SWAT2012/Release 622 was the version of the model used for this study. SWAT is dynamic model that is typically executed on a daily time step, although sub-daily options are also provided. The model is comprised of climate, soil, hydrology, management, nutrient cycling and transport, pesticide fate and transport, and several other components. Release 622 also features enhanced algorithms that account for more accurate representation of important switchgrass and miscanthus growth phenomena related to belowground biomass, plant respiration and nutrient uptake, which were developed by Trybula et al. (2015) and ported to standard SWAT versions starting with SWAT2012/release 615. A watershed is typically delineated into subbasins in SWAT, based on topography, and each subbasin is then divided into multiple hydrological response units (HRUs), which consist of homogeneous soil, land use, topographic and management characteristics (Neitsch et al., 2011; Arnold et al., 2012). At present, HRUs are not spatially identified in applications of standard versions of SWAT although incorporation of expanded spatial detail is being developed (Duku et al., 2015; Arnold et al., 2010). Water and pollutants discharged at the HRU level are input at the respective subbasin outlet and routed through the stream system to the watershed outlet. Neitsch et al. (2011) and Arnold et al. (2012) provided additional details about specific SWAT components, functions and/or input data requirements.

Baseline model testing (Teshager et al., 2015) was performed using 10 weather stations distributed fairly uniformly across the watershed, and streamflow and in-stream pollutant data measured at a gauge located near Van Meter, which drains 95 % of the RRW. The model was calibrated and validated for the RRW for the years 2002 to 2010 for flow, total suspended solids (TSS), nitrate (NO3) and mineral phosphorus (MINP) at daily, monthly and annual timescales (Teshager et al., 2015). Land use/land cover (LULC) from the USDA Cropland Data Layer (CDL; USDA-NASS, 2012) for the years 2002 to 2010 was used to develop crop rotations for calibration/validation of the watershed. According to Teshager et al. (2015), about 14 % of the watershed was planted in continuous corn (CC), 30 % was in 3-year rotations with 1 year of soybean and 2 years of corn (CCS/CSC/SCC), 31 % was in 2-year corn–soybean rotations (CS/SC), 6 % was in 3-year rotations consisting of 2 years of soybean and 1 year of corn (SSC/SCS/CSS), and 10 % was pasture/grass (Fig. 1). The rest of the watershed included developed areas, forest or water bodies. The SWAT model was able to replicate flow, TSS, NO3 and MINP satisfactorily at daily, monthly and annual timescales.

The most common approach in assessing climate change impacts is scenario construction (Öborn et al., 2011). The objective of a specific scenario, and the subject's complexity and time horizon shapes the method chosen for constructing scenarios (Dreborg, 2004). Due to the absence of a direct method for predicting future farming choices, the agricultural scenarios developed in this study were developed based mainly on the need for more corn production for food, livestock feed and biofuel production, and the promising potential of switchgrass (SWG) for bio-energy production (Khanna et al., 2008; Schmer et al., 2008; Secchi et al., 2008; Vadas et al., 2008). Accordingly, five agricultural scenarios were considered for the overall impact analysis (Table 1).

The first scenario considered in this study assumed that future agricultural land use (crop type and rotation) matches historical agricultural land use patterns and is referred to as the baseline (BL) scenario. In addition to crop types and rotations, fertilizer/manure applications, tillage practices and tile drainage were held constant through all three future simulation periods. Hence, the distributions of crop rotations described in the “Model description and setup” section and Fig. 1, as well as with the management practices stated in Table 2, were used for the BL simulations.

The second scenario reflects projections developed by the US Department of Agriculture (USDA) that demand for corn will increase in the future based on an analysis of the world's agricultural sector in general and the US agricultural sector in particular for the next decade (USDA-NASS, 2015). According to this report, US corn acreage is projected to remain high and production to rise gradually taking all uses of corn into account. Thus, this scenario is termed partial-corn (PC) and is simulated by converting selected HRUs into CC, as a function of baseline crop rotation, land use, and topographical conditions, accommodating the projected increase in corn production. All baseline CCS/CSC/SCC rotations were converted to CC, due to the fact that those land parcels were already managed with relatively intense corn production. Next, pasture HRUs with an average slope less than or equal to the current maximum cropland average slope were converted to CC; the slope constraint prevented conversion of extremely high sloped pasture land. About 52 % of the watershed was planted in CC for this scenario, CS/SC and SSC/SCS/CSS rotations percentages remained the same, and about 2 % of the watershed was still under pasture (Fig. 2a). Fertilizer applications to corn for CC cropping systems was 202 kg N ha-1 and 65 kg P ha-1 (as recommended by Duffy, 2013), in combination with conventional tillage, for the HRUs that were changed from other rotations or land uses to the CC rotation. The presence of tile drainage was held constant relative to the baseline scenario.

The third scenario reflects adoption of switchgrass on selected RRW HRUs and is called the partial switchgrass (PS) scenario. The HRUs selected for this scenario were chosen based on baseline land use and topographical conditions. First, all pasture HRUs in the baseline scenario were converted to switchgrass. Moreover, cropland HRUs with an average slope of greater than or equal to the average slope of pasture in the baseline were changed to switchgrass, to maximize environmental benefits of converted cropland. Accordingly, about 41 % of the watershed was converted to switchgrass in this scenario, resulting in decreases of 29, 34, 42 and 69 % in CC, CCS/CSC/SCC, CS/SC and SSC/SCS/CSS relative to the BL scenario. As a result, about 10, 19, 18 and 2 % of the remaining cropland was partitioned between CC, CCS/CSC/SCC, CS/SC and SSC/SCS/CSS, respectively (Fig. 2b). A nitrogen fertilizer application of 90 kg ha-1 was simulated for all converted cropland planted to switchgrass based on recommendations by Duffy (2008), Schumer et al. (2008), and McLaughlin and Kszos (2005). Tillage practices are not part of a perennial switchgrass cropping system and thus no till was the simulated tillage level by default. The PS scenario criteria underscore that the most productive corn-dominated cropland is located in very low slope areas.

The final two scenarios feature extreme conversions of all cropland and pasture land, representing 90 % of the RRW, to either CC (all corn scenario or AC) or switchgrass (all switchgrass or AS). The same respective fertilizer and tillage assumptions described for the PC and PS scenarios were also used for these two scenarios.

The last two scenarios bracket hypothetical extreme future land use changes in the watershed and represent the extent of the possible trade offs in food and fuel production, water quality and water quantity. The two partial scenarios are more realistic and illustrate potential land use changes at a very fine resolution associated with climate change, global market forces, and energy and conservation policies. For example, the PS scenario could be associated with very aggressive climate mitigation and conservation policies, and the effective deployment of cellulosic ethanol and the corresponding phasing of corn ethanol.

The climate projections were developed by downscaling output from multiple coupled AOGCMs to the locations of watershed weather stations. AOGCMs represent the primary tools available to assess the large-scale climatic response to changes in forcing such as the expected changes in 21st century greenhouse gas concentrations. In this study, eight AOGCMs (Table 3), which were all included in CMIP5 (Taylor et al. 2012), were utilized in developing climate change projections for the RRW land use change scenario simulations. Using AOGCM ensembles incorporates information from different models, often increasing the value of the climate information obtained (Knutti et al., 2010; Martre et al., 2015; Pierce et al., 2009; Weigel et al., 2010) and thus an improved overall climate change impact analysis.

Each of these eight climate models were forced with three representative concentration pathways (RCPs) representing low (RCP2.6), medium (RCP4.5) and high (RCP 8.5) levels of radiative forcing from GHGs (Moss et al., 2010; van Vuuren et al., 2011a). The RCP2.6 pathway depicts future conditions, which represent a “medium development” of global population, income and energy use and land use, resulting in a peak atmospheric CO2 concentration prior to 2100 (van Vuuren et al., 2011a, b). A cost-minimizing approach is used in the RCP4.5 pathway, which assumes that simultaneous efforts occur worldwide to mitigate emissions, including taking into account the cost of reducing emissions per the 100-year warming potential of a respective GHG, resulting in stabilization of atmospheric CO2 concentrations in 2100 (Thomson et al., 2011; van Vuuren et al., 2011a). High energy demand and GHG emissions characterize the RCP8.5 pathway, which occur due to an assumed high population and slow income growth with modest rates of technological change and energy improvement, without implementation of climate change adaptation policies (Riahi et al., 2011; van Vuuren et al., 2011a).

The PC and AC agricultural scenarios reflect land use patterns, management systems and energy use levels that could potentially contribute to higher GHG emissions (Davis et al., 2012), which would be consistent with the RCP8.5 pathway. Planting switchgrass, on the other hand, has a potential to sequester carbon (Keshwani and Cheng, 2009; Davis et al., 2012) and help reduce CO2 emission in the long term. Thus the AC scenario could be viewed as being consistent with the RCP8.5 pathway and the AS scenario could be considered as a system consistent with the RCP2.6 pathway, due to expected lower GHG emissions that would occur during the next century due to the expanded switchgrass production. These hypothetical relationships between the future agricultural scenarios and the RCP pathways are investigated to some extent per the interactions of different agricultural scenarios and climate projections in the Results and discussions section.

Contemporary AOGCMs are archived with a resolution of approximately 2∘, although there is substantial variability in model resolution among participating modeling groups. To conduct impact analysis using models like SWAT, higher-resolution information is required. Thus, downscaling to a finer resolution is crucial to incorporate local climate variability for detailed watershed assessments. Here, a statistical downscaling approach involving regression-based models and stochastic weather simulation, as described by Schoof et al. (2007) and Schoof (2010), was used to derive station-based projections consistent with the projections of the parent AOGCMs under each emissions pathway. These downscaled climate data were then post-processed to produce a comprehensive daily weather data set (precipitation, minimum and maximum temperature, relative humidity, solar radiation and wind speed) for the years 2011 to 2100 to be used in the SWAT model scenario simulations.

In addition to the three emission scenarios (RCPs), the weather data were divided into three temporal blocks of 20 years to represent early (2016–2035), mid- (2046–2065) and late (2076–2095) century climate conditions. As a result, a total of 72 (8 climate models × 3 emission scenarios × 3 temporal scenarios) climate scenarios were created. Moreover, simulating climate change scenarios in SWAT requires the CO2 concentration for the simulation time periods. Accordingly a single average value of CO2 concentration was used in simulating each 20-year temporal block, similar to the approach used by Ficklin et al. (2009), for a given RCP scenario (Table 4). These scenarios were used to run simulations through the calibrated SWAT model for each agricultural scenario discussed in the “Model description and setup” section at the annual timescale.

Reporting SWAT output values for each year was not feasible due to the fact that 360 total land use change and climate change combinations (72 climate × 5 agricultural scenarios) were simulated in the study. Therefore, annual average and standard deviation values for each temporal block (early, mid- and late century), RCP pathway (2.6, 4.5 and 8.5) and agricultural land use change scenario were reported for each output indicator of interest: streamflow (Q), TSS, total nitrogen (TN) and total phosphorous (TP). This approach allowed us to capture both the trends across temporal blocks and agricultural scenarios, and variations within temporal blocks and across climate models. Moreover, the predicted average corn and switchgrass yields were also determined for each temporal block (consisting of eight climate models) for the AC and AS agricultural scenarios, respectively.

Table 5 shows a comparison between historical-observed and future projected average annual precipitation and annual average temperature values along with standard deviations across the years and among AOGCMs. The results show that, on average, annual precipitation and temperature values increase from early to late century (and from RCP2.6 to RCP8.5). Compared to the average historical observations between the years 1991 and 2010, the annual average temperature values for the RCP2.6, RCP4.5 and RCP8.5 pathways within the early, mid- and late century time periods all increased by 1.5–4.2 ∘C (Fig. 3), depending on the RCP and time period. In contrast, there were decreases in average annual precipitation values for all of the scenarios except the late century RCP4.5 and RCP 8.5 scenarios (Fig. 3).

Similar results have been reported in previous studies. Chien et al. (2013) reported that, compared to 1990–1999, the average temperature increased by up to ∼ 3 ∘C (∼ 5 ∘C) for 2051–2060 (2086–2095) and the percentage change in annual precipitation was about -28 to +8 % (-33 to +16 %) for 2051–2060 (2086–2095), using data from nine GCMs (general circulation models) for four watersheds, which cover portions of Illinois, Indiana and Wisconsin. Similarly, Ficklin et al. (2013) analyzed downscaled temperature and precipitation projections from 16 GCMs (two emission scenarios, low (B1) and high (A2)) for Mono Lake basin, California, and found that the 2070–2099 annual average temperature increased by 2.5 and 4.1 ∘C for B1 and A2 scenarios, respectively, compared to 1961–1990. However, they also reported that there was a slight but statistically insignificant decrease in annual precipitation on average. These previous studies confirm the results found here that there is a consistent trend of increases in temperature across climate models and geographical locations, while precipitation could increase or decrease depending on the choice of AOGCMs, projection pathway and geographical location of the analysis.

The interannual variation (standard deviation) was much higher (≤ factor of 4) for the historical-observed temperature and precipitation vs. the corresponding future projections (Table 5). The variations among climate models increased for both temperature and precipitation from early to late century. Moreover, the standard deviations among AOGCMs were higher than (≥ factor of 2) the interannual variations for annual average temperature values. Chapman and Walsh (2007) found similar differences between models and interannual variabilities (standard deviation) of temperature using 14 AOGCMs. For average annual precipitation values, the standard deviations among AOGCMs were slightly higher than interannual variations. These results were mainly due to the consideration of an ensemble of AOGCMs that has an effect of reducing interannual variations compared to interannual variations from individual AOGCMs (Knutti et al., 2010). Therefore, one should take into account these effects in using ensembles of AOGCM results for impact analysis. Moreover, variations among AOGCMs may indicate that the choice of models within an ensemble for climate change impact analysis may result in different conclusions.

The historical (1991 to 2010) annual average Q at the watershed outlet was about 212 mm (63 m3 s-1). There were both predicted decreases (1 to 24 %) and increases (3 to 75 %) in Q for the BL, AC and PC agricultural scenarios in response to the different climate projections (Fig. 4a–c), relative to the historical average Q. For the PS and AS scenarios, however, there were decreases (15 to 83 %) in Q for all but one of the climate projections (Fig. 4a–c). Despite decreases in precipitation and increases in temperature, an increase in Q in some of the scenarios indicates the possible occurrence of larger and more frequent high intensity precipitation events than the historical-observed values in the projected climate data (Schoof, 2015; Kharin and Zwiers, 2000). Moreover, a reduction in ET (evapotranspiration) due to increased CO2 levels, especially in the mid- and late century periods, also contributed to simulated increases in streamflow in mid- and late century scenarios similar to results reported by Jha et al. (2006) and Wu et al. (2012).

The PS and AS scenarios resulted in lower estimated Q compared to the other scenarios and the historical baseline, for a given climate scenario, while very small difference were observed between the BL, AC and PC scenarios (Fig. 4a–c). The AS agricultural scenario exhibited the highest decrease in streamflow (or water yield) as expected. Similar results were indicated by previous studies (e.g., Kim et al., 2013; Parajuli and Duffy, 2013; Schilling et al., 2008; Wu et al., 2013). This reveals that large-scale conversion to switchgrass could result in reduced water availability due to increased ET and conversely reduced Q, which could render it less desirable as a climate change adaptation strategy in the watershed for future climate conditions that manifest lower precipitation levels. Also, as noted previously, the AC and AS scenarios reflect agricultural production schemes that are consistent with the high GHG emission RCP8.5 pathway and the low emission RCP2.6 pathway, respectively. A comparison on this basis reveals that the AS scenario resulted in a much higher reduction in Q compared to the AC scenario, relative to the previous comparison (Fig. 4a–c), which further underscores that widespread adoption of just switchgrass in current intensively cropped Corn Belt watersheds may not be a viable strategy in mitigating climate change impacts on water availability.

Comparisons were also made between climatic projections for a given agricultural scenario. The results show a decrease in Q relative to historical-observed values for early century under all RCPs (Fig. 4a). At mid-century, decreases in Q were predicted for the majority of agricultural scenario–climate projection combinations, except for the BL, AC and PC scenarios in response to the projected RCP8.5 pathway. However, there was a consistent increase in Q, during the late century time period, across agricultural scenarios in response to the RCP8.5 projection and for the BL, AC and PC scenarios when impacted by the RCP4.5 projection (Fig. 4a–c). These increases in Q from early-to-late century could be attributed to the precipitation increase in the same manner as discussed in Sect. 4.1. Except for the early century time period, Q increased from RCP2.6 to RCP8.5 for all agricultural scenarios. The maximum increases (≤ 75 % of historical Q) were simulated under the late century RCP8.5 for all agricultural scenarios.

Previously, Jha and Gassman (2014) used an ensemble of GCMs projections, developed within the framework of CMIP Phase 3 (CMIP3; PCMDI, 2016), to simulate the impacts of projected future climate change on the RRW with SWAT. They concluded that there was an overall average decrease in total Q of 17 % in the mid-century period, compared to Q for the years 1961 to 2000. Similar BL scenario results were obtained in this study for the RCP2.6 projections (14.7, 11.0 and 14.7 % for the early, mid- and late 21st century, respectively) and RCP4.5 and RCP8.5 early century projections (23.7 and 12.4 %, respectively). The mid-century RCP4.5 scenarios showed a slight decrease (2.6 %) in Q, whereas increases in Q were simulated for the RCP8.5 scenario in both the mid- and late century (12.3 and 72.6 %, respectively), and for the late century RCP4.5 scenario (9 %).

The standard deviations of annual Q between AOGCMs and future time periods (Fig. 4 and Tables A1 and A2) followed trends similar to the temperature and precipitation results discussed in Sect. 4.1. The standard deviation across time periods for the historical period was greater than for any of the future temporal periods for all of the agricultural scenarios. Similarly, the standard deviation between AOGCMs is greater than that across future time periods. These trends are also similar for all TSS, TN and TP values (Fig. 4 and Tables A1 and A2).

Simulated TSS impacts for the different agricultural scenario–climate projection combinations were compared to each other and vs. the simulated historical TSS values. The historical (1991–2010) annual average TSS concentration at the watershed outlet was about 113 mg L-1 (2.25 × 105 t). Compared to the historical TSS concentration, there were increases in TSS for the AC and PC scenarios across all climate projections, decreases for the PS scenario for most of the climate projections and decreases for AS in all three climate projections (Fig. 4d–f). The increases in TSS were the highest for the AC (≤ 67 %) scenario, followed by the PC (≤ 65 %) and BL (≤ 63 %) scenarios. Peak TSS decreases were 74 and 27 % for the AS and PS scenarios, respectively.

For a given climate scenario, there were 1.6–7.1 % increases in TSS for PC and 2.3–11.1 % increases for AC compared to the BL scenario. This indicates how intensively the RRW is utilized for agricultural production already. It was only when switchgrass was introduced (AS and PS scenarios) that significant decreases in TSS were observed (18–27 % for PS and 56–74 % for AS) relative to the BL scenario. Hence, switchgrass seems to be a good adaptation strategy with respect to addressing TSS reductions. This result is magnified when results are assessed based on agricultural scenarios simulated with the appropriate climate scenarios, as discussed in “Streamflow (Q)” section. Generally, the predicted TSS values followed the Q trends for all of the climate projection and agricultural scenario categories (Fig. 4d–f). For a given agricultural scenario, TSS increased continuously from the early-to-late 21st century. Considerable reduction in TSS was simulated in the AS agricultural scenario under all climate scenarios compared to historical levels. However, the AS scenario must be viewed as extreme and impractical, due to the importance of corn as a crop in the RRW and Corn Belt region in general. However, the PS agricultural scenario, which is a more plausible scenario, may require additional best-management practices to significantly reduce TSS yield and transport from the watershed.

The annual average historical-simulated (1991 to 2010) TP at the watershed outlet was roughly 4.52 × 103 t (or 7.6 mg L-1). Comparisons were made between the different scenario results, and between historical and scenario results. Due mainly to the absence of phosphorus fertilizer application and reduction in surface runoff when planting switchgrass, there were significant reductions in TP in the PS and AS agricultural scenarios compared to historical-simulated values (up to 66 and 99 %, respectively) and BL scenario (up to 49 and 99 %, respectively) (Fig. 4d–f). The differences in tillage practices between agricultural scenarios also contributed to the difference in TP output among scenarios, due to the shifts in tillage practices used in the BL scenario vs. just conventional tillage for CC in the PC and AC scenarios, and elimination of tillage for the PS and AS scenarios. Conventional tillage practices result in higher sediment and phosphorus yields but conservation and no-till tillage practices can result in lower yields under some conditions. Various researchers (e.g., Parajuli et al., 2013; Tomer et al., 2008; Andraski et al., 2003; Bundy et al., 2001) have demonstrated similar effects of tillage practices on sediment and/or phosphorous outputs from agricultural fields.

For a given climate scenario, the PS and AS scenarios exhibited similar reductions in TP output (≤ 49 and 99 %, respectively) compared to the BL scenario. Both of the CC-based scenarios (PC and AC) resulted in large increases in TP, compared to both the BL scenario for all climate projections (≤ 36 and 41 % for PC and AC, respectively) and historical simulated values (≤ 62 and 67 % for PC and AC, respectively).

The annual average historical-simulated (1991 to 2010) TN load value at the watershed outlet was about 2.14 × 104 t (or 36 mg L-1). Comparisons were made between simulated historical and scenario annual average TN load values at the watershed outlet, and also among scenarios. These comparisons reveal two important insights: (1) the AC scenario resulted in lower TN loads relative to the BL scenario, which was not originally expected, and (2) the PC scenario resulted in the highest TN loads of all of the agricultural scenarios (Fig. 4a–c). This implies that, with respect to TN output, the current agricultural management conditions (BL scenario) in the RRW are already extremely intensive, and are comparable to planting continuous corn everywhere with conventional tillage and 202 N kg ha-1 of fertilizer (AC scenario). Even though the fertilizer application rates were less than 202 N kg ha-1 in the BL scenario, manure was applied in addition to the fertilizer (Teshager et al., 2015). This resulted in a slightly higher TN load for the BL scenario. However, as previously described, the PC scenario reflects a combination of BL scenario cropping system and management practices, and conversion of some land parcels to CC, resulting in slightly higher TN loads as compared to both the BL and AC scenarios. Also, the introduction of switchgrass in the RRW AS and PS scenarios has the potential to reduce the total nitrogen outflow from the watershed significantly relative to historical levels (≤ 84 % for AS and ≤ 35 % for PS) as shown in Fig. 4a–c. For a given climate projection, annual average TN loads were reduced by to 81 % for AS and 18 % for the PS scenarios in comparison to the BL scenario. This was due to both the elimination of tillage in switchgrass cropping systems and the capability of switchgrass to scavenge nitrate from the soil–water matrix. Planting switchgrass in select areas of a watershed, similar to the PS scenario approach, and implementing effective best-management practices could further reduce nitrogen losses to Corn Belt stream systems. The effects of expanded adoption of switchgrass depicted in the PS and AS scenarios on reductions in TN loads are further magnified when examining the results within the context of the RCP4.5 and RCP2.6 pathways, which were previously identified as the two respective pathways that the PS and AS scenarios were most correlated with, especially for the late century time period. Similar to the Q and TSS results, the TN loads increased from the early part of the century to the late part of the century, especially for the RCP4.5 and RCP8.5 pathways (Fig. 4a–c).

Crop yield analyses were done to point out the potential impacts of climate change on corn and switchgrass yields, assuming that the current production technologies for both crops remain the same, based on crop yield estimates obtained from the AC and AS scenarios. The 20-year (1991 to 2010) historical-simulated average yields across the entire RRW was 10 t ha-1 for corn and 15.5 t ha-1 for switchgrass. The AC scenario corn yields and AS scenario switchgrass yields were predicted to decline across future climate conditions, as compared to the historical-simulated yields, especially during the mid- and late centuries for the higher RCP4.5 and RCP8.5 GHG emission pathways (Fig. 5).

The reduction in corn yields ranged from 7 % during the early century time period to 25 % in the late century time period (Fig. 5). However, no reductions were predicted for switchgrass yields initially in the early century, but estimated declines in switchgrass yields of ≤ 19 % occurred in the latter part of the century. In the early century, the effects of the emission pathways on the crop yields were insignificant; however, the emission pathway effects became more pronounced in the mid- and late century simulations (Fig. 5). There were essentially no differences in corn or switchgrass yields between the early, mid- and late century time period simulations for the low emission RCP2.6 pathway. The highest yield reductions, 25 % for corn and 19 % for switchgrass, were simulated in response to the high emission RCP8.5 pathway at the end of the century. Lower percentage crop yield reductions were found in this study compared to similar previous research results (e.g., Miao et al., 2015; Ummenhofer et al., 2015; Cai et al., 2009; Schlenker and Roberts, 2008). One possible reason that lower reductions in crop yields were predicted within this study could be the inclusion of CO2 concentrations during the simulations, and the capability of SWAT to account for positive effects of CO2 concentration on crop yield. In addition, higher precipitation amounts that characterize the RCP8.5 pathway late century time period could have partly offset the effects of increased temperatures on yield. However, the predicted corn and switchgrass yields for the RCP8.5 pathway late century time period were lower than other time periods, even though the average annual precipitation was higher than the historical or any other future projected precipitation. This result is consistent with the results presented in Sect. 4.2 because the increase in annual precipitation was due mainly to more high intensity daily precipitation events (Schoof, 2015), which will not necessarily be beneficial for crop growth.