| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Dep. de Producción Vegetal: Fitotecnia, Escuela Técnica Superior de Ingenieros Agrónomos, Tech. Univ. of Madrid, Ciudad Univ. E-28040 Madrid, Spain
Corresponding author (iminguez{at}pvf.etsia.upm.es)
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMethodologyResults and discussionConclusionsREFERENCES |
|---|
Abbreviations: ANOVA, analysis of variance • AOGCM, atmosphere–ocean general circulation model • CV, coefficient of variation • CVspat, spatial coefficients of variation • CVtemp, temporal coefficients of variation • ET, evapotranspiration • ETmax, maximum evapotranspiration • GCM, general circulation model • GIS, geographic information system • RCM, regional climate model • WUEb, water use efficiency for aboveground production
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMethodologyResults and discussionConclusionsREFERENCES |
|---|
General circulation models acceptably simulate many features of the global climate system. They recreate the current and past climate (paleoclimate) on a large scale reasonably well, and in general, the reliability of their future predictions on climate is considered acceptable (IPCC, 1998). Nevertheless, on a regional scale, the results are less satisfying due to the lack of detail in the influence of topographical features and the ocean–land boundary among other reasons (Grotch and MacCracken, 1991).
The shortcomings of current GCMs may prevent their results from being directly applied to impact assessments. Therefore, relative (2xCO2, or future climate)/(1xCO2, or current climate) outputs were analyzed (see Iglesias, 1994, for the Iberian Peninsula). There are some factors that reduce the reliability of GCMs in the Iberian Peninsula such as (i) the peninsula's critical geographic location in terms of climatic processes (Gallardo, 1998), (ii) its small area (
500000 km2), (iii) its complex orography [58% of the territory is at an altitude of >600 m (Fig. 1)
], and (iv) the great diversity of its vegetation (Castro et al., 1995). Furthermore, the water that surrounds the peninsula is considerably warmer than expected for its latitude, and this is a determining factor in the local climate (Font Tullot, 1983). Figure 2
shows the nine cells that are taken into consideration by an AOGCM for Spain. Most of the cells are comprised of cultivated land areas at altitudes as low as 0 to 150 m and as high as 1000 m within a range of 200 km. The experiments available with statistical downscaling techniques and nested regional models have shown that the problem of complex topographical features such as large lake systems and narrow land masses, which are not solved at the resolution of current GCMs, significantly affect the simulation of regional and local scenarios for both precipitation, and to a lesser extent, temperature (Kattenberg et al., 1996; McGregor, 1997; Wilby and Wigley, 1997; Giorgi and Mearns, 1999).
|
|
This report shows how crop models that use the scenarios from RCM-PROMES (Prognostic Model at the Mesoscale) (Castro et al., 1993) enhance the representation of spatial and temporal variability in crop growth and development in Spain. A comparative analysis was performed between the use of output from an AOGCM and RCM to drive crop simulations with DSSAT (Tsuji et al., 1994). The impact of climate change is reflected in the phenology, grain yield, aboveground biomass, and water use of winter wheat, winter barley, and maize, which are regarded as being representative of the common field crops grown in Spain. The direct use of climate model outputs as crop model inputs was considered here as another way of comparing the performance of the climate models. This may highlight how the weather perturbations simulated by the climate models are reflected in the crop model outputs.
|
Methodology |
|---|
|
TOPABSTRACTINTRODUCTIONMethodologyResults and discussionConclusionsREFERENCES |
|---|
Mountains and pastures were not included. A crop model system was used that simulates the behavior of these three reference cereal crops in terms of their phenological development, growth rate, yield, and irrigation demand under both current and future climate conditions (see below).
Climate Scenarios
Global climate scenarios were generated using the AOGCM-HadCM2 model (Mitchell et al., 1995; Mitchell and Johns, 1997) that was developed at the Hadley Centre for Climate Research and Prediction of the United Kingdom Meteorological Office. Regional climate scenarios were generated using the RCM-PROMES model, a 3-D primitive equation, hydrostatic, high-resolution mesoscale model that was developed at the Department of Geophysics and Meteorology at the Complutense University of Madrid, Spain (Castro et al., 1993). The RCM-PROMES model applies a horizontal grid of 50 by 50 km2 (Fig. 1 and 2) and 25 vertical layers. It was one-way nested in the HadCM2, which uses a grid size of 2.5 by 3.75° (latitude by longitude) and 19 vertical layers. The regional model was run for a total area of 2250 by 1950 km2, centered in the Iberian Peninsula, with an orography deduced from the global database of the National Geophysical Data Center (Edwards, 1986). The soil water content was initialized with Mintz and Serafini (1992) dataset values. In the RCM, the soil–atmosphere water exchange was modeled using a parameterization called SECHIBA (Schematisation des Exchanges Hydriques à l'Interface entre la Biosphère et l'Atmosphère) (Ducoudré et al., 1993), in which seven vegetation classes are defined along with eight types of land surfaces, including bare soil. Over each of these land covers, water transfers were computed: Evaporation from soil, transpiration from plants through stomatal and architectural resistances, and water intercepted by the canopy (Ducoudré et al., 1993). A correspondence between SECHIBA land cover types and the CORINE (Coordination of Information of the Environment) Land Cover database (European Environmental Agency, 1997), which defines 89 land uses, had to be previously established according to leaf area index evolution, architectural resistance, and ETmax. The process of extracting and simplifying is of the upmost importance for the parameterization.
The climatic change experiment included two different integrations of the AOGCM: A current climate (1xCO2) control run and a modified climate (2xCO2) scenario run. Both simulations started in 1860 and finished in 2100. Throughout the control run, the greenhouse gas (CO2, N2O, and chlorofluorocarbons) concentration was kept constant and equal to the 1990 concentration (473 ppmv CO2). Perturbed simulation also began in 1860, but the initial CO2 concentration was 341 ppmv (corresponding to 1860 CO2 conc.). Until 1990, the increase of greenhouse gases followed observations (Shine et al., 1990), and from 1990 to 2100, the increment was 1% per year. This increase corresponds to the "business as usual" scenario (Mitchell and Gregory, 1992). The SO4 aerosol effect was not considered in these GCM runs. The RCM-PROMES experiments consisted of two 10-yr simulations: Control and scenario. The time range chosen lasted from 2040 to 2049. The reason for choosing these years is that the CO2 concentration in this period will be twice that of the preindustrial concentration.
Comparisons were performed between the monthly surface temperatures and precipitation and the measured and simulated values for the 10-year period of the 1xCO2 scenario obtained from both models in the Iberian Peninsula (Gallardo et al., unpublished data, 2000). Although more details of the comparison can be found in the previous reference, the most notable results can be summarized as follows: (i) The spatial distributions of the monthly surface temperatures and precipitation are generally quite close to those measured in the RCM-PROMES model. (ii) The annual variation of the mean air temperature throughout the Iberian Peninsula is satisfactorily reproduced by both models although the RCM simulates temperature with greater spatial detail than the AOGCM. (iii) Both models simulate more precipitation than is measured. The RCM values are generally higher than those of the AOGCM in winter and autumn, whereas the opposite is true in the summer months.
For this study, daily data of rainfall, solar irradiance, wind speed, and air humidity and temperature generated from the RCM were used while only monthly values from the GCM were considered. Because crop models need daily climate inputs (see below), a weather generator was used for the GCM-derived scenarios when the models were run but not when a direct comparison was performed between climate scenarios, such as in Table 1 and Fig. 3 and 4 . Weather generators, based on a first-order Markov chain, have been shown to reduce the variance of monthly rain because they assume constant transition probabilities. Synthetic weather data in semiarid regions can then have a reduced variance compared with the variance obtained from measured data. In particular, this was shown for rain data in Spain by Villalobos et al. (1999).
|
|
|
For both scenarios, Table 1 shows the maximum and minimum mean annual air temperatures and the total annual irradiance and rain for central Spain (Basin 1) and southwestern Spain (Basin 2). The direct outputs from the climate models were used. In Basin 1, the data corresponded to GCM Cell 3-3 and the average of 16 RCM cells within Cell 3-3 where land is currently cultivated. The Basin 2 data were those of GCM Cells 2-2 and 3-2 and the corresponding 11 RCM cells (Fig. 2 and 3). There were significant differences in the minimum and maximum air temperatures between the scenarios, as well as in the solar irradiance in both climate models, except for the GCM in Basin 2. However, no significant changes were found in the annual precipitation. The mean annual values indicated that differences between the models arose mainly in the temperatures, solar irradiance, and minimum differences in precipitation. Air temperatures were generally greater in the RCM scenarios than in the GCM scenarios.
The CVtemp and CVspat reflected the high rain variability in all scenarios. Differences between the models were greater at minimum air temperatures. For both basins and each scenario, the differences between the RCM and GCM were approximately 6°C for minimum air temperatures and 4°C for maximum air temperatures. The CVtemp of minimum temperatures for 1xCO2 GCM scenarios were the highest in both basins. In the other cases, the CVtemp of the GCM temperatures were equal or lower to the CVtemp of the RCM temperatures.
The evolution of the average of 10 yr of monthly maximum and minimum air temperatures in Basin 1 is shown in Fig. 4. Each RCM value corresponds to the mean of the cells chosen within Cell 3-3 of the GCM. As in Table 1, the RCM temperatures were generally higher than the GCM temperatures for a given scenario. Nevertheless, the maximum air temperatures in the 1xCO2 RCM scenario were similar to those of the 2xCO2 GCM scenario in autumn, winter, and spring but lower in summer (Fig. 4a).
Crop Simulations
The crop simulation models used in this study are included in DSSAT 3.1 (Tsuji et al., 1994). The CERES (Crop Estimation through Resource and Environment Synthesis) models were: CERES-Wheat (Godwin et al., 1989), CERES-Barley (Otter-Nacke et al., 1991), and CERES-Maize (Jones and Kiniry, 1986; Ritchie et al., 1989).
Model calibration and validation of CERES for wheat and maize were based on Iglesias (1994), Iglesias and Mínguez (1995), and Mínguez and Iglesias (1996). The field experiments carried out at La Canaleja Experimental Farm in Madrid, central Spain (40° 30' N lat; 600 m altitude) during the 1995 through 1996 and 1996 through 1997 growing seasons were used for the calibration and for validation of rainfed barley (Guereña et al., 1998; Díaz-Ambrona, 1999). The wheat variety chosen for the study was a cultivar that could be sown in all irrigated areas with moderate vernalization needs (genetic coef. for vernalization: 
). The sowing dates were those currently used for these cultivars. To account for the different air temperatures in spring and summer among the areas studied, two maize cultivars were chosen: One short-season cultivar representing those sown in the northern regions (thermal time from emergence to end of juvenile stage: P1 = 140 and thermal time from 75% silking to physiological maturity: P5 = 750) and one long-season cultivar (P1 = 220 and P5 = 910) for the rest of the country. Some of the irrigated areas of the Northwest, like the Douro Valley, are >700 m in altitude. In contrast, the southwest areas of the Guadalquivir Valley are <150 m in altitude. The sowing dates were then chosen as early as the end of frost events would allow, which is the current practice. Barley was not simulated for the southern areas because wheat is the crop that is mainly cultivated there (MAPA, 1998).
A modification in DSSAT 3.1 accounts for changes in CO2 concentration through the direct effect on photosynthesis and ET rates. The ET was calculated by Priestley-Taylor (Priestley and Taylor, 1972) because the models had been calibrated for this subroutine. A modification of the 
coefficient was included to account for maximum air temperatures >35°C for maize, as described by Mínguez and Iglesias (1996). Simulations under the 2xCO2 scenario were performed with a CO2 concentration of 550 ppmv. The physiological effect of CO2 on net photosynthesis rates is simulated by a multiplicative coefficient that increases the daily potential dry matter production at optimum temperature and soil water availability, based on a set of observational studies made by Cure and Acock (1986) and Kimball (1983). For the considered CO2 concentration (550 ppmv), the coefficients are 1.06 for maize and 1.17 for wheat and barley. The model also simulates the effect of CO2 on stomatal resistance by means of a ratio applied to the calculation of transpiration rates that accounts for stomatal closure under higher CO2 concentrations (Hoogenboom et al., 1995). The model assumes that increases in stomatal resistance under elevated CO2 levels are independent of water, nutrient, and temperature stresses. This response is similar to the canopy transpiration rate calculated by Allen (1990) that was derived from experimental data obtained under no limiting conditions.
AEGIS/WIN Geographic Information System
AEGIS/WIN (Agricultural and Environmental Geographic Information System for Windows) (Engel and Jones, 1995; Engel et al., 1997) was used to combine the crop model system, DSSAT 3.1, to the geographic information system (GIS) software, ArcView 3.0.a (ESRI, 1994). The new version of DSSAT, 3.5, now includes this tool (Hoogenboom et al., 1999; Hartkamp et al., 1999).
A digital map of the Spanish borderline without the islands was used as a background and managed with ARC/INFO 7.1 (ESRI, 1993). Another layer with the regional model grid (50 by 50 km2) was also created. Both layers or coverages, i.e., digital maps linked with data tables, were projected in the universal transverse mercator (UTM) system so that they could be overlapped, and thereby obtain cells corresponding to the Spanish territory. The methodology applied for the nine cells from the GCM was the same, but only the one applied to RCM-DSSAT will be described.
Eighty-six RCM cells or polygons corresponding to irrigated areas were selected to run the simulations. The resulting coverage was managed with ArcView 3.0.a because ARCINFO is not compatible with AEGIS/WIN. To accomplish the link, the input of soil and weather codes was needed in the coverage attribute table that was attached to the map. These codes referred to the soil and weather files of DSSAT.
The soil file was obtained by overlaying the coverage with the digital European Soil Map at a scale of 1:1000000, based on the classification of the Food and Agriculture Organization (FAO and UNESCO, 1988). The most representative soil type in terms of area and agricultural use was selected in each cell. Therefore, one soil was assigned to each polygon. These soil types were characterized using soil taxonomy classification (USDA, 1998).
Daily meteorological data simulated by the PROMES model for the 1xCO2 and 2xCO2 scenarios were processed and adapted to the needs of the weather files managed by DSSAT. Files were built with daily maximum and minimum temperatures, precipitation, solar radiation, and wind and air humidity covering the 10-yr period from 2040 to 2049.
The coverage containing the identifying codes could then be used by any AEGIS/WIN project to run simulations from an ArcView environment. The first set of simulations was run for 10 yr of baseline climate (1xCO2) and 10 yr of modified climate (2xCO2), both with current crop management practices. For each scenario and crop, DSSAT experimental files were created that specified the crop and varieties as well as the sowing date, density, and depth. One polygon was considered the smallest unit of analysis.
After reading and joining the simulation outputs to the polygon attribute table, results were displayed in thematic digital maps, which made spatial analysis possible. A simulation of different climate scenarios was performed by changing the weather files in DSSAT. Results from the 1xCO2 and 2xCO2 scenarios were also combined so that relative changes in the yield and development components were displayed in maps. An analysis of variance (ANOVA) was done with STATGRAPHICS Plus 2.1 and CVs were obtained using a spreadsheet and the statistical function of AEGIS/WIN. Geospatial weighted statistics were not calculated because of the formal, more than physical, nature of the polygons (equal area cells).
|
Results and discussion |
|---|
|
TOPABSTRACTINTRODUCTIONMethodologyResults and discussionConclusionsREFERENCES |
|---|
|
|
|
|
|
|
|
|
The grain yield and biomass of rainfed barley (Table 2) were overestimated in both models (Díaz-Ambrona, 1999). Both the GCM and RCM-DSSAT underestimated the measured maximum grain yields for wheat in Basin 1 (12000–13000 kg ha-1) (MAPA, 1998) but not in Basin 2 where the yields were equivalent (9500 kg ha-1) (Mariscal, 1993; MAPA, 1998). The maximum grain yield and biomass obtained in field experiments of maize in Basin 2 (17000 kg ha-1) (Aguilar, 1990) was also closer to the simulated values in RCM-DSSAT (Table 4).
Several differences between the GCM and RCM-DSSAT in the 1xCO2 scenario could be highlighted. The development rates during the vegetative phase were greater in RCM-DSSAT simulations than in GCM-DSSAT and were a consequence of the higher air temperatures of the RCM scenario during this period (Fig. 4). Nevertheless, this did not take place in the summer crop, maize (Table 4), and no significant differences were found in the crop duration. The lower air temperatures in the 1xCO2 GCM scenario (similar evolution as in Fig. 4 and Table 1 for annual temperatures) did not lead to lower development rates in the simulated maize. The differences between the GCM and RCM-DSSAT were attenuated in the summer months.
The grain yield of wheat in Basin 1 was significantly greater from GCM-DSSAT simulations, but the differences in amount and significance were less than in biomass. In fact, the sowing date for irrigated wheat in Basin 1 was the beginning of November, so the grain filling period lasted from mid-April to the end of May in RCM-DSSAT and from the beginning of June until mid-July in GCM-DSSAT (Table 3). The same happened in Basin 2 where the sowing date of wheat was at the beginning of December, and the grain filling took place from mid-May to the 3rd wk of June in RCM-DSSAT and from the beginning of July until mid-August in GCM-DSSAT (Table 5). The general anticipation of the crop cycle brought the grain filling period to periods where the incident radiation was smaller and the days were shorter, which limited the daily net photosynthesis in that phase.
Effect of the 2xCO2 Climate
The development rates during the vegetative phase were greater in the 2xCO2 scenarios from both climate models, especially for winter crops. This led to a general anticipation of anthesis and physiological maturity, but the duration of the grain filling period was not as affected (Tables 2 and 3). Because air temperatures in the GCM-derived climate were lower in the 1xCO2 scenario, the effect of climate change on the shortening of crop duration was relatively higher than in the RCM-derived scenarios (Tables 2 to 5). As shown in Fig. 4, the minimum air temperatures in winter were near 0°C and changed to 5°C in the GCM scenarios.
In Basin 2 and in all RCM cells corresponding to GCM Cell 2-2, increases in air temperature under the 2xCO2 scenario prevented wheat from satisfying its vernalization requirements. For the 10-yr simulations, the wheat crop duration was >365 d, and the simulated biomass was >45000 kg ha-1. In the RCM cells corresponding to GCM Cell 3-2, the same happened for several years. As a result, all wheat crop outputs were unrealistic and were excluded from Table 5. This showed the higher sensitivity of RCM-DSSAT to detect failures due to the lack of vernalization in wheat cultivars with cold requirements. Under the same conditions, the shoot biomass based on GCM-DSSAT reached a value of approximately 23000 kg ha-1 and a shorter vegetative phase than in the 1xCO2 scenario (Table 5).
No statistically significant differences were found in the biomass and grain yield between the 1xCO2 and 2xCO2 scenarios within the same climate model.
The ET and irrigation requirements generally decreased for all crops and both climate models. The reductions were closely related to the shortening of crop duration, so they were greater for the RCM-DSSAT simulations than for the GCM-DSSAT simulations (Tables 2–5).
Both the GCM and RCM-based simulations predicted an increased water use efficiency for aboveground production (WUEb) under 2xCO2 conditions that reached values of approximately 40 kg ha-1 mm-1 (Table 6). Both GCM-DSSAT and RCM-DSSAT predicted more pronounced WUEb increases for wheat and barley (Table 6), as expected from the modeling of the beneficial effects of higher CO2 concentrations on C3 metabolism.
|
Physiological maturity of barley was reached >35 d earlier in the RCM-based simulations for both climate scenarios (Table 2) and 40 d earlier in wheat in the 1xCO2 scenario (Tables 3 and 5).
In Basin 2, maize was sown once the frosts ended, which always took place in mid-March for the RCM-derived climate and in mid-May for the GCM-derived climate. The delay in sowing dates in the GCM scenarios made the vegetative phase occur in a warmer period, so the days to anthesis were significantly fewer compared with the RCM-DSSAT simulation for both scenarios. The effect of warmer air temperatures from the RCM on thermal time accumulation was reduced because it was partially masked by the threshold temperature 
together with an early sowing date (Table 4).
For all crops and climate scenarios, the aboveground biomass of the RCM-DSSAT simulations was always significantly greater than that of the GCM-DSSAT simulations in spite of the shorter crop cycles. In the 2xCO2 scenario, milder RCM temperatures during the winter and early spring accelerated crop establishment and reduced the low-temperature limitation of photosynthesis for the C3 species in both basins, thereby allowing for a higher shoot biomass to accumulate in a shorter time. The positive effect of the warmer spring air temperatures in the C3 species was compensated by the shorter crop duration.
Differences in the crop duration of maize between the GCM-DSSAT and RCM-DSSAT simulations were much less than in C3 species, so the effects of warmer RCM temperatures in the photosynthetic rates and earlier sowing dates led to increases in biomass production. These increases were >3000 kg ha-1 in the RCM-DSSAT simulations under the 1xCO2 scenario and >4500 kg ha-1 under the 2xCO2 scenario (Table 4).
Due to the greater biomass production of RCM-DSSAT compared with GCM-DSSAT, simulations of C3 species did not result in significant increases in barley grain yield (Table 2) or wheat grain yield under the 2xCO2 conditions of Basin 1 (Table 3) and the 1xCO2 conditions of Basin 2 (Table 5). The lower grain yield response of the RCM-DSSAT simulations of C3 species is due to a general decrease in the harvest index when compared with the GCM-DSSAT simulations.
The barley and wheat grain filling duration was almost the same for the RCM and GCM-DSSAT simulations in all conditions. The lower dry matter distribution to the grain in RCM-DSSAT may have been due to the general anticipation of the crop cycle that brought the grain filling to periods where the incident radiation was smaller and the days were shorter, which limited the daily net photosynthesis in that phase.
In maize, a higher harvest index was reached in RCM-DSSAT for both CO2 scenarios (Table 4), so the differences with GCM-DSSAT of 4000 kg ha-1 under the 1xCO2 scenario and >5000 kg ha-1 under the 2xCO2 scenario were even greater in the harvest index than in the biomass. The differences in yield could have come from the direct effects of warmer RCM temperatures on the photosynthesis rates during the vegetative and grain filling phases because the differences in crop duration were small and much lower than in C3 species. The same considerations could have been made for Basin 1 (Fig. 6).
The ET from the RCM-DSSAT simulations was higher than the corresponding values from GCM-DSSAT in all crops and climate scenarios (Tables 2– 5). Considering the shorter crop cycles of wheat and barley in RCM-DSSAT, the daily ET increased even more. Once again, the warmer RCM temperatures in each basin and the greater biomass of the crops may have been responsible for the high ET. This also explained the similar values of the WUEb, between 32 and 38 kg ha-1 mm-1, that were found under the 1xCO2 conditions of both the GCM-DSSAT and RCM-DSSAT simulations (Table 6).
The mean ET increases in the RCM-DSSAT simulations of C3 species were 15.2 and 9.5% in the 1xCO2 and 2xCO2 scenarios, respectively (Tables 2, 3, and 5), while they were 25.9 and 14.5% for maize because no reduction occurred in the maize crop duration (Table 6).
The increase in the daily ET from RCM-DSSAT was compensated in C3 species by a shorter crop duration. The irrigation requirements were then similar to GCM-DSSAT for wheat in Basin 1 under both climate scenarios (Table 3) and under the 1xCO2 scenario in Basin 2 (Table 5). They were even significantly lower for maize in both scenarios (Table 4). Earlier sowing dates were possible under the RCM scenarios, and the crop cycle was matched to periods with lower ET demands and higher precipitation. Both climate models simulate more precipitation than is measured, so the irrigation requirements should be carefully considered.
The temporal variability was generally greater in the RCM-DSSAT simulations than in the GCM-DSSAT simulations. This was more pronounced in the phenological characteristics and irrigation requirements because of their direct dependence on air temperature and rain. The temporal variability is not so marked in the growth outputs because differences in the crop duration are compensated by the opposite effects on radiation interception and efficiency.
In the RCM-DSSAT simulations, the CVspat, and as stated above, CVtemp related to irrigation needs were high for winter wheat, thereby enhancing the importance of high-resolution studies in areas such as the Iberian Peninsula (Tables 3 and 5). The interannual variability in irrigation requirements was also found to be high, especially in the future climate scenario. The CVspat and CVtemp of the irrigation requirements for maize were lower than for wheat, but with the same general trend, greater for the RCM-DSSAT simulations. The coefficients were without a clear difference between climate scenarios (Table 4). The CVspat and CVtemp were also found to increase for other crop outputs in 2xCO2 scenarios, and in general, for the winter crops with respect to the 1xCO2 scenario.
|
Conclusions |
|---|
|
TOPABSTRACTINTRODUCTIONMethodologyResults and discussionConclusionsREFERENCES |
|---|
The RCM-derived 1xCO2 and 2xCO2 scenarios had warmer air temperatures than those of the GCM, leading to faster development rates, and in general, a higher aboveground biomass and ET. The ET rates were always significantly higher in the RCM-derived scenarios for both irrigated crops, but the irrigation requirements were not different in irrigated wheat and significantly lower in maize. This was due to the increase in the interannual variability under the 2xCO2 scenario as well as changes in the RCM seasonal rain.
The minimum air temperatures in the 1xCO2 GCM scenario, especially during winter, were near to or lower than the base or threshold temperatures established by the crop models. Thus, the crop differences between the 1xCO2 simulations of the RCM and GCM were more significant than those between the 2xCO2 and 1xCO2 simulations.
The results obtained with RCM-DSSAT highlighted differences among closely located areas that could not be detected with GCM-DSSAT, thanks to the higher spatial resolution of the RCM. The GCM scenarios consider a single mean altitude for each 250- by 250-km2 cell, which masks the differences in altitude, and consequently, may project inaccurate temperature data. As a result, the ability to simulate spatial differences in crop phenology, grain yield, and water requirements is limited. The comparison between the GCM and RCM-based analysis showed a higher sensitivity of the RCM-DSSAT predictions; for instance, RCM-DSSAT detected crop failures that were not detected by GCM-DSSAT simulations, which were due to a lack of vernalization in the wheat cultivars with these requirements. Nevertheless, the interannual variability in GCM-DSSAT was underestimated. The increase in temperatures during winter could affect crops with cold requirements, as shown here with wheat. Rainfed winter grown crops will be subject to increased precipitation variability.
The meteorological data generated by RCMs is better matched with the resolution that can be achieved in soil and crop parameters. The use of direct output of GCMs is questionable given the orography, land cover distribution, and size of the Iberian Peninsula. The methodology presented here proved to be adequate for impact and vulnerability assessments using RCM-DSSAT in areas with complex orographical features. The GIS and AEGIS/WIN were key tools for processing the information and performing spatial analysis by means of combining the GIS with DSSAT, the crop simulation model-based decision support system.
|
ACKNOWLEDGMENTS |
|---|
Received for publication October 29, 1999.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMethodologyResults and discussionConclusionsREFERENCES |
|---|
ova (ed.) Proc. Eur. Soc. of Agron. Congr., 5th, Nitra, Slovak Republic. 27 June–3 July 1998. Inst. Natl. de la Recherche Agron., Grignon, France.This article has been cited by other articles:
|
|
 |
|
  F. Sau, K. J. Boote, W. M. Bostick, J. W. Jones, and M. I. Minguez Testing and Improving Evapotranspiration and Soil Water Balance of the DSSAT Crop Models Agron. J., September 1, 2004; 96(5): 1243 - 1257. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. J. Carbone, L. O. Mearns, T. Mavromatis, E. J. Sadler, and D. Stooksbury Evaluating CROPGRO-Soybean Performance for Use in Climate Impact Studies Agron. J., May 1, 2003; 95(3): 537 - 544. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||
