evapotranspiration; Ln=rainwash index; Ih=humidity index; VR=variation of soil water reservoir; xGSs=growing seasons.

Computing agroclimate according to Papadakis’ classification

The methodology developed by Papadakis (1966) classifies the world’s climates in terms of their agricultural potential. This classification reflects thermal and humidity regimes in detail because it uses elaborate definitions and extreme values rather than direct average parameters (Verheye, 2008). To characterize climates in terms of crop ecophysiology at a specific location, Papadakis’ procedure takes into account winter severity, summer thermal stress, and water availability throughout the year (Fig. 2B). It uses extreme temperatures, intra-seasonal thermal and rainfall differences and soil water balance.

Winter type

The winter type defines the severity of the cold season as a function of the following parameters:

• absolute minimum temperature of the coldest month,

• minimum temperature of the coldest month,

• maximum temperature of the coldest month.

Summer type

To define the summer type, the following parameters were considered:

• Growing seasons (xGS): They are defined as the pe-riod in which temperatures are above a threshold of 0 ºC (mean GS - MGS), 2 ºC (available GS - AGS) or 7 ºC (minimum GS - minGS) at all times. To calculate them, linear regressions are made between tminabs (y axis) and months (x axis) for the two semesters of the year.

where z (m) is tminabs for month m and the subscripts in-dicate the semester of the regression. By making z=0, iso-lating m1 and m2 and subtracting eq (8) – eq (7), MGS is obtained. Similarly, for z=2 and z=7, the subtraction will yield AGS and minGS. It should be noted that, following Papadakis’ protocol, in the semester in which monthly tminabs increases (1st semester in the Northern hemis-phere), tminabs is considered to occur at the beginning of the month, while in the semester with decreasing monthly tminabs, it is considered to occur at the end of the month.

• Average of the Tmax of the n warmest months, for n=2, 4 or 6,

• Highest Tmax,• tmin of the month with the highest Tmax,

• Average of the tmin of the two months with highest Tmax.

Humidity regime

To describe the water availability that climate provi-des, Papadakis’ method assumed a maximum soil water content of 100 mm in the most humid month of the year. This volume is the reservoir that suffers losses due to evapotranspiration and receives input from rainfall, and a monthly water balance is performed. In the study area, this month was set as December, after the autumn rains and when day length and mean temperatures drop.

where Rm is the mean soil water content for month m.

The following parameters must be taken into account in order to determine the humidity regime:

• Annual humidity index (Ih): ratio of annual precipita-tion over evapotranspiration

• Level of humidity of every month:

where ΔRm=Rm – Rm-1 is the monthly variation of soil wa-ter content.

• Rainwash index (Ln): Sum of the balance (Pm-ETPm) over the 12 months, for all months in which Pm>ETPm

The criteria to define each category are provided as supplementary material (Table S1 [suppl.]). Papadakis’ work included the description of the climate requirements of common crops in terms of his agroclimatic classifica-tion (Table S2 [suppl.]). All maps were plotted by means of the qGIS software.

Suitability of cultivation

To quantify and visualize the impact of climate change on agriculture, five typical Mediterranean crops (almonds, olives, barley, alfalfa and oranges) were selected. Winter and summer types and humidity regimes under which the most common crops can thrive are easily available, for example from the Spanish Ministry of Agriculture (De León Llamazares, 1989; Table S2 [suppl.]). For each spe-cies, the coldest winter types and summer types that are entirely tolerated are indicated, along with excess heat constraints; from these data, a list of acceptable catego-ries is derived. Similarly, the minimum humidity regime required by each species gives rise to a list including that level of moisture or higher.

The suitability of cultivation was computed with a lo-gical algorithm. For each crop, every location was clas-sified according to its ability to withstand temperature and water availability conditions at each location: viable (V), all three parameters (winter and summer types and humidity regime) allow cultivation of the variety; not viable (NV), at least one of the thermal parameters (type of summer or type of winter) is not adequate; and viable with irrigation (Virr), the humidity regime does not grant a sufficient amount of water to ensure optimal productivity, but the thermal conditions are adequate.

Adequacy of Papadakis’ methodology based on present-day crop occurrence

To evaluate the fitness of the methodology developed by Papadakis to predict the agricultural potential of a re-gion’s climate with high-resolution data, the types of winter and summer of each location in which either of the five target species are present today were checked against the list of tolerated climate types of the corresponding species. Ideally, all present-day crops should be placed in locations with climate regimes that they can tolerate, and none of them should be found in areas with a climate in which they cannot thrive. To perform this analysis, the 2012-2020 agroclimatic scenario was generated by means of the va-riation constants shown in Table 1. The suitability of the humidity regime or presence of irrigation systems were not taken into account, because crops can hardly be present in areas that are too cold or too warm for them; however, some may exist in places that are too dry but have unregistered irrigation support, and some others may be grown in irriga-ted fields even if they could be grown as rainfed crops, just because water supply is readily available. Therefore, the rate of coincidence between “Papadakis climate suitability prediction” and “presence of a crop” was calculated using the type of summer and type of winter at each location only.

Water deficit

The water deficit (WD) of all five crops was calculated for every month by multiplying the Kc crop constant by monthly potential evapotranspiration (ETP0) (Allen et al., 1998), and subsequently obtaining the balance between culture evapotranspiration and effective rainfall.

Following Papadakis’ methodology, monthly ETP0 was calculated as

where eTmax and etmin are the vapor pressures correspon-ding to Tmax or tmin in every month:

Monthly effective precipitation (Pef)m was calculated according to Brouwer & Heibloem (1986):

Kc constants were obtained from the FAO repository (FAO, 2002) except for olive (Pastor & Orgaz, 1994) and orange (Castel, 2001) (Table S3 [suppl.]).

Absolute minimum temperature (tminabs) data

To perform agroclimatic analyses following Papada-kis’ approach, three of the four required variables (ave-rage monthly Tmax, tmin and P) were already publicly available from the DCAC, but data layers for the additio-nally required variable (average monthly tminabs) were generated in this work. The four independent variables (Alt, DistSea, CosLat, and Rad) showed behaviours and correlation coefficients (Table S4 [suppl.]) comparable to those of other DCAC thermal data layers. Maps of tmi-nabs for the 12 months of the year were obtained using the UK interpolation technique (Fig. S1 [suppl.]), which had a RMSE error between 1.13 and 1.69 ºC, lower than multiple regression (Table S5 [suppl.]).

Papadakis’ agroclimatic zoning

In the baseline scenario, the warm and cool mariti-me thermal regimes occupied each approx. 40% of the surface of Catalonia and coincided with semi-arid agri-cultural areas in the interior (which is very dependent on irrigation) and the coast (which has a tradition for seve-ral rainfed crops). The cold maritime regime coincided approximately with the Pyrenees subregion and occupied more than 18% of the study area. There was also a fourth type, the semi-warm subtropical regime, present only in the Ebro Delta and in some sparse points in the southern half of the coast (Table 2).

As for the humidity regime, the Pyrenees subregion was found to have humid or humid Mediterranean re-gimes, the coastal subregion mostly dry Mediterra-nean, and the interior mostly semi-arid Mediterranean (Table 3).

When visualized as maps, the comparison between the baseline and future scenarios showed that the warm coas-tal zone will advance towards the interior as the colder regimes retreat, mainly due to the expansion of Oryza and Gossypium summer types and the Citrus winter type (Fig. 3), which will push colder types inland. Simultaneously, the semi-arid Mediterranean humidity regime will expand from the interior towards the coast as humid regimes re-cede remarkably toward higher altitudes.

Table 2.Definition and percentage of territory occupied by each thermal regime in the baseline and future scenarios under study.

Figure[1] Ci=cool Citrus; av=cool Avena; Ti=warm Triticum; g=cool Gossypium. [2] O=Oryza; M=maize; T=Triticum; P=polar. [3] The definitions of types of winter and summer are provided in Table S1 [suppl.].

Table 3.DPercentage of territory occupied by each humidity regime in the baseline and future scenarios under study.

FigureThe definitions of humidity regimes are provided in Table S1 [suppl.].

Suitability of cultivation

The impact of the changes described in temperature and moisture conditions was evaluated by calculating the suitability of five typical Mediterranean crops which together represent about 45% of the cultivated land in Catalonia, according to the 2018 Single Agrarian Declaration (DUN2019) (Table S6 [suppl.]).

Initially, the performance of Papadakis’ methodology as an instrument for predicting the potential distribution of crops was evaluated by calculating the percent coincidence between the existence of a crop and Papadakis’ suitability class at the corresponding location. In the year 2018, 100% of the registered fields growing alfalfa, barley and olive trees were located in areas where winter and summer types are suitable for the corresponding crop according to Papadakis (i.e., none of these traditional crops are grown in areas that Papadakis’ classification would consider inadequate for the crop) (Table 4). A more recently established crop, orange, showed a rate of coincidence above 90%. In contrast, the viability of almond seems to be poorly predicted by this classification system (nearly 40% of present-day almond orchards are in areas that Papadakis’ method would consider too cold, as discussed below).

When climate suitability of the five species was visualized as maps (this time taking into account humidity along with thermal regimes), the current real distribution of crops was found to largely reflect these agroclimatic computations (Fig. 4). However, two apparent contradictions may be observed, caused by water availability (rather than need) and the presence of varieties adapted to local climates.

On the one hand, in the Littoral subregion, barley is grown with irrigation in locations where these crops would be suitable in the drylands, just because irrigation is readily available in these areas. On the other hand, in the Interior subregion some almond plantations exist which, according to Papadakis’ categorization, are in an excessively cold area. This fact probably indicates a limitation of Papadakis’ system to predict the viability of almond cultivars, some of which can withstand harsher winters than the coldest category that can be reasonably applied (i.e., almond withstands warm Avena winters or warmer; some cultivars may withstand the milder versions of the cool Avena category, above a threshold that is not discerned by this classification).

This procedure was then applied to compare agroclimatic suitability for the baseline scenario and the 203150 scenario. No transitions from V to NV or Virr to NV were detected for any of the studied crops, but the opposite changes (NV to V or NV to Virr) do take place (Table 5). Therefore, the change to warmer winter and summer conditions will not reduce the viability area of any of the crops, but it will enable some of them to be grown in areas that have been traditionally too cold for them, as long as water demands are covered. On the other hand, for most crops the transitions V to Virr and NV to Virr are abundant, mainly due to the expansion of the semi-arid Mediterranean regime and the retraction of the humid Mediterranean regime (Fig. 3 and Table 5). The woody crops (almond, orange and olive trees) are the species that gain larger potential cultivation areas (19.4%, 12.2% and 7.2% of the study area, respectively).

Upon inspection of these data as maps, it appeared that the suitability of almond will increase thanks to irrigation (19.4%) in highland areas of the plains and inland mountains (Fig. 5, Table 5). Nevertheless, an important part of the study area (39.6%) will remain unsuitable for this crop, except for some late-flowering varieties that may be suited for slightly colder conditions than those of the coldest Papadakis categories that can be applied. Orange trees may become a viable crop along almost the entire Catalan coast in the next few decades, as long as irrigation is a possibility. In general terms, however, 86.6% of the study area will remain unsuitable for orange growing (Fig. 5e, Table 5). Olive trees will still be suitable as a rainfed crop on 41.8% of the land, while if irrigation is provided they will be viable in an additional 44.9% of the area. This implies an increase of 15.4% in areas in which irrigation would be mandatory for olive production.

Regarding herbaceous crops, barley will maintain a larger area of cultivation suitability without irrigation (49.6%), although 45% of the land will require irrigation and, in the timeframe under study, this land will increase by 15.4%. This will inevitably lead to a drop of productivity in many of the areas that need irrigation, since barley production has a meagre economic margin and is often not worth irrigating. Alfalfa will have a larger area requiring irrigation (a 9.8% increase), which will reach 75.1% of the land. As with barley, in many areas of the country, rainfed lands will presumably be less productive for alfalfa compared to the present day.

Figure 3.Maps of summer and winter types and humidity regime according to Papadakis in the baseline and future scenarios.

Table 4.Percentage of coincidence with Papadakis thermal suitability for each of the five selected crops over the study area.

FigureThe definitions of humidity regimes are provided in Table S1 [suppl.].

Figure 4.Left panel: Cultivation suitability of the five targeted crops in the 2012-21 scenario. Green, blue and red indicate V, Virr and NV, respectively. Right panel: Actual fields in which each crop is grown with (blue) or without (yellow) irrigation.

Combined water deficit of the five species under study

In order to quantify the increasing difficulties of cultivation conditions in terms of water stress, the water deficit of all irrigated fields of the DUN2019 that contain one of the five target crops was calculated, and water requirement of every irrigated field according to its surface was obtained. This subset added up to 67,535 fields (82,850 ha) and therefore constituted about 11.5% of the total farmland of Catalonia and one third of the total irrigated f ields according to the DUN2019 (Table S6 [suppl.]).

The combined water demand of the fields that currently have one of the five crops under irrigation will increase by more than 16% between the base and future scenarios, going from 330 to 383 hm3 (Table 6). Percentage-wise, this increase in water demand will be more pronounced in coastal areas and the Pyrenees, although most of the consumption will continue to be in the Interior subregion (more than 80%), largely due to alfalfa and barley (Table S7 [suppl.]).

The calculation of tminabs following previously established methodology generated results with a statistical significance comparable to that of other layers of information already published (Ninyerola et al., 2000); therefore, it provides a solid basis for subsequent agroclimatic studies. The fact that UK is superior to multiple linear regression as an interpolation method for climate variables has been observed previously (Bostan et al., 2012).

Papadakis’ classification was chosen as a well-known procedure to visualize the agricultural potential of a region’s climate. The current distribution of crops in the study area was found to fit Papadakis’ thermal suitability predictions in four out of the five selected species. In combination with climate change projections, this methodology enabled the visualization and quantification of the impacts of climate change on agriculture in the region.

Agroclimatic zoning of baseline and future scenarios showed that a substantial increase in subtropical thermal regime (+20.1%) and a drastic reduction of cool and cold maritime regimes (-20.1% taken together) will occur, accompanied by the expansion of the semi-arid Mediterranean humidity regime (which will become the most common regime in the future scenario). Our results for the baseline scenario are comparable to the Papa-dakis climate layers available at the SIGA agricultural data viewer (Spanish Ministry of Agriculture, 2012), which have a lower resolution than those used in the present study.

As a consequence of these changes, the majority of crops will be viable in larger areas, but the new high-alti-tude areas with a higher rainfall (where cultivation suita-bility changes from NV to V) will not quite compensate the increase of the areas in which irrigation will become mandatory. Therefore, all crops will have a larger poten-tial cultivation area, but they will be more dependent on irrigation availability. Irrigation is known to decrease sen-sitivity to heat stress, and for some areas and crops, war-mer conditions may lead to increased yields, as long as water needs are met (Zaveri & Lobell, 2019).

The changes in humidity regime will bring about an increase in water demand of the studied crops (together representing about a third of the irrigated land of Cata-lonia) of about 16% on average. In absolute terms, the increase of 53 hm3 to irrigate the five targeted species are comparable to the increase of 197 hm3 for the total irriga-ted land of Catalonia predicted in the TRCCC (Sebastià et al., 2016). Most of the water consumption of these species (> 80%) will continue to be in the Interior, which is where most of the irrigation infrastructure in Catalonia is found. Through improved water-management and irrigation techniques, however, reductions of water use of up to 40% can be achieved (Kourgialas et al., 2019).

It should be noted that in regional predictive models of climate change, precipitation projections of each season have a greater uncertainty than temperature projections (Gonçalves et al., 2014). The need for better rainfall pre-dictive models to improve agricultural performance pre-dictions has been underscored in multiple studies (Watson & Chalinor, 2013; Ullah et al., 2017). In addition, Papa-dakis’ methodology assumes a soil water carrying capa-city of 100 mm if real values are unknown. When sound high-resolution data for this variable become available, the methodology followed in this study will generate more precise predictions of evapotranspiration, and therefore of humidity regime and water deficit (Folberth et al., 2016). A high-resolution soil map of Catalonia is currently under development (ICGC, 2019), which will enable the esti-mation of soil water carrying capacity all over the region.

Climate change, therefore, can significantly affect the distribution of crops in Catalonia and generate structural changes in production. With continued improvement of data and modelling, priority areas and types of interven-tions can be defined with a very detailed resolution. Herba-ceous and woody crops may use different adaptation stra-tegies. For almonds, the poor performance of Papadakis’ methodology to predict present-day orchards may reflect an ongoing adaptation to climate change: old orchards are being abandoned or replaced with hybrids adapted to shor-ter cold seasons and late-flowering varieties that enable harvest in mountainous areas (Vargas et al., 2008), where almonds can still be produced as a rainfed crop. Oranges are predicted to have a large expansion potential; however, this new potential cultivation area is restricted to the coast, where urban uses of water and land are greatest and new irrigation developments are unlikely. Therefore, the current citric production areas will remain viable with irrigation, but the expansion of this crop will depend on strongly determining factors beyond climate suitability. On the other hand, olive trees will have a larger potential cultivation area with extended growing seasons (Benlloch-González et al., 2018); however, in certain areas of the Western Mediterranean a loss of productivity is expected due to increased water stress (Fraga et al., 2019). The remarkable expansion of suitable olive-growing areas with a strong dependence on irrigation observed in this study is consistent with previous studies with a European scope (Tanasijevic et al., 2014). Regarding herbaceous crops, barley has traditionally been grown during winter and spring, and it is likely to be strongly affected by the decreased spring precipitation. Rainfed barley fields may require the use cultivars suited for early autumn sowing and early spring harvest, given that this species is more sensitive to water deficiency and temperature extremes in the early stages of its growth cycle (Cammarano et al., 2019). Alfalfa has higher water requirements; in areas were precipitation is sufficient, the increased temperature and CO2 concentration may benefit productivity (Thivierge et al., 2016), but these environments will be restricted to the Pyrenees mountain range, and this crop will require irrigation in most of the study area.

In irrigated areas, if sufficient water resources are available, the impacts on production and crop replacement may be lower, but efficient water-management systems will be needed. In traditional rainfed agriculture areas, especially the coastal and inland locations South of Barcelona, these impacts are likely to be drastic, and a substitution of crops or the establishment of support irrigation systems must be considered to avoid a significant drop in productivity, as suggested by a simulation of rainfed crop yield that covers all of Europe at a lower resolution for the 2030 time-horizon (Donatelli et al., 2015). The transition to warmer conditions will also favour some pests associated with the different crops, which may influence the adaptive response, economic viability and productive area (Ponti et al., 2014). In some instances, however, pests may be negatively influenced by warmer conditions and advanced flowering times (Aurambout et al., 2009). In the Pyrenees subregion, some Mediterranean crops may become interesting new options, relative to alfalfa and other animal fodder that are currently grown. However, even in these mountainous regions, some varieties will need irrigation due to the strong Mediterraneanisation of the area. In addition, the higher frequency of extreme phenomena (warm nights, extreme temperatures and precipitation) that are expected in the study area (Barrera-Escoda et al., 2014) can have a stronger negative impact than the changes in average temperatures and precipitation (Trnka et al., 2015; Ullah et al., 2017). Most importantly, multiple social and economic factors determining adaptive capacity will lead to different impacts of climate change in different areas (Kahil et al., 2015; Parker et al., 2019). The highest-altitude areas, above 1500 m, will remain clearly inhospitable for agriculture.

The case-study presented here underscores the importance of improved predictions of seasonal precipitation and temperature changes for local agricultural planning and recovers a classical methodology to evaluate the impact of these changes on climatic suitability of common crops. The procedure used in this work enables the detection of climate-derived risks and opportunities for the agricultural sector with a high spatial resolution.

Table 5.Percentage of territory occupied by each combination of suitability in the baseline scenario and suitability in the future scenario for the studied crops. The last three rows are the result of operating with the rows above (M=E+G-B-J; N=-C+G-I-J; O=M-N).

FigureFigure 5.Cultivation suitability of the five targeted crops in the baseline (left) and future (right) scenarios. Green, blue and red indicate V, Virr and NV, respectively.

Table 6.PWater deficit (WD) of irrigated (R) and dryland fields (S) with one of the five targeted crops in the base scenario (mm, average ± SD), increase estimated for the 2031-50 scenario (var WD), total gross irrigation needs (gIN) in the baseline and future scenarios (hm3), and percent-increase of the total gIN.