Four long-term and presently on-going agricultural research experiments within the dryland cropping region of the iPNW were used for our study (Table 1). Additional description of these sites can be found in Morrow et al. (2016). The four field experiments are located at the following research centers: Kambitsch Farm near Genesee, ID, operated by the University of Idaho; the Washington State University and USDA Palouse Conservation Field Station near Pullman, WA; the Columbia Basin Agricultural Research Center near Pendleton, OR; and the Columbia Basin Agricultural Research Center near Moro, OR. These long-term studies were selected based on the cropping systems present as well as the range of climatic conditions collectively represented by the sites. Together, these four locations span three agroecological classes (AECs), or land-use classifications that represent unique biophysical and socioeconomic conditions resulting in distinct cropping systems (Huggins et al., 2014): (1) annual cropping (limited annual fallow) represented by Kambitsch Farm and Palouse Conservation Field Station; (2) annual crop-fallow transition (e.g., 3-years rotations with fallow every third year) represented by the Pendleton Research Station; and (3) grain-fallow (e.g., 2-years wheat [Triticum aestivum L.]-fallow rotation) represented by the Moro Research Station.

The four study sites represent a range of tillage and cropping intensities. The Natural Resources Conservation Service (NRCS) soil tillage intensity rating (STIR) (USDA-NRCS, 2006) was applied to assess the soil disturbance in each treatment. The STIR rating captures the range and severity of soil disturbance caused by management operations over an entire crop rotation, from spraying herbicide to plowing, planting and harvesting. The final STIR value is annualized based on the operations required for the entire rotation (Table 1).

The NRCS establishes a STIR of 30 as the maximum value for no-tillage (NT) operations (USDA-NRCS, 2011). In the site descriptions and in subsequent tables, all treatments with a STIR exceeding 30 are referred to as conventional tillage (CT), while those under 30 are referred to as NT. The primary tillage implements for CT treatments are listed in Table 1.

Cropping intensity of each treatment was quantified according to the proportion of time there was growing crop cover (Table 1). Each crop or phase was assigned a value, and these values were averaged over the length of the rotation. The following cropping intensity values were applied: perennial crops—0.83 (10 of 12 months), fall planted, winter annual crops—0.75 (9 of 12 months), spring planted, annual crops—0.33 (4 of 12 months).

Each site was sampled in June or July of 2013 prior to harvest from 0- to 10-cm depth, collecting 50–60 samples across each plot using a hand-operated soil probe (1.8 cm inside diameter). Four cores (3.0 cm inside diameter) were also collected from 0- to 10-cm depth from each subject plot using a hand-operated soil probe to determine soil bulk density and gravimetric water content. All samples were immediately placed on ice in the field and transferred to cold storage at 4°C until further processing.

Actual values of soil C and N properties and agricultural management treatment effects can be found in Morrow et al. (2016), whereas in this paper we are primarily using these data for correlation and regression analyses. Stepwise multivariate analysis using SAS (SAS Institute, 2014) was used to model the influence of climate variables and management factors on the soil C and N properties. Initial model entry was set at p < 0.25, with significance for variables remaining in the model set at p < 0.10. Pearson correlations were also determined for all variables in the study with significant correlations set at p < 0.10. A less rigorous alpha was selected due to the inherent spatial variability of many of the soil C and N properties measured. In consideration of this, an alpha of 0.10 was considered a judicious step toward minimizing type II errors (failing to reject a false null hypothesis).

Two climate scenarios form the basis for future climate in 2030 and 2070 at the four study sites. Representative concentration pathway (RCP) 4.5 represents moderate climate action (Thomson et al., 2011), while RCP 8.5 is a baseline scenario that assumes no target reduction in greenhouse gas emissions (Riahi et al., 2011). Using these two climate scenarios, projected MAT and MAP of each site for 2030 and 2070 were determined by averaging across the MAT and MAP outputs for each site from 19 to 20 climate models considered most appropriate for the iPNW and where data were available (data unavailable for one model). The climate data were from a statistical downscaling of global climate data from the Coupled Model Intercomparison Project 5 or CMIP5 (Taylor et al., 2011) utilizing the Multivariate Adaptive Constructed Analogs (MACA) method (Abatzoglou and Brown, 2012) and can be accessed at <http://maca.northwestknowledge.net/GCMs.php>. The resulting MAT and MAP of each site for 2030 and 2070 were then used to determine the future climate ratio, which is the ratio of MAT to MAP; subsequently a higher climate ratio indicates a warmer, drier climate. The present day relationship between the climate ratio and average SOC and total N at all sites was used to create a regression equation that was then used to predict SOC and total N using future climate ratios determined for each of the climate scenarios described.

POXC, SOC, total N, HC, NHC, HN, and NHN had positive Pearson correlations with MAP, ranging from 0.7 to 0.84, whereas these same variables were negatively correlated, although less strongly, with MAT (Table 2). Variable means and treatment effects were previously reported (Morrow et al., 2016). The MAP/MAT had the greatest Pearson correlations with measured SOM-related variables ranging up to 0.88 (Table 2). MAP explained 57 and 69%, respectively, of the variation in SOC and total N in the multivariate regression analysis that also included MAT, STIR, and cropping intensity as explanatory variables (Table 3). When MAP was removed from the regression model, MAT was the dominant significant variable, explaining 42% of SOC variation and 49% of total N variation (Table 3). These data show that management variables of STIR and cropping intensity explained little of the variability in SOC and total N as compared to climate related variables (Table 3).

The hydrolyzable and non-hydrolyzable fractions of soil C and N are also correlated with MAP and MAT, similar to SOC and total N (Table 3). MAP explains 50% of NHC variation, 55% of NHN variation, and a slightly greater percent of variation in the hydrolyzable fraction, explaining 57% of HC variation and 63% of HN variation. As with SOC and total N, if MAP is removed from the model, MAT becomes the major significant factor in explaining the variation in the hydrolyzable and non-hydrolyzable fractions of SOC and total N (Table 3).

Some of the more labile soil properties measured in this study, including WEON, MBC, and Cmin_0−1d, were positively correlated with MAP, whereas only WEON along with WEOC displayed a significant correlation with MAT (Table 2). The microbial metabolic quotient (qCO₂) is an indicator of microbial efficiency, where increases in qCO₂ indicate a decrease in microbial efficiency (Smith et al., 2002). In the present study, qCO₂ had a negative relationship with MAP (Table 2), indicating a decrease in MAP was associated with a decrease in microbial efficiency. MAP was also an important factor, more so than management factors, in explaining the variation in MBC (r² = 0.22), 1-day Cmin (r² = 0.12), qCO₂ (r² = 0.26), and WEON (r² = 0.14; Table 3). Variation in MBN was not explained by management or climate factors (Table 3). When MAP was removed from the model, MAT was not a signficant variable in explaining variation in these more active soil properties (Table 3).

The multivariate analysis of climate along with management indicates that soil C and N of surface soils within the scale covered by this study were influenced by MAP and MAT more so than tillage and cropping intensity. Similarly, Colman and Schimel (2013) found that on a continental scale across 84 different soils, MAP and MAT could explain 60% of the observed variation in SOC. Results also showed that more active soil C and N pools displayed lower or no significant relationship to MAT, whereas the more recalcitrant pools were strongly correlated with MAT (Table 2). These results are consistent with enzyme kinetic theory established by Arrhenius (Davidson and Janssens, 2006), which dictates that reactants with higher activation energies (recalcitrant SOM) have higher temperature sensitivities. In other words, as MAT increases across the temperature gradient, the more recalcitrant soil C and N pools experience a greater relative increase in decomposition rate compared to more labile SOM (Davidson and Janssens, 2006).

PNM, PDR, and BDR were all positively correlated with MAP and negatively correlated with MAT (Table 2). These trends between MAP, MAT, and soil N transformations reflect the opposing gradients of MAP and MAT across the iPNW. Greater MAP is associated with colder MAT and greater SOM accumulations and labile SOM constituents. Laboratory incubations occur under favorable environmental conditions for promoting soil N transformations and the results express the combined influence of MAT and MAP on SOM accumulations and consequent measures of PNM, BDR, and PDR.

Variability in N mineralization has been reported to be primarily correlated with MAT, MAP, as well as total N and N deposition (Chapman et al., 2013). In agreement with our results, Colman and Schimel (2013) concluded that the direct drivers of N mineralization were soil C and N as well as clay content and that precipitation drives N mineralization indirectly through its influence on soil C and N content. Regarding PDR, Niboyet et al. (2011) found that experimentally increasing precipitation 50% for 8 years resulted in a 22% increase in PDR, in agreement with the positive correlation of PDR with MAP found in our study. Niboyet et al. (2011) did not, however, detect sensitivity of PDR to warming of 1°C for 8 years. The MAT gradient in our study spans 1.9°C (Table 1) and the negative correlation of MAT with PDR represent long-term climatic conditions rather than short-term increases that have been experimentally induced.

Soil organic C and total N as well as the hydrolyzable and non-hydrolyzable fractions were negatively correlated with the ratio of MAT to MAP, called the climate ratio (Weil and Magdoff, 2004; Figure 1, Table 4). Here, SOC and total N along with the hydrolyzable and non-hydrolyzable fractions are greatest at the Palouse Conservation Field Station and Kambitsch Farm, where the climate ratio approaches one, and decrease as the climate ratio increases (Figure 1). Similarly, Weil and Magdoff (2004) reported the greatest SOC when the climate ratio is near one, decreasing as the ratio increases. Consequently, the climate ratio provides a simple indicator of future climate change effects on SOC and related properties.

The relationship between the climate ratio and soil C and N indicates that a rise in MAT without a proportionate rise in MAP will result in degradation of surface SOM across the iPNW. For the iPNW, RCP 4.5 and RCP 8.5 both predict disproportionate increases in MAT relative to MAP across the region for 2030 and 2070 average projections of 19 climate models as compared to present day averages (Abatzoglou and Brown, 2012). Consequently, future climate ratios are projected to increase as much as 27–46% by 2070 for RCP 8.5 across the study sites (Table 4). Increases in the climate ratio will drive a decline in average SOC and total N across the region for 2030 and 2070 (Figure 2), under the assumption that a new steady state of SOC and total N is realized. For example, SOC values for 2030 and 2070 at Pendleton, based on forecasted shifts in MAT/MAP under the RCP 4.5 scenario, are 3.6 and 12.3%, respectively, below the present day average for the Pendleton site across all treatments (Figure 2). Likewise, at the same site under the RCP 8.5 scenario and based on forecasted shifts in MAP/MAT for 2030 and 2070, SOC levels may be reduced by 6.9 and 21.2%, respectively. Calculated another way and for scalability, the slope of the regression lines in Figure 2 coupled with current soil bulk densities indicate that every 1/10 increase in MAT/MAP is associated with a decrease in soil C and N of 725 kg ha⁻¹ and 58 kg ha⁻¹, respectively. Taking regional averages from the present data set, this equates to an average reduction in soil C and N of 3.3% and 3.4%, respectively, for every 1/10 increase in MAT/MAP. This analysis does not take into account potentially offsetting unknown microbial responses to temperature increases, such as changes in enzyme production that could alter the present day relationship between decomposition and temperature, or greater water-use efficiencies and potential biomass production resulting from increasing concentrations of atmospheric CO₂ (Stöckle et al., 2010). This analysis, however, does provide scale to potential implications of climate change to soil C and N, and subsequently soil health, across the iPNW. Loss of soil C is associated with a reduction in soil aggregate stability, reduced water holding capacity and porosity, reduction in root aeration and overall root health, and a decrease in cation exchange capacity (CEC), all soil properties critical to soil health and function (Delgado et al., 2011).

While overall SOM stocks are susceptible to projected changes in climate within the iPNW, understanding how SOM pools representing different turnover times are impacted by future climate scenarios is also important. Both hydrolyzable C (HC) and non-hydrolyzable C (NHC), which has a longer turnover time than the HC fraction (Paul et al., 2006; Rovira and Vallejo, 2007), were negatively correlated with MAT and positively correlated with MAP (Table 2). Under elevated temperatures, Plante et al. (2010) demonstrated accelerated decomposition of SOM resistant to acid hydrolysis as well as particulate organic matter, typically consisting of more labile C. A comparison of the hydrolyzable and non-hydrolyzable fractions of SOC with the climate ratio revealed that both declined similarly as the climate ratio increased (Figure 1). SOM pools with longer turnover times reportedly exhibit a greater relative increase in decomposition rate than SOM pools with shorter turnover times as temperature increases (Lefevre et al., 2014). Therefore, our results indicate that interactive effects of MAT and MAP on C inputs and decomposition are operational and that relative changes in temperature and precipitation must be considered when assessing climate change impacts on SOC pools. In contrast to HC and NHC, the HN fraction, composed primarily of amino acids and amino sugars, displayed greater sensitivity to the climate ratio than the NHN fraction (Figure 1). This is consistent with the concept that a portion of HN may be stabilized by the presence of more processed material, represented in NHC (Pare et al., 1998), and therefore a decline in NHC would further accelerate a decline in HN.

Loss of recalcitrant SOM fractions may not be easily replaced under traditional cropping systems. Annual inputs from agriculture to more recalcitrant SOM pools as compared to that derived from native vegetation have often been reported to be minimal (Balesdent et al., 1990; Huggins et al., 1998; Poirier et al., 2006). Contributing factors are reductions in quantity of annual C inputs, decreases in recalcitrant inputs (e.g., lignin), and/or increases in disturbance or erosion that result from a switch from native to agricultural conditions (Hassink, 1997; Huggins et al., 1998; Krull et al., 2003). In part, this may explain why management factors, particularly tillage and cropping intensity, were not very significant factors influencing soil C and N properties as compared to MAT and MAP across the four study sites (Table 3).

Declines in labile and recalcitrant constituents of SOM with increasing climate ratio have implications for associated soil functions. More recalcitrant SOM benefits soil chemical and physical properties, such as CEC and water holding capacity, while more labile SOM is essential for soil biological processes such as nutrient mineralization (Wander, 2004). The demonstrated sensitivity of various constituents of SOM to the climate ratio suggests that soil chemical, physical, and biological properties are susceptible to degradation resulting from predicted climate change in the iPNW.

Climate variables were major drivers of total N and, in turn, total N was the dominant driver of both PNM and PDR (Table 5). Therefore, the primary influence of future climate change on N transformations will be through soil N impacts. Replacing total N with NHN and HN in the multivariate model shows that HN is the dominant variable for explaining variation in both PNM (R² = 0.58) and PDR (R² = 0.51). MAT, however, was the main variable that explained BDR variability (Table 5). Li et al. (2010) modeled the influence of climate, along with soil and management factors associated with a maize-wheat rotation and found SOC had the greatest impact on N₂O emissions; the model did not include soil N. This emphasizes along with present findings that substrate availability is a critical driver of N transformations. The negative correlations of TN, PNM, and PDR with both MAT and MAT/MAP (Table 2) indicate that these variables will decrease below present day values with projected climate change. Reductions in PNM may lead to a greater reliance on inorganic fertilizers and possible decreases in N-use efficiency as applied inorganic N is subject to losses via multiple pathways (Culman et al., 2013). This places greater importance on precision N applications under future climate scenarios to help increase N-use efficiency and minimize potential losses.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.