Agricultural and Forest Meteorology

Transcription

1 Agricultural and Forest Meteorology 151 (2011) Contents lists available at ScienceDirect Agricultural and Forest Meteorology jou rn al h om epa g e: Trend and uncertainty analysis of simulated climate change impacts with multiple GCMs and emission scenarios X.-C. Zhang a,, W.-Z. Liu b, Z. Li c, J. Chen d a USDA-ARS, Grazinglands Research Laboratory, 7207W. Cheyenne St., El Reno, OK 73036, USA b State Key Laboratory of Soil Erosion and Dryland Farming on Loess Plateau, Institute of Soil and Water Conservation, Chinese Academy of Sciences, Northwest A & F University, Yangling, Shaanxi , China c College of Resource and Environment, Northwest A&F University, Yangling, Shaanxi , China d Department of Construction Engineering, École de technologie supérieure, Université du Québec, 1100 Notre-Dame Street West, Montreal, QC H3 C 1K3, Canada a r t i c l e i n f o Article history: Received 21 February 2011 Received in revised form 27 April 2011 Accepted 15 May 2011 Keywords: Climate change Impact assessment Soil erosion Soil hydrology Wheat production a b s t r a c t Trends and uncertainty of the climate change impacts on hydrology, soil erosion, and wheat production during at El Reno in central Oklahoma, USA, were evaluated for 12 climate change scenarios projected by four GCMs (CCSR/NIES, CGCM2, CSIRO-Mk2, and HadCM3) under three emissions scenarios (A2, B2, and GGa). Compared with the present climate, overall t-tests (n = 12) show that it is almost certain that mean precipitation will decline by some 6% (>98.5% probability), daily precipitation variance increase by 12% (>99%), and maximum and minimum temperature increase by 1.46 and 1.26 C (>99%), respectively. Compared with the present climate under the same tillage systems, it is very likely (>90%) that evapotranpiration and long-term soil water storage will decease, but runoff and soil loss will increase despite the projected declines in precipitation. There will be no significant changes in wheat grain yield. Paired t-tests show that daily precipitation variance projected under GGa is greater than those under A2 and B2 (P = 0.1), resulting in greater runoff and soil loss under GGa (P = 0.1). HadCM3 projected greater mean annual precipitation than CGCM2 and CSIRO (P = 0.1). Consequently, greater runoff, grain yield, transpiration, soil evaporation, and soil water storage were simulated for HadCM3 (P = 0.1). The inconsistency among GCMs and differential impact responses between emission scenarios underscore the necessity of using multi-gcms and multi-emission scenarios for impact assessments. Overall results show that notill and conservation tillage systems will need to be adopted for better soil and water conservation and environmental protection in the region during the next several decades. Published by Elsevier B.V. 1. Introduction The Global Climate Projections in the Fourth Assessment Report of Climate Change 2007: The Physical Science Basis prepared by the Intergovernmental Panel on Climate Change (IPCC, 2007) has concluded that global mean atmospheric temperature will likely increase between 1.8 and 4.0 C by the end of this century, depending on the greenhouse gas (GHG) emissions scenarios. As a result of temperature rise, global hydrological cycles will intensify, and globally averaged mean water vapor, evaporation, and precipitation will likely increase in the century. Many General Circulation Models (GCMs) have consistently projected considerable increases in (1) the frequency and magnitude of extreme events including both droughts and heavy downpours during the century, (2) average daily rainfall intensity, especially for extreme rainfall events, and Corresponding author. Tel.: x264; fax: address: John.Zhang@ars.usda.gov (X.-C. Zhang). (3) spatial and temporal variability of precipitation (IPCC, 2007). The projected increases in the frequency and magnitude of extreme precipitation events are in line with the observed trends in many parts of the world. For example, across the contiguous United States, precipitation has increased by some 10% since 1910, and the increase has been primarily in the form of heavy and extreme daily storms (Karl and Knight, 1998). Specifically, about 53% of the total national increase is due to the precipitation increase within the upper 10% of all the daily precipitation amounts since Groisman et al. (2001) analyzed the trends in share of total annual precipitation occurring in heavy (>95th percentile), very heavy (>99th), and extreme daily precipitation events (>99.9th) in the contiguous U.S. between and , and reported that the linear trends between the two periods increased by 4.6, 7.2, and 14.1% per decade, respectively, which are in contrast to the 1.2% increase per decade for total annual precipitation. This projected increase towards more intense rainfall events is of great concern for assessing the potential impacts on surface hydrology, soil erosion, crop production, and environmental protection because severe soil erosion and catastrophic environ /$ see front matter. Published by Elsevier B.V. doi: /j.agrformet

2 1298 X.-C. Zhang et al. / Agricultural and Forest Meteorology 151 (2011) mental destruction are often caused by infrequent heavy storms. A special report prepared by the Soil and Water Conservation Society in 2003 (SWCS, 2003) stated that under climate changes, the potential for such projected changes to increase the risk of soil erosion and related environmental consequences is clear, but the potential damage is not known and needs to be assessed. These insights are needed to determine (1) whether a change in soil and water conservation practices is warranted under changed climate and (2) what practices should be taken to adequately protect soil and water resources if a change is warranted. Therefore, proper assessment of potential impacts of climate change is critical to developing strategic plans and adaptations to mitigate the adverse effects and protect ecosystems and environments. Mismatches in time and space between GCM resolution and input requirements of hydrological and crop models are major challenges for impact assessment. Both dynamic and statistical downscaling methods have been used to couple GCM output and simulation models. Using various statistical downscaling methods, many researchers have coupled GCM projections with hydrological and agricultural systems models and evaluated the potential impacts of projected climate changes on crop productivity (e.g., Rosenzweig and Parry, 1994; Semenov and Porter, 1995; Mearns et al., 1997; Mavromatis and Jones, 1998), and on surface runoff and soil erosion (e.g., Favis-Mortlock et al., 1991; Boardman and Favis-Mortlock, 1993; Favis-Mortlock and Savabi, 1996; Pruski and Nearing, 2002a,b; Zhang et al., 2009). General results indicated that surface runoff and soil erosion would very likely increase as rainfall intensity increases under climate change. The rates of the increases are directly linked to the frequency and intensity of extreme precipitation events (Pruski and Nearing, 2002a; Zhang et al., 2009). In recent years, the Water Erosion Prediction Project (WEPP) model has been used to simulate the potential impacts of climate change on soil erosion and surface runoff in U.S. (e.g., Pruski and Nearing, 2002a,b; O Neal et al., 2005; Zhang and Nearing, 2005), Belgium (Nearing et al., 2005), U.K. (Favis-Mortlock and Savabi, 1996; Mullan and Favis-Mortlock, 2010), Austria (Klik and Eitzinger, 2010), South Korea (Kim et al., 2009), and China (Zhang et al., 2009; Li et al., 2010). Proper spatial and temporal treatments of climate variation and change during downscaling are crucial for yielding reliable impact assessment. Zhang (2005) developed a statistical downscaling method that treats spatial and temporal climate variation explicitly by using transfer functions for spatial downscaling and a stochastic weather generator for temporal downscaling. Compared with the conventional downscaling approaches, this method tends to produce greater surface runoff and soil erosion due to more detailed treatment of spatiotemporal climate variation (Zhang, 2007). It has been used to study the impact of climate change at particular locations on rainfall erosivity (Zhang et al., 2010) and soil hydrology and soil erosion (Zhang et al., 2009; Li et al., 2010). The objective of this work is to quantify trends and uncertainties of climate change and its impacts on runoff, soil water, soil erosion, and wheat production at a unit watershed near El Reno, Oklahoma, USA, during The climate change scenarios, projected by four GCMs under three GHG emission scenarios, were downscaled to the target location using an explicit spatiotemporal downscaling method of Zhang (2005). 2. Materials and methods 2.1. Watershed description Three experimental watersheds, located at the Grazinglands Research Laboratory, 7 km west of El Reno, Oklahoma were used in the WEPP model calibration and for future climatic impact simulation. Each watershed is 80 m wide and 200 m long with a drainage area of 1.6 ha. The longitudinal slope of the watershed is approximately 3 to 5%. Soils are primarily silt loam with an average of 23% sand and 56% silt in the tillage layer. The climate at the location is characterized as semiarid to subhumid with large seasonal and interannual precipitation variability. The mean monthly precipitation is bimodal with the primary peak in May June and the secondary peak in August October. A common regional cropping system (annual winter wheat summer fallow) with three contrasting tillage systems of no-till, conservation (disks) and conventional (moldboard) tillage systems was primarily studied on each watershed between 1980 and Measured weather data, soil properties, wheat yield, surface runoff and soil loss were used to calibrate the plant growth, effective hydraulic conductivity, and soil erosion components of the WEPP model in a sequential and iterative manner (Zhang, 2004; Zhang and Nearing, 2005) WEPP model and calibration The WEPP model (version ), which is a process-based, continuous daily simulation model (Flanagan and Nearing, 1995), simulates soil erosion, hydrological processes, daily water balance, plant growth, and residue decomposition components. For simulating the impacts of climate change, the plant growth and water balance components in the WEPP model were modified to account for the CO 2 effects on evapotranspiration (ET) and biomass production as is used in the APEX models (Williams et al., 2008). The WEPP model simulates eventual, monthly, and annual soil loss and runoff from a hillslope or small watershed. It uses four input files including soil, topography, climate, and crop management. It also includes a climate generator (CLIGEN), which generates daily rainfall pattern, daily temperature (maximum, minimum, and dew point), solar radiation, and wind speed and direction (Nicks and Gander, 1994). Precipitation occurrence is generated using a first-order, two-state Markov chain based on conditional transition probabilities of a wet day following a wet day (P w/w ) and a wet day following a dry day (P w/d ). Daily precipitation amounts are generated using a skewed normal distribution, and daily temperature and radiation are generated using normal distributions. CLIGEN was evaluated for the study region at four weather stations across Oklahoma (Zhang and Garbrecht, 2003). The mean absolute relative errors were 4.7 and 1.7% for simulating means of daily and monthly precipitation amounts, respectively, and 3.7 and 6.7% for simulating standard deviations. The key WEPP parameters calibrated here included effective hydraulic conductivity for runoff prediction, harvest index and energy-biomass conversion ratio for biomass and grain yields, and interrill and rill erodibilities for soil loss. Remaining parameters retained the model s default values. Since the physiographic conditions of the three watersheds were similar, only one set of these parameters was calibrated, which were 7.95 mm/h for effective saturated hydraulic conductivity, 0.26 for harvest index, 23 g/kj for energy biomass ratio, kg s/m 4 for interrill erodibility, s/m for rill erodibility, and 6.3 Pa for critical shear stress. These calibrated soil erodibility parameter values are well within the ranges recommended based on soil properties by the model. More information on the parameter calibration can be found in Zhang (2004) and Zhang and Nearing (2005) GCMs and emissions scenario Four GCMs (CCSR/NIES, CGCM2, CSIRO-Mk2, and HadCM3) and three emissions scenarios (A2, B2 and GGa) from the Third Assessment Report (IPCC, 2001) were used, and the respective spatial resolutions of the four GCMs were (long.) (lat.), , , and The three emissions

3 X.-C. Zhang et al. / Agricultural and Forest Meteorology 151 (2011) Variables considered: Precipitation, Tmax, Tmin Parameters modified: Mean, variance, probability of rainfall occurence WEPP input data: Soil, tillage, crop, topography Smoothing over 4 neighboring boxes using inverse distance weighting Spatial downscaling of monthly projections from GCM cell to a target station using transfer functions Temporal downscaling from monthly to daily values at the station using a weather generator Impact assessment using erosion models such as WEPP 2.5. Simulated management systems under changed climate All measured soil properties and watershed configurations used in WEPP calibration were used in the future impact simulation to minimize model prediction errors. A common regional cropping system of continuous annual winter wheat-summer fallow was simulated. No-till, conservation tillage (3 disks in summer at a 45- day interval), and conventional tillage (one moldboard plow plus 3 disks at a 30-day interval) were simulated to evaluate the effect of surface residue management and tillage operations on runoff and soil loss. The first tillage was made 2 weeks following wheat harvest. Wheat was planted on 15 October and harvested on 20 June under the baseline climate, while it was planted on 20 October and harvested on 15 June to accommodate increased temperature. The WEPP model was run for 100 years for each climate change scenario of and for the baseline climate as well. Fig. 1. Logical flow chart of the downscaling and impact assessment procedures. scenarios were reported in the Special Report on Emissions Scenarios (i.e., SRES 2000) by IPCC, and atmospheric CO 2 concentration was assumed to be 444 ppmv (parts per million by volume) for GGa, 416 for B2a, and 592 for A2a (in the high end of A2 family) by Monthly mean precipitation and maximum and minimum temperature projected by the four GCMs under the three emissions scenarios were downloaded from the IPCC Data Distribution Center ( for the U.S. southern Great Plains for the periods of and Statistical spatiotemporal downscaling The logical flow chart used in this work is shown in Fig. 1. First, monthly GCM projections at the four nearest neighboring gridpoints were interpolated or smoothed to the target station using an inverse distance weighting method as was used by Li et al. (2010). Second, the smoothed monthly GCM projections were spatially downscaled to the target station using linear or nonlinear transfer functions derived from quartile plots (i.e., QQ-plot). Transfer functions were fitted between interpolated monthly GCM projections and station monthly values for the baseline period of One function was fitted for each variable and each calendar month, and thus 36 functions (3 variables 12 months) were derived for each climate change scenario. Those functions were used to downscale GCM projections at grid box to the target station for the future period of Third, at the target station spatially downscaled monthly values of future climate were further disaggregated to daily values using CLIGEN (v5.3). For precipitation, baseline conditional probability of precipitation occurrence, daily precipitation mean and variance (computed from daily station weather data of the baseline) were adjusted using spatially downscaled monthly precipitation means and variances from step 2. For maximum and minimum temperature, spatially downscaled mean maximum and minimum temperature were directly used as the adjusted means of the changed climate. Adjusted daily temperature variances were estimated by multiplying the baseline daily temperature variances by the monthly variance ratios of spatially downscaled monthly values of to those of These parameter adjustments were made for each calendar month, and were repeated for each climate change scenario. All adjusted parameter values were input to CLIGEN to generate 100 years of daily weather series for each climate change scenario for the target station. Readers are referred to Zhang (2005) for detailed description of the downscaling procedures Statistical analysis One hundred years of CLIGEN-generated daily weather series were used to compute annual mean climate changes for each GCM and emission scenarios as well as for the baseline climate. All annual means of all GCMs and emission scenarios (n = 12) were tested against the known means of the baseline climate for significant changes using a one tail t-test. Furthermore, annual mean climate changes were grouped either by emission scenario to test for significant differences between scenarios or by GCMs to test for differences between GCMs. A one tail t-test was paired by GCMs (n = 4) in the case of the former or by emission scenario (n = 3) in the case of the latter. Similarly, WEPP-simulated annual means such as runoff and soil loss were tested against the known baseline means for each tillage system (n = 12). For the scenario contrasts, t-tests were paired by the GCMs and tillage systems (n = 12), while for the GCM contrasts they were paired by the scenarios and tillage systems (n = 9). It is worth noting that the multiple pairwise comparisons of two population means were run for all possible contrasts. This procedure is simple and straightforward, but does have the disadvantage of increasing the probability of committing Type 1 error (i.e., falsely rejecting at least one of the null hypotheses) as the number of t- tests increases. However, we only had 3 pairwise comparisons for scenario contrasts and 6 for GCM contrasts, and the small numbers of the multiple t-tests are not expected to alter the test results and overall conclusions. 3. Results 3.1. WEPP calibration Measured annual mean yields of wheat grain during the calibration period were 1466, 1906, and 1916 kg/ha for the no-till, conservation, and conventional tillage, respectively, and the respective WEPP-calibrated yields were 1920, 1750, and 1850 kg/ha. Zhang (2004) reported that the coefficient of determination (r 2 ) between measured and simulated annual wheat grain yields was 0.54 on the study site, and the r 2 between measured and simulated annual runoff depths was 0.56 using calibrated parameter values. Measured annual mean soil losses were 278, 2571, and 6179 kg/ha for the no-till, conservation, and conventional tillage, respectively, and the corresponding WEPP-calibrated soil losses were 436, 2334, and 6300 kg/ha. The good agreement between measured and calibrated soil losses under the three tillage systems (one in each watershed) with one set of the calibrated soil erodibility parameters indicated that the WEPP model simulated the effects

4 1300 X.-C. Zhang et al. / Agricultural and Forest Meteorology 151 (2011) Table 1 GCMs-projected mean annual precipitation, changes in maximum (T max) and minimum (T min ) temperature, and daily variance (var.) ratios during relative to the baseline at El Reno, Oklahoma after downscaling. Emission scenarios GCMs Annual precipitation (mm) Precipitation var. ratio T max shift ( C) T max var. ratio T min shift ( C) T min var. ratio A2 (592 ppmv) CCSR CGCM CSIRO HadCM B2 (416 ppmv) CCSR CGCM CSIRO HadCM GGa (444 ppmv) CCSR CGCM CSIRO HadCM Mean Baseline ( ) P-value a < < a Pooled data of emission scenarios and GCMs (n = 12) were tested against the known baseline means using a one tail t-test. of tillage systems and surface residue management on soil erosion reasonably well Downscaled climate changes during Average annual precipitation and temperature anomalies along with their variance ratios of the downscaled future climates were shown in Table 1 for all climate change scenarios. The baseline means were also included for comparison. The means of the 12 climate change scenarios were tested against the known baseline means using a one tail t-test (n = 12). Compared with the baseline (or present) climate, the average annual precipitation during would decline by some 50 mm or 6.2% (P = 0.015), while the mean daily precipitation variance would increase by 12% (P = 0.004). The mean maximum and minimum temperature would increase by 1.46 and 1.26 C, respectively (P < 0.001). However, no significant changes in temperature variance were detected. The downscaled climate changes of Table 1 were grouped by emission scenarios to test for significant differences between the scenarios using a paired t-test (n = 4). There were no statistical differences for the mean precipitation amounts among the scenarios; however, the precipitation variance in GGa was greater than those of A2 and B2 at P = 0.1 (Table 2). The mean increase in the maximum temperature was greater in GGa than in A2, but was indifferent from B2. The mean increases in the minimum temperature were greater in B2 and GGa than in A2. Similar paired t-tests for GCM contrasts, paired by scenarios (n = 3), showed that HadCM3 predicted the greatest mean annual precipitation, which was statistically different from those of CGCM2 and CSIRO but indifferent from CCSR at P = 0.1 (Table 3). CSIRO predicted the maximum precipitation variance, which was significantly greater than those of CCSR and Table 2 GCM-averaged annual precipitation, changes in maximum (T max) and minimum (T min ) temperature, and daily variance ratios during relative to the baseline at El Reno, Oklahoma after downscaling. Emission scenarios Annual precipitation (mm) Precipitation variance ratio T max shift ( C) T min shift ( C) A2 735a a 1.050a 1.44a 1.10a B2 710a 1.079a 1.69ab 1.31b GGa 783a 1.260b 1.61b 1.68b a Numbers followed by different letters within each column are statistically different at P = 0.1. One tail t-tests between emission scenarios were paired by GCM models (n = 4). CGCM2 but not different from HadCM3. CCSR and CGCM2 tended to predict greater increases in both maximum and minimum temperature than CSIRO and HadCM Climatic impact assessment during WEPP s mean annual output simulated under all GCMs and emission scenarios was tested against the known means simulated under the baseline climate individually for each tillage system using a t-test (Table 4). Compared with the baseline, under the identical tillage system, the mean annual surface runoff and soil loss during would increase, while the mean annual plant transpiration, soil evaporation, and soil water storage in the 1.2 m profile would decrease. There was no significant trend in the wheat grain yields. The WEPP output was contrasted by emission scenarios, with the contrast being paired by GCMs and tillage systems (n = 12). The mean annual runoff and soil loss in GGa were greater than those in A2 and B2 (Table 5). The wheat grain yield in A2 was higher than those in B2 and GGa. The plant transpiration was lower in A2 than in GGa, but was indifferent from B2. The soil evaporation was greater in A2 and GGa than in B2. The soil water storage was greater in A2 than in B2 and GGa. The WEPP output was further broken down into tillage systems within each emission scenario, and tested individually against the corresponding tillage systems under the baseline climate of Table 4. In all tillage systems, compared with the baseline, the mean annual runoff and soil loss in GGa would increase significantly during ; the wheat grain yields in B2 decrease; the transpiration in A2 decrease; the soil evaporation in all Table 3 Scenario-averaged annual precipitation, changes in maximum (T max) and minimum (T min ) temperature, and daily variance ratios during relative to the baseline at El Reno, Oklahoma after downscaling. GCMs Annual precipitation (mm) Precipitation variance ratio T max shift ( C) T min shift ( C) CCSR 730 a 1.032a 1.87b 1.67c CGCM2 722a 1.129ab 1.91bc 1.63bc CSIRO 729a 1.226c 1.37a 0.99a HadCM3 789b a 1.17b a Pairs followed by different letters within each column are statistically different at P = 0.1. One tail t-test between GCMs was paired by emission scenarios (n = 3). Note that numbers without letters are not different from the others within each column.

5 X.-C. Zhang et al. / Agricultural and Forest Meteorology 151 (2011) Table 4 Mean annual output simulated under all GCMs and emission scenarios for runoff, soil loss, wheat yield, plant transpiration (PT), soil evaporation (SE), and daily mean water storage in the 1.2 m profile (SW) using 100 years of generated baseline and future climate. Tillage Runoff (mm) Soil loss (kg/ha) Yield (kg/ha) PT (mm) SE (mm) SW (mm) Baseline Conv Cons No-till Conv. 105** 7768* ** 218** 270** Cons. 104* 4325** ** 226** 264** No-till 89** 322** ** 197** 263** * and ** are different from the corresponding tillage treatment of the baseline at P = 0.1 and 0.05, respectively. Pooled data for three emissions and four GCMs (n = 12) for each tillage system were tested against the known baseline means for that treatment using a one tail t-test. Table 5 GCM-averaged annual runoff, soil loss, wheat yield, plant transpiration (PT), soil evaporation (SE), and daily mean water storage in the 1.2 m profile (SW) using 100 years of generated climate for the period of Scenario Runoff (mm) Soil loss (kg/ha) Yield (kg/ha) PT (mm) SE (mm) SW (mm) A2 93a a 2955a 1473b 398a 218b 270b B2 89a 3354a 1214a a 261a GGa 115b 6105b 1262a 432b 218b 265a a Numbers followed by different letters within each column are statistically different at P = 0.1. One tail t-tests between emission scenarios were paired by GCMs and tillage systems (n = 12). Table 6 GCM-averaged annual runoff, soil loss, wheat yield, plant transpiration (PT), soil evaporation (SE), and daily mean water storage in the 1.2 m profile (SW) using 100 years of generated climate for Emissions/Tillage Runoff (mm) Soil loss (kg/ha) Yield (kg/ha) PT (mm) SE (mm) SW (mm) A2 Conv * 403** 225* 274 Cons ** 231* 269 No-till ** 199* 267 B2 Conv ** ** 266** Cons ** ** 259** No-till ** ** 258** GGa Conv. 118** 11416** * 269 Cons. 121** 6494** * 264 No-till 106** 406** * 263 * and ** are different from the corresponding tillage treatment of the baseline at P = 0.1 and 0.05, respectively. Pooled data for four GCMs (n = 4) for each emission scenario and tillage system were tested against the known baseline means in Table 4 for that tillage using one tail t-tests. scenarios decrease; and the soil water storage in B2 decrease (Table 6). The tillage systems exhibited a substantial impact on the simulated mean annual soil loss within each emission scenario, with the losses from the no-till and conservation tillage being consistently less that those of the conventional tillage. The WEPP output was also analyzed and tested by GCMs (n = 9, Table 7). Two important trends deserve to be mentioned. There were no significant differences in the mean annual soil loss among all GCMs. The mean annual surface runoff, wheat yield, transpiration, soil evaporation, and soil water storage of HadCM3 were greater than those of most other GCMs Discussion There is a strong east-west precipitation gradient in the study region (Fig. 2). Precipitation amounts during within the 4 neighboring grid boxes ranged from about mm/yr. A previous study showed that simulated climatic impact at the El Reno station depended upon the number of grid boxes used in developing climate change scenarios (Zhang, 2006). This is especially true if a strong spatial gradient exists and the target station is near the intersection of GCM grid boxes (cf. Fig. 2). Hewitson (2003) suggested the skillful resolution may likely be lower than Table 7 Scenario-averaged annual runoff, soil loss, wheat yield, plant transpiration (PT), soil evaporation (SE), and daily mean water storage in the 1.2 m profile (SW) using 100 years of generated climate for the period of GCMs Runoff (mm) Soil loss (kg/ha) Yield (kg/ha) PT (mm) SE (mm) SW (mm) CCSR 99 a a a 265 CGCM2 94a a 408a 206a 259a CSIRO 102b a 404a 213b 265b HadCM3 102b b 422b 234c 272c a Pairs followed by different letters within each column are statistically different at P = 0.1. One tail t-tests between GCMs were paired by emission scenarios and tillage systems (n = 9). Note that numbers without letters are not different from the others within each column.

6 1302 X.-C. Zhang et al. / Agricultural and Forest Meteorology 151 (2011) Fig. 2. Mean precipitation depths for the period of for U.S. southern Great Plains overlaid with four HadCM3 grid cells. The figure was cropped from the U.S. national maps from National Weather Service. (Numbers are in inches, 1 in. = 25.4 mm). the native resolutions of the GCMs and proposed the use of multiple GCM grid boxes rather than a single box to develop climate change scenarios for assessing the first-order climatic impact. This suggestion was in line with the finding that the grid box overlying the local observations may not provide the best predictive skill for the locations (Wilby and Wigley, 2000; Brinkmann, 2002; Diaz- Nieto and Wilby, 2005). Thus, the inverse distance weighting of the four nearest boxes was to better use the GCMs predictive skill or to capture the true climate change signal by averaging out noises. The true climate change signal was then spatially and temporally downscaled to the target station to develop future climate change scenarios for the station. It is worth noting that the inverse distance weighting is not a true downscaling but rather a spatial smoothing, and should not be used over grid boxes having contrasting geophysical properties such as over sea and land surfaces. Based on the likelihood terminology defined in the Fourth Assessment Report (AR4) of IPCC (i.e., extremely likely for P > 95%, and very likely for P > 90%), it is extremely likely that mean annual precipitation will decrease and precipitation variance increase (Table 1). An increase in daily precipitation variance means an increased probability of occurrence of heavy storms. These trends are consistent with those reported in AR4 (IPCC, 2007) in that (1) precipitation in the mid continent of north America will decrease, (2) average rainfall intensity, especially of heavy storms, will increase including places where precipitation is projected to decline, and (3) the frequency and magnitude of extreme rainfall events in the study region during has increased remarkably. The increase in the frequency and magnitude of heavy storms has significant implications for soil and water conservation and environmental protection. For example, it is well documented worldwide that most soil loss often occurs in only a few heavy storms. It is almost certain that air temperature will increase (Table 1). The downscaled increases at El Reno are 1.46 C for the maximum temperature and 1.26 C for the minimum, which are comparable to the increases of 1.61 C for the maximum and 1.24 C for the minimum computed between and using the smoothed raw GCM projections. The differences resulted from the downscaling procedures used here. Both best fitted linear and nonlinear transfer functions were used in the downscaling. The nonlinear transfer functions were used when future projections were in the ranges within which the transfer functions were fitted; otherwise the linear functions were used. Overall, approximately 10% of data points were outside the ranges, and the use of linear approximation resulted in a slightly lower estimate of the maximum temperature compared with the raw GCM projections. The paired t-tests, which were either paired by GCMs in the case of the scenario contrasts (Table 2) or by scenarios in the case of the GCM contrasts (Table 3), were used to minimize the model s effects on the test results in the former or scenario s influence in the latter. Compared with A2 and B2, GGa forcing produced much greater precipitation variance (Table 2), which has significant impact on soil erosion and runoff generation. The increases in the maximum and minimum temperature in A2 tended to be lower than those in B2 and GGa. The discrepancy might be partially caused by the linear approximation of transfer functions when data points were outside the ranges, implying a potential shortcoming of downscaling temperature using this type of transfer function approach. In addition, differences in radiative forcing of other GHG apart from CO 2 among the three emission scenarios might have played a role. The raw GCM-projected temperature increases (before downscaling) relative to the baseline were 1.68 C for A2 and 1.69 C for B2 for the maximum temperature, and 1.05 C for A2 and 1.31 C for B2 for the minimum temperature. It should also be mentioned that the CO 2 concentration of 592 ppmv estimated for A2 by 2025 is in high end of the A2 family, and an average concentration between 450 and 500 ppmv would be more likely. There were no consistent trends between GCMs regarding the predicted mean precipitation, precipitation variance, and temperature shifts (Table 3). Despite the GCM-projected decreases in precipitation, the WEPP-predicted mean annual surface runoff increased (Table 4), resulting from the projected increases in rainfall variance under changed climate (Table 1). An increase in rainfall variance increases probability of occurrence of extreme rainfall events. Heavier storms often result in greater runoff coefficient, and subsequently greater annual runoff totals. Annual mean soil losses increased as runoff increased. Due to the decrease in precipitation amounts and the increase in runoff, WEPP-simulated plant transpiration, soil evaporation, and long-term soil water storage decreased significantly. It should be mentioned that the lesser soil evaporation in no-till was counteracted by the increased evaporation from surface residue (data not shown), resulting in similar soil water balances among the three tillage systems. The greater increase in precipitation variance in GGa relative to B2 and A2 (Table 2) as well as the numerically greater precipitation amount (though statistically insignificant) resulted in much greater increases in the WEPP-simulated runoff and soil loss in GGa (Table 5). The simulated percent increases of soil loss in GGa, compared with A2 and B2, were much greater than those of runoff, indicating that simulated soil erosion is extremely sensitive to increases in runoff rate and volume, especially during heavy storms. It is well documented that greater ambient CO 2 concentration enhances plant radiation use efficiency and biomass production while it suppresses transpiration by increasing stomatal resistance, and therefore increases water use efficiency (Stockle et al., 1992). These mechanisms were incorporated in WEPP and were responsible for the simulated results that the wheat grain yield in A2 (592 ppmv) was greater than those in B2 (416) and GGa (444), while transpiration was lower in A2 than in B2. The lesser precipitation amount in B2 (though insignificant) might be responsible for the lower soil evaporation in the scenario. The lesser runoff and transpiration as well as the intermediate rainfall amount in A2 might have resulted in the greatest soil water balance in the scenario. Compared with the baseline climate, the projected greatest increase in precipitation variance as well as the largest precipitation amount (though insignificant) was responsible for the significant

7 X.-C. Zhang et al. / Agricultural and Forest Meteorology 151 (2011) increases in runoff and soil loss for the GGa scenario (Table 6). The significant decreases in grain yield and soil water balance for B2 were largely caused by the lowest projected precipitation (though insignificant) because precipitation is the major limiting factor for dryland farming in the region. The insignificant changes in grain yield for A2 and GGa resulted from the counterbalanced effects of elevated CO 2 (positive), and decreased precipitation and increased temperature (negative). The significant decrease in transpiration for A2 was attributable to the most increase in CO 2. The decreases in soil evaporation for all scenarios were mainly caused by the decreases in precipitation. As for the model comparison in Table 7, the greater precipitation amount projected by HadCM3 was largely responsible for the greater runoff, grain yield, transpiration, soil evaporation, and soil water balance as was predicted by WEPP. Overall, inconsistency in projected climate changes and simulated impacts between GCMs and discrepancy between emission scenarios underscore the necessity of using multi-gcms and multiscenarios in order to take into account uncertainties introduced by GCMs and scenarios for more reliable impact assessment. Besides uncertainties caused by differences in GCMs and emission scenarios representing various assumptions of future socio-economic conditions, downscaling methods and hydrological simulation models are also important sources of uncertainties, which cannot be quantify here due to singularity and should be considered in future studies. 4. Conclusions It is extremely likely that mean annual precipitation will decrease by some 6% (>98.5% probability), daily precipitation variance will increase by 12% (>99%), and maximum and minimum temperature will increase by 1.46 and 1.26 (>99%), respectively, near El Reno during Compared with the present climate, under the identical tillage systems, it is very likely (>90%) that ET and long-term soil water storage will decease, but runoff and soil loss will increase despite the projected decrease in precipitation. There will be no significant changes in wheat grain yield. The greatest daily precipitation variance projected under GGa is responsible for the greatest runoff and soil loss predicted by WEPP for the scenario (P = 0.1). HadCM3 projected greater precipitation than other GCMs, resulting in greater runoff, grain yield, transpiration, soil evaporation, and soil water storage under HadCM3 as simulated by WEPP (P = 0.1). Compared with the conventional tillage systems, no-till and conservation tillage systems consistently reduced soil loss in all 12 climate change scenarios (4 GCMs 3 emission scenarios). Simulated results strongly indicated that adoption or implementation of conservation systems, especially no-till, will be effective in keeping soil erosion low under future climate in the study region. In addition, the inconsistency among GCMs and disparate responses between emission scenarios underscore the necessity of using multi-gcms and multi-emission scenarios for reliable impact assessment. Acknowledgements This work was partially supported by the Outstanding Overseas Chinese Scholars Fund of Chinese Academy of Sciences (no ) and the International Exchange and Cooperation Program of National Natural Science Foundation of China (no ). References Brinkmann, W.A.R., Local versus remote grid points in climate downscaling. Climate Res. 21, Boardman, J., Favis-Mortlock, D.T., Climate change and soil erosion in Britain. Geog. J. 159, Diaz-Nieto, J., Wilby, R.L., A comparison of statistical downscaling and climate change factor methods: implications on low flows in the River Thames, United Kingdom. Clim. Change 69, Favis-Mortlock, D.T., Evans, R., Boardman, J., Harris, T.M., Climate change, winter wheat yield, and soil erosion on the English South Downs. Agri. Syst. 37, Favis-Mortlock, D.T., Savabi, M.R., Shifts in rates and spatial distribution of soil erosion and deposition under climate change. In: Anderson, M.G., Brooks, S.M. (Eds.), Advances in Hillslope Processes. John Wiley, New York, pp Flanagan, D.C., Nearing, M.A. (Eds.), USDA-water erosion prediction project: hillslope profile and watershed model documentation. NSERL Report, No. 10. West Lafayette, Ind.: USDA-ARS Nat. Soil Erosion Research Lab. Groisman, P.Y., Knight, R.W., Karl, T.R., Heavy precipitation and high streamflow in the contiguous United States: trends in the twentieth century. Bull. Am. Meteorol. Soc. 82, Hewitson, B., Developing perturbations for climate change impact assessments. Trans. Am. Geophys. Union, EOS 84, IPCC (Intergovernmental Panel on Climate Change), Working Group I, Climate change 2001: the scientific basis. Contribution of Working Group I to the Third Assessment Report of the IPCC. Cambridge University Press, Cambridge, UK. IPCC (Intergovernmental Panel on Climate Change), Working Group I, Climate change 2007: the physical science basis. In: Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor and H.L. Miller (Eds.), Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 996 pp. Karl, T.R., Knight, R.W., Secular trends of precipitation amount, frequency, and intensity in the United States. Bull. Am. Meteorol. Soc. 79, Kim, M.-K., Flanagan, D.C., Frankenberger, J.R., Meyer, C.R., Impact of precipitation change on runoff and soil erosion in Korea using CLIGEN and WEPP. J. Soil Water Conserv. 64, Klik, A., Eitzinger, J., Impact of climate change on soil erosion and the efficiency of soil conservation practices in Austria. J. Agr. Sci. 148, Li, Z., Liu, W.Z., Zhang, X.C., Zheng, F.L., Assessing the site-specific impacts of climate change on hydrology, soil erosion and crop yields in the Loess Plateau of China. Clim. Change, doi: /s Mavromatis, T., Jones, P.D., Comparison of climate change scenario construction methodologies for impact assessment studies. Agric. For. Meteorol. 91, Mearns, L.O., Rosenzweig, C., Goldberg, R., Mean and variance change in climate scenarios: methods, agricultural applications, and measures of uncertainty. Clim. Change 35, Mullan, D., Favis-Mortlock, D.T., Modeling the impacts of climate change on future rates of soil erosion: a case study from the north of Ireland. Geophysical Research Abstracts 12, EGU Nearing, M.A., Jetten, V., Baffaut, C., Cerdan, O., Couturier, A., Hermandez, M., Le Bissonnais, Y., Nichols, N.H., Nunes, J.P., Renschler, C.S., Souchere, V., van Oost, K., Modeling response of soil erosion and runoff to changes in precipitation and cover. Catena 61, O Neal, M.R., Nearing, M.A., Vining, R.C., Southworth, J., Pfeifer, R.A., Climate change impacts on soil erosion in Midwest United States with changes in crop management. Catena 61, Pruski, F.F., Nearing, M.A., 2002a. Runoff and soil-loss responses to changes in precipitation: a computer simulation study. J. Soil Water Conserv. 57 (1), Pruski, F.F., Nearing, M.A., 2002b. Climate-induced changes in erosion during the 21st century for eight U.S. locations. Water Resour. Res. 38, art. no Rosenzweig, C., Parry, M.L., Potential impact of climate change on world food supply. Nature 367, Nicks, A.D., Gander, G.A., CLIGEN: a weather generator for climate inputs to water resource and other models. In: Proceedings of the 5th International Conference on Computers in Agriculture, American Society of Agricultural Engineers, St. Joseph, Michigan, pp Semenov, M.A., Porter, J.R., Climatic variability and the modeling of crop yields. Agric. For. Meteorol. 73, Stockle, C.O., Williams, J.R., Rosenberg, N.J., Jones, C.A., A method for estimating the direct and climatic effects of rising atmospheric carbon dioxide on growth and yield of crops: part I modification of the EPIC model for climate change analysis. Agric. Syst. 38, SWCS, Conservation implications of climate change: Soil erosion and runoff from cropland. A report from the Soil and Water Conservation Society, Soil and Water Conservation Society, Ankeny, Iowa. ( Wilby, R.L., Wigley, T.M.L., Precipitation predictors for downscaling: observed and general circulation model relationships. Int. J. Climatol. 20, Williams, J.W., Izaurralde, R.C., Steglich, E.M., Agricultural policy/environmental extender (APEX) model: theoretical documentation. BREC Report # , Blacklands Research and Extension Center, Temple, TX. Zhang, X.C., Calibration, refinement, and application of the WEPP model for simulating climatic impact on wheat production. Trans. ASAE 47, Zhang, X.C., Spatial downscaling of global climate model output for sitespecific assessment of crop production and soil erosion. Agric. For. Meteorol. 135,

8 1304 X.-C. Zhang et al. / Agricultural and Forest Meteorology 151 (2011) Zhang, X.C., Spatial sensitivity of predicted soil erosion and runoff to climate change at regional scales. J. Soil Water Conserv. 61, Zhang, X.C., A comparison of explicit and implicit spatial downscaling of GCM output for soil erosion and crop production assessments. Clim. Change 84, Zhang, X.C., Garbrecht, J.D., Evaluation of CLIGEN precipitation parameters and their implication on WEPP runoff and erosion prediction. Trans. ASAE 46, Zhang, X.C., Liu, W.Z., Li, Z., Zheng, F.L., Simulating site-specific impacts of climate change on soil erosion and surface hydrology in southern Loess Plateau of China. Catena 79, Zhang, X.C., Nearing, M.A., Impact of climate change on soil erosion, runoff, and wheat productivity in central Oklahoma. Catena 61, Zhang, Y.G., Nearing, M.A., Zhang, X.C., Xie, Y., Wei, H., Projected rainfall erosivity changes under climate change from multimodel and multiscenario projections in Northeast China. J. Hydrol., doi: /j.jhydrol