Sugar Beet Agronomic Performance Evolution in NW Spain in Future Scenarios of Climate Change

: Changes in environmental conditions resulting from Climate Change are expected to have a major impact on crops. In order to foresee adaptation measures and to minimize yield decline, it is necessary to estimate the e ﬀ ect of those changes on the evapotranspiration and on the associated irrigation needs of crops. In the study presented herein, future conditions extracted from RCP4.5 scenario of IPCC, particularized for Castilla-y-Le ó n (Spain), were used as inputs for FAO crop simulation model (AquaCrop) to estimate sugar beet agronomic performance in the medium-term (2050 and 2070). A regional analysis of future trends in terms of yield, biomass and CO 2 sequestration was carried out. An annual ET 0 increase of up to 200 mm was estimated in 2050 and 2070 scenarios, with ETc increases of up to 40 mm / month. At current irrigation levels, temperature rise would be accompanied by a 9% decrease in yield and a ca. 6% decrease in assimilated CO 2 in the 2050 and 2070 scenarios. However, it is also shown that the implementation of adequate adaptation measures, in combination with a more e ﬃ cient irrigation management, may result in up to 17% higher yields and in the storage of between 9% and 13% higher amounts of CO 2 .


Introduction
Spring-sown sugar beet is an industrial crop of great importance in Castilla-y-León region (Northwestern Spain), which accounts for 87% of Spanish production of spring-sown sugar beet, with over 24,000 ha [1], and which is the area of the European Union that achieves the highest yields per hectare [2]. As it is also the case for many other crops in Continentalized Mediterranean climate areas, sugar beet requires more water than that provided by rainfall, and thus irrigation is necessary to satisfy its water requirements. Besides, irrigation is the most determining factor in its production, being an indispensable practice in Spain [3]. However, the availability of water for crop irrigation is expected to decrease in the future due to increased demands from other sectors (drinking and household needs, recreation, industry and commerce, etc.) and because of changes in environmental conditions [4,5]. The latter are, in fact, the main source of uncertainty for the viability of sugar beet cultivation in this region in the future.

Baseline Scenario Climatic Data
As indicated in the flowchart shown in Figure S1, daily climatic data from 2001 to 2014 was collected from 29 weather stations that belong to the SIAR (Agroclimatic Information System for Irrigation) network of the MAPA (Spanish Ministry of Agriculture, Fisheries and Food). This climatic data was used to calculate a representative meteorological year for each season, in order to build the baseline scenario for the study. This time series was chosen over other longer (but less local) data series because field data can improve the regional projections of crop models [31]. Mann-Kendall tau test was conducted to detect trends in the dataset.

Projected Climatic Data for 2050 and 2070 Scenarios
Climate data projected for 2050 and 2070 in the locations of the SIAR stations were obtained through the WorldClim global climate layers [32](http://www.worldclim.org). In this project, different global climate models output data from CMIP5 (IPCC Coupled Model Intercomparison Project Phase 5) were downscaled and calibrated (bias corrected) using WorldClim 1.4 as a reference baseline "current" climate [33]. More information on CMIP5 coordinated multi-model dataset, which ensembles 40 GCMs (Global Climate Model) from 20 research groups, may be found in Taylor et al. [34]. Future climate data generated with those GCMs usually has a spatial resolution of hundreds of kilometers, which is problematic for regional studies that consider variation at much higher spatial resolution. Hence, high-resolution information from low-resolution variables needs to be inferred through a downscaling process, which can be conducted in different ways [35]. In particular, WorldClim project uses a methodology that assumes that change in climate is relatively stable over space (high spatial autocorrelation).
The layers selected in this study (monthly average minimum temperature, monthly average maximum temperature and monthly total precipitation) were projections of the Earth system model MPI-ESM-LR, developed by the Max Planck Institute for Meteorology (MPI-M) for RCP4.5, and had

Baseline Scenario Climatic Data
As indicated in the flowchart shown in Figure S1, daily climatic data from 2001 to 2014 was collected from 29 weather stations that belong to the SIAR (Agroclimatic Information System for Irrigation) network of the MAPA (Spanish Ministry of Agriculture, Fisheries and Food). This climatic data was used to calculate a representative meteorological year for each season, in order to build the baseline scenario for the study. This time series was chosen over other longer (but less local) data series because field data can improve the regional projections of crop models [31]. Mann-Kendall tau test was conducted to detect trends in the dataset.

Projected Climatic Data for 2050 and 2070 Scenarios
Climate data projected for 2050 and 2070 in the locations of the SIAR stations were obtained through the WorldClim global climate layers [32] (http://www.worldclim.org). In this project, different global climate models output data from CMIP5 (IPCC Coupled Model Intercomparison Project Phase 5) were downscaled and calibrated (bias corrected) using WorldClim 1.4 as a reference baseline "current" climate [33]. More information on CMIP5 coordinated multi-model dataset, which ensembles 40 GCMs (Global Climate Model) from 20 research groups, may be found in Taylor et al. [34]. Future climate data generated with those GCMs usually has a spatial resolution of hundreds of kilometers, which is problematic for regional studies that consider variation at much higher spatial resolution. Hence, high-resolution information from low-resolution variables needs to be inferred through a downscaling process, which can be conducted in different ways [35]. In particular, WorldClim project uses a methodology that assumes that change in climate is relatively stable over space (high spatial autocorrelation).
The layers selected in this study (monthly average minimum temperature, monthly average maximum temperature and monthly total precipitation) were projections of the Earth system model MPI-ESM-LR, developed by the Max Planck Institute for Meteorology (MPI-M) for RCP4.5, and had a 30-s (of a longitude/latitude degree) spatial resolution (about 900 m at the equator), and 1 month temporal resolution. The MPI-ESM consists of coupled general circulation models for the atmosphere and the ocean, as well as subsystem models for land and vegetation, and for the marine biogeochemistry. Thus, the carbon cycle has been added to the model system [36].
RCP4.5 scenario, which represents stabilization without overshoot pathway to 4.5 W·m −2 at stabilization after 2100 [37,38], was chosen because it is not as optimist regarding GHG reduction as RCP2.6, but it does consider a reduction in greenhouse gases starting before 2050. In this model, global mean surface temperature change is estimated at 1.4 • C in 2046-2065, and at 1.8 • C in 2081-2100.

Calculated Reference Evapotranspiration and Crop Evapotranspiration
Penman-Monteith simplified equation [39], adopted by the FAO, was used for the calculation of monthly and annual ET 0 values, both in 2050 (average for 2041-2060) and 2070 (average for 2061-2080), using temperature and precipitation data from previous section. Wind, humidity and radiation parameters were assumed stationary. The choice of this method would be supported by the fact that many studies have successfully applied it to different climates and time scales [40]. Moreover, in Spain it has been used, for example, by Espadafor et al. [9] and by Vicente-Serrano et al. [10] to examine historical trends of ET 0 .
To validate this method, calculated values were compared with real values coming from the stations, and regression lines were obtained, obtaining a coefficient of determination R 2 of 0.998 (thus confirming that the method of calculation was perfectly acceptable).
Monthly crop evapotranspiration (ETc) was calculated as the product of monthly ET 0 and Kc crop-specific coefficients. Monthly values for sugar beet K C in the area of study were obtained from AIMCRA (Spanish Research Association for Sugar Beet Crop Improvement) [41].

AquaCrop
A detailed description of the model can be found in [42] and [18]. Minimum and maximum temperatures, ET 0 , rainfall and CO 2 concentrations (from Mauna Loa Observatory (Hawaii) records and estimated values for the future, discussed in Section 2.3) were supplied as climate inputs. An appropriate irrigation schedule for crops in this region, based on AIMCRA recommendations [41], with a total dose of 553 mm during the whole cycle, was chosen. This irrigation dose was considered as fixed for predictions in the future scenarios.

Crop Model Calibration
By default AquaCrop offers files for the simulation of different crops, and in the case of sugar beet, the model is automatically calibrated and validated to Foggia (Italy) in 2000. Stricevic et al. [43] calibrated the model for the specific conditions of northern Serbia, concluding that this calibration only implied small changes of a few of the default model coefficients, illustrating the resilience of the model. Garcia-Vila et al. [25] recently calibrated and validated the model for different irrigation water allocations in the two main producing areas in Spain. Consequently, in this study the crop parameters for sugar beet were adjusted taking into consideration aforementioned works to obtain typical yields in the area of study (of over 100 t/ha).
In general, it is more suitable to study the different crop stages through the growing-degree days (GDD), to better reflect the plant physiology [17]. In this study, GDD was used for the baseline scenario, and both GDD and days were used for future projections of the crop growth cycle (GDD for comparisons with the baseline scenario and days for the extended cultivation period calculations). The necessary GDD to achieve each growth stage (Table 1) were chosen on the basis of field data collected in different locations in the area of study [44] and on data available in the literature. Details on sowing and harvesting dates have been reported in a previous paper [44]. As in the calibration proposed by Garcia-Vila et al. [25], it was also decided to slightly increment the water productivity (WP) parameter. This modification would be supported by the conclusions of previous research works that suggest that, although sugar beet is a C3 species, it is very efficient in water use, with a behavior closer to that of C4 crops [48]. Planting density in the model was increased to 125,000 plants/ha, which was the density used in real cultivation conditions [41,44]. The harvest index (HI) was kept at 70%, in agreement with Martínez Quesada [49] and with field data. As in the study by Stricevic et al. [43], soil fertility was not addressed, given that nutrient requirements were fully satisfied following AIMCRA recommendations [41]. Soil types were not specified either.
Validation was conducted by calculating the error between real production data, available from experiments conducted in 2011 and 2012 growing seasons, and that calculated with AquaCrop. For this, as in other works [50], the root mean square error (RMSE) and the normalized root mean square error (RMSEn) were calculated. The global Root Mean Square Error (RMSE) was determined as: where Ai = experimental yield, Si = simulated yield and n = number of observations. Units are t/ha of dry matter yield. The normalized root mean square error was calculated as: whereĀ = mean observed data. Values of RMSEn smaller than 10% are considered as excellent, between 10 and 20% as good, between 20% and 30% as fair and, if larger than 30%, as poor [51].

Carbon Sequestration
Data about carbon content in the different parts of the sugar beet was obtained from experimental data [44], with average values of 43.5% in roots and 37.5% in leaves. Net CO 2 uptake (i.e., carbon storage) was obtained by multiplying dry matter, C content and the C to CO 2 conversion factor (44/12, i.e., the ratio between the molar mass of CO 2 and the molar mass of C).

Cartography and Spatial Interpolation
ArcGIS v.10 software (Esri, Redlands, CA, USA) was used to extract monthly average minimum and maximum temperatures and monthly total precipitation data (used as inputs to the crop model) from WorldClim layers for the coordinates of every weather station. It was also used to generate annual reference evapotranspiration, monthly crop evapotranspiration, annual yield, annual biomass and annual CO 2 sequestration maps. These are shown in the results section below. As for spatial interpolation, ordinary kriging method in Geostatistical Analyst toolbox was used. A spherical isotropic semivariogram [52] was chosen on the basis of RMSE criterion. IDW (Inverse Distance Weighting) was avoided in order to minimize points isolated by bull's-eye effect, while Spline was not chosen either, since it would be more appropriate for smaller scales [53]. The spatial analyst toolbox from ArcGIS was used to implement this interpolation.
Interpolation validation was conducted by "leave one out" cross-validation, which consists in using one of the stations as the validation dataset, and using the rest of stations as the training set to calculate the error of the resulting model in that validation station; and then repeating this process for each and every station in the dataset.

Evapotranspirationn Baseline
Monthly average temperature, precipitation and ET 0 values from the 29 weather stations, with daily data from 2001 to 2014, are shown in Figure 2a.

Evapotranspirationn Baseline
Monthly average temperature, precipitation and ET0 values from the 29 weather stations, with daily data from 2001 to 2014, are shown in Figure 2a. It may be observed that the highest ET0 values were found between June and August, while the lowest ones corresponded to the November to March period, as expected in a continental Mediterranean climate.
Despite the fact that the general trend in Spain in the last decades has been towards an increase in ET0 [10], it is worth noting that Mann-Kendall trend test did not identify any clear trend in ET0 data for the years considered herein. This would be beneficial when it comes to the creation of a "representative year" model.

WorldClim Data
Annual average temperatures for each station in the area of the study are shown in Figure 2b. The average temperature increase of the whole set of stations was 2.4 and 2.7 °C for 2050 and 2070, respectively. It should be noted that these values would be higher than the global averages for the RCP4.5 scenario. Ribalaygua et al. [54], working with SRES scenarios, reported increases in maximum and minimum temperature averages ranging from 1.5 to 2.5 °C (depending on the scenario), relative to the 1971-2000 period, for mid-21st century in Aragón (north-eastern Spanish region). In that study, regarding rainfall, authors explained that there was not a clear trend, but rather higher uncertainties. However, all the scenarios suggest a moderate decrease in rainfall for the midcentury (2%-4%). In this work, the percentage of change of annual precipitations for 2050 and 2070 in the area of study (not shown) was minimum compared to the current scenario, so it should not influence the results in a significant manner.

Crop Model Validation
The minor adjustments made to the calibration discussed in Section 2.6.1 showed an adequate performance when applied to the area of study, in agreement with Heng et al. [55]. In comparison It may be observed that the highest ET 0 values were found between June and August, while the lowest ones corresponded to the November to March period, as expected in a continental Mediterranean climate.
Despite the fact that the general trend in Spain in the last decades has been towards an increase in ET 0 [10], it is worth noting that Mann-Kendall trend test did not identify any clear trend in ET 0 data for the years considered herein. This would be beneficial when it comes to the creation of a "representative year" model.

WorldClim Data
Annual average temperatures for each station in the area of the study are shown in Figure 2b. The average temperature increase of the whole set of stations was 2.4 and 2.7 • C for 2050 and 2070, respectively. It should be noted that these values would be higher than the global averages for the RCP4.5 scenario. Ribalaygua et al. [54], working with SRES scenarios, reported increases in maximum and minimum temperature averages ranging from 1.5 to 2.5 • C (depending on the scenario), relative to the 1971-2000 period, for mid-21st century in Aragón (north-eastern Spanish region). In that study, regarding rainfall, authors explained that there was not a clear trend, but rather higher uncertainties. However, all the scenarios suggest a moderate decrease in rainfall for the mid-century (2%-4%). In this work, the percentage of change of annual precipitations for 2050 and 2070 in the area of study (not shown) was minimum compared to the current scenario, so it should not influence the results in a significant manner.

Crop Model Validation
The minor adjustments made to the calibration discussed in Section 2.5.1 showed an adequate performance when applied to the area of study, in agreement with Heng et al. [55]. In comparison with experimental data from 2011 and 2012 growing seasons, errors remained below 10% in all cases, except for one plot in 2012 (14.72%) ( Table 2). The yield RMSE value, of 2 t/ha, and the yield RMSEn, below 10% (rated as 'excellent', according to [51]), imply that the model would be adequate with a view to analyzing spatial trends for the crop in future scenarios of Climate Change.

Interpolation Validation
A "leave one out" cross-validation of ET 0 values was used to assess the behavior of the spatial kriging interpolation applied to the variables under study. With a RSME of 69.5 mm (5.2% mean error) between observed and interpolated values, this interpolation method may be regarded as a consistent tool for the creation of maps that are representative of the spatial distribution of the studied variables (Table S1).

ET 0 and ETc
In the baseline scenario, the central and southern zones of the area of study showed higher ET 0 values, decreasing as one moves towards the periphery, especially towards the north and northeast (Burgos province). In future scenarios, a clear annual ET 0 increase was calculated for all the area under study both in 2050 and in 2070 (Figure 3), more marked for the latter, with differences vs. the baseline scenario of up to 200 mm ( Figure 4). This increase would be a direct consequence of the temperature rise, given that temperature was the only parameter modified in FAO-56 PM equation. below 10% (rated as 'excellent', according to [51]), imply that the model would be adequate with a view to analyzing spatial trends for the crop in future scenarios of Climate Change.

Interpolation Validation
A "leave one out" cross-validation of ET0 values was used to assess the behavior of the spatial kriging interpolation applied to the variables under study. With a RSME of 69.5 mm (5.2% mean error) between observed and interpolated values, this interpolation method may be regarded as a consistent tool for the creation of maps that are representative of the spatial distribution of the studied variables (Table S1).

ET0 and ETc
In the baseline scenario, the central and southern zones of the area of study showed higher ET0 values, decreasing as one moves towards the periphery, especially towards the north and northeast (Burgos province). In future scenarios, a clear annual ET0 increase was calculated for all the area under study both in 2050 and in 2070 (Figure 3), more marked for the latter, with differences vs. the baseline scenario of up to 200 mm ( Figure 4). This increase would be a direct consequence of the temperature rise, given that temperature was the only parameter modified in FAO-56 PM equation.  Within the area of study, the areas that would be most affected by the increase in annual ET0 would be the province of Valladolid, the south of Palencia, the north of Salamanca and Ávila, the   Within the area of study, the areas that would be most affected by the increase in annual ET0 would be the province of Valladolid, the south of Palencia, the north of Salamanca and Ávila, the northeast of Segovia, and the east of Zamora. That is, the annual ET0 increase in future scenarios would make the spatial distribution differences that already exist in the baseline scenario more dramatic. The increase would be more marked in the central zone, with a clear southwards direction Within the area of study, the areas that would be most affected by the increase in annual ET 0 would be the province of Valladolid, the south of Palencia, the north of Salamanca and Ávila, the northeast of Segovia, and the east of Zamora. That is, the annual ET 0 increase in future scenarios would make the spatial distribution differences that already exist in the baseline scenario more dramatic. The increase would be more marked in the central zone, with a clear southwards direction (and slightly to the west) so that areas with lower annual ET 0 would move towards the north and northeast over time.
This same trend was also reflected in the monthly differences in ETc values with respect to the baseline scenario, depicted in Figure 5. Between March and May, monthly ETc increases ranging from 1.5 to 20 mm/month are foreseen in the future scenarios. July would be the month in which the monthly ETc increase would be maximum, of up to 40 mm/month. (and slightly to the west) so that areas with lower annual ET0 would move towards the north and northeast over time. This same trend was also reflected in the monthly differences in ETc values with respect to the baseline scenario, depicted in Figure 5. Between March and May, monthly ETc increases ranging from 1.5 to 20 mm/month are foreseen in the future scenarios. July would be the month in which the monthly ETc increase would be maximum, of up to 40 mm/month.

Yield
Assuming that irrigation doses are not increased, yield would be affected by changes in ET0 (Figure 6a). Taking into consideration the pattern for annual ET0 evolution described above, yield would decrease by ca. 9% in both 2050 and 2070 in the central zone of the area of study. An in-depth analysis of modelled effects of sugar beet responses to different irrigation doses may be found in [25].
It is worth noting that the decrease in yield would be slightly larger in 2050 than in 2070. This unexpected result may be tentatively ascribed to a higher photosynthetic activity in 2070 resulting from higher CO2 concentrations. That is, the negative effect of higher temperatures would be partly compensated by larger photosynthetic rates due to the expected CO2 increase [56]. This would be consistent with studies that have carried out FACE (Free Air Carbon Enrichment) experiments to simulate future scenarios with larger CO2 concentrations, such as those by Manderscheid et al. [57], finding a yield increase between 7% and 16% for the CO2 concentration levels foreseen for mid-21st

Yield
Assuming that irrigation doses are not increased, yield would be affected by changes in ET 0 (Figure 6a). Taking into consideration the pattern for annual ET 0 evolution described above, yield would decrease by ca. 9% in both 2050 and 2070 in the central zone of the area of study. An in-depth analysis of modelled effects of sugar beet responses to different irrigation doses may be found in [25]. crop, the region and the chosen adaptation strategies, such as matching crops to soils [29]. Other strategies such as shifting the sowing date, changing the required cultivar growth duration, the development of heat tolerant plants may have to be adopted depending on the location and crop, as noted by Khordadi et al. [30]. Concerning the study presented herein, an important point that should be taken into consideration is that the yield simulations generated by the model in 2050 and 2070 involve changes in the crop cultivation period. While in the baseline scenario sugar beet cultivation goes from March until November, in future simulations the crop season would start earlier, and would finish by October or by the end of September, depending on the area. This responds to the fact that crops, with increasing temperatures, take less time to reach the necessary GDD in each growth stage. Although the temperature rise can increase the developmental rate of the crop, resulting in an earlier harvest, such "heat stress" may have negative effects on crop production [16].
However, the future increase in temperatures could also allow a lengthening of the cultivation period, allowing for earlier sowing. The temperature rise would allow an earlier development of the photosynthetic organ (the leaves), making the most of solar radiation: Due to the slow leaf development in spring, sugar beet crop achieves its highest canopy when the maximum solar radiation of the year has already passed [62,63]. By getting the largest field coverage in the least amount of time, and keeping this coverage for as long as possible, the plant can thus optimize solar radiation interception [64]. In this way, the softer winter temperatures could lead to higher productivity in that part of the year, somehow balancing the losses from the other seasons [56].
According to the simulations, as long as the water requirements of the plant were met with the same water allocation-which involves higher irrigation efficiencies and/or the development of new irrigation strategies [65]-, if the cultivation period was extended, bigger yields could be achieved. In this sense, data about the positive influence of cycle extension on yield, based on field studies conducted in the area of study [66], suggests that yield increases of up to 20% could be attained. This would be in agreement with Hoffmann et al. [67] and Hull et al. [68], who claimed that lengthening of the growing season would have a strong positive effect on sugar beet yield. In Figure 6b, the simulations have been extended until November, generally obtaining larger yields, both in 2050 and 2070 (with an increase of up to 17%). These increases would be more noticeable in the north and eastern zones than in the center and in the south. In 2070, because of higher CO2 concentrations, the increase would be more marked, especially in the northern zone (Palencia and León) and in the east (Burgos). It is worth noting that the decrease in yield would be slightly larger in 2050 than in 2070. This unexpected result may be tentatively ascribed to a higher photosynthetic activity in 2070 resulting from higher CO 2 concentrations. That is, the negative effect of higher temperatures would be partly compensated by larger photosynthetic rates due to the expected CO 2 increase [56]. This would be consistent with studies that have carried out FACE (Free Air Carbon Enrichment) experiments to simulate future scenarios with larger CO 2 concentrations, such as those by Manderscheid et al. [57], finding a yield increase between 7% and 16% for the CO 2 concentration levels foreseen for mid-21st century in the A1B IPCC scenario (SARS scenario). The projections of Vanuytrecht et al. [58] for sugar beet in Belgium showed that although higher temperatures and a shorter growth period alone would reduce potential yield, sugar beet would substantially benefit from the CO 2 fertilization effect (with a mid-century yield increase between 6% and 13%).

Biomass and CO2
On the other hand, an increase in yield would be expected in 2050 in northwestern León and northeastern Burgos, an area that would expand to include also the north of Palencia in 2070, reaching a 10% yield increase. This can be explained because, in these colder zones in the baseline scenario, the plants would actually benefit from the higher temperatures. Therefore, as it has been reported for other crops, future global warming may be beneficial in some regions [50,59], but may reduce productivity in zones where optimal temperatures already exist [60]. Moreover, a shorter crop cycle may result in a reduction of attainable yield [30,61] or in a yield increment [31,59] depending on the crop, the region and the chosen adaptation strategies, such as matching crops to soils [29]. Other strategies such as shifting the sowing date, changing the required cultivar growth duration, the development of heat tolerant plants may have to be adopted depending on the location and crop, as noted by Khordadi et al. [30].
Concerning the study presented herein, an important point that should be taken into consideration is that the yield simulations generated by the model in 2050 and 2070 involve changes in the crop cultivation period. While in the baseline scenario sugar beet cultivation goes from March until November, in future simulations the crop season would start earlier, and would finish by October or by the end of September, depending on the area. This responds to the fact that crops, with increasing temperatures, take less time to reach the necessary GDD in each growth stage. Although the temperature rise can increase the developmental rate of the crop, resulting in an earlier harvest, such "heat stress" may have negative effects on crop production [16].
However, the future increase in temperatures could also allow a lengthening of the cultivation period, allowing for earlier sowing. The temperature rise would allow an earlier development of the photosynthetic organ (the leaves), making the most of solar radiation: Due to the slow leaf development in spring, sugar beet crop achieves its highest canopy when the maximum solar radiation of the year has already passed [62,63]. By getting the largest field coverage in the least amount of time, and keeping this coverage for as long as possible, the plant can thus optimize solar radiation interception [64]. In this way, the softer winter temperatures could lead to higher productivity in that part of the year, somehow balancing the losses from the other seasons [56].
According to the simulations, as long as the water requirements of the plant were met with the same water allocation-which involves higher irrigation efficiencies and/or the development of new irrigation strategies [65]-, if the cultivation period was extended, bigger yields could be achieved. In this sense, data about the positive influence of cycle extension on yield, based on field studies conducted in the area of study [66], suggests that yield increases of up to 20% could be attained. This would be in agreement with Hoffmann et al. [67] and Hull et al. [68], who claimed that lengthening of the growing season would have a strong positive effect on sugar beet yield. In Figure 6b, the simulations have been extended until November, generally obtaining larger yields, both in 2050 and 2070 (with an increase of up to 17%). These increases would be more noticeable in the north and eastern zones than in the center and in the south. In 2070, because of higher CO 2 concentrations, the increase would be more marked, especially in the northern zone (Palencia and León) and in the east (Burgos).

Biomass and CO 2
Besides offering data on yield, AquaCrop also provides results about the total biomass achieved by the crop (Figure 7a), thus allowing to estimate CO 2 sequestration ( Figure 7c). As it occurred with crop yield, a decrease in the average biomass production and in CO 2 assimilation of around 7% in 2050 may be expected when the entire area of study is considered. Such decrease would be slightly mitigated in 2070 (5% decrease) for the same reason discussed above (related to the increase in CO 2 levels). In absolute terms, the assimilated CO 2 would decrease from 49 t/ha in the baseline scenario to ca. 46 and ca. 47 t/ha in the 2050 and 2070 scenarios, respectively.
As in the case of yield, if the cultivation period was expanded, larger average quantities of biomass ( Figure 7b) and captured CO 2 (Figure 7d) would be obtained throughout the area of study, with 8% and 12% increases in 2050 and 2070, respectively. In this case, captured CO 2 , would evolve from 49 t/ha in the baseline scenario to 53 and 55 t/ha (on average) in 2050 and 2070, respectively (9 and 13% increase, approximately).
To sum up, for the studied variables (annual ET 0 , monthly ETc, yield, biomass and CO 2 assimilation), the trends predicted in this work would match in a regional scale what other studies have observed: A northward movement of crop suitability zones, as well as increased crop productivity in Northern Europe [69]. Furthermore, the simulations suggest that there is a greater potential for adaptation in northern, cooler zones, in which the reduction in yields can be compensated by shifting the crop growing season to cooler months [30], by advancing sowing, and by taking advantage of an extended growing period through the use of suitable varieties [70]. Although a regional approach is necessary to assess the effects of Climate Change on future yields and changes in crop suitability, it must be kept in mind that there are many uncertainties associated with this kind of yield simulations, including uncertainties in the GCM models and projections of future climate [70], crop model uncertainties [61,71], assumptions, and observation errors [72]. To these, other likely factors such as an increase in extreme rainfall events and droughts may be added, which should also be taken into consideration in future studies [59,69].
by the crop (Figure 7a), thus allowing to estimate CO2 sequestration ( Figure 7c). As it occurred with crop yield, a decrease in the average biomass production and in CO2 assimilation of around 7% in 2050 may be expected when the entire area of study is considered. Such decrease would be slightly mitigated in 2070 (5% decrease) for the same reason discussed above (related to the increase in CO2 levels). In absolute terms, the assimilated CO2 would decrease from 49 t/ha in the baseline scenario to ca. 46 and ca. 47 t/ha in the 2050 and 2070 scenarios, respectively. As in the case of yield, if the cultivation period was expanded, larger average quantities of biomass ( Figure 7b) and captured CO2 (Figure 7d) would be obtained throughout the area of study, with 8% and 12% increases in 2050 and 2070, respectively. In this case, captured CO2, would evolve from 49 t/ha in the baseline scenario to 53 and 55 t/ha (on average) in 2050 and 2070, respectively (9 and 13% increase, approximately).
To sum up, for the studied variables (annual ET0, monthly ETc, yield, biomass and CO2 assimilation), the trends predicted in this work would match in a regional scale what other studies have observed: A northward movement of crop suitability zones, as well as increased crop productivity in Northern Europe [69]. Furthermore, the simulations suggest that there is a greater potential for adaptation in northern, cooler zones, in which the reduction in yields can be compensated by shifting the crop growing season to cooler months [30], by advancing sowing, and by taking advantage of an extended growing period through the use of suitable varieties [70]. Although a regional approach is necessary to assess the effects of Climate Change on future yields and changes in crop suitability, it must be kept in mind that there are many uncertainties

Conclusions
The spatial distribution of evapotranspiration in 2050 and 2070 was simulated with AquaCrop at a regional scale for spring-sown sugar beet in the region of Castilla-y-León (Spain), based on data from the MPI-ESM-LR climate model and the RCP4.5 emission scenario from the AR5. A clear annual ET 0 increase was observed in all the area of study in 2050 and 2070, with differences vs. the baseline scenario of up to 200 mm, which would result in monthly ETc increases of up to 40 mm in July. Yield (at current irrigation levels) would decrease by ca. 9% in both 2050 and 2070 in the central zone of the area of study. This overall yield decrease would be aggravated in the case of decreasing precipitation levels, or increased frequency of extreme events of drought, which seem to be likely in the future. In a similar fashion, the assimilated CO 2 would decrease from 49 t/ha in the baseline scenario to 46 and 47 t/ha in the 2050 and 2070 scenarios, respectively. However, new opportunities for adaptation may arise by lengthening the sugar beet cultivation cycle, delaying the harvest and advancing the sowing. These measures, along with more efficient strategies of irrigation, could result in higher yields (up to 17% higher) and higher amounts of stored CO 2 (9% and 13% higher in 2050 and 2070, respectively). Keeping in mind the uncertainty and errors associated to these methodologies, the results coincide with the findings of other studies at different scales and in different regions of Europe: The most suitable cultivation zones for some crops would move northwards to cooler zones, and even higher yields may be obtained by the implementation of appropriate adaptation measures. In this context, in order to have tools available to face potential future adverse situations, efforts aimed at minimizing the uncertainty of the methodologies used for the projection of future scenarios, at evaluating the existing strategies of adaptation of the different crops to Climate Change, and at devising new ones are more necessary than ever.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4395/10/1/91/s1, Figure S1: Flowchart summarizing the methodology used in the study; Table S1: Comparison between observed and interpolated annual ET 0 values for the weather stations for "leave one out" cross-validation.