Potential Impacts of Changing Precipitation Patterns on Biological Nitrogen Fixation in Soybean (Glycine Max L.) as Mediated by Landscape Position and Tillage

Purpose Expected changes in rainfall patterns will affect the timing of N-mineralization and other N transformations, potentially promoting or suppressing biological nitrogen �xation (BNF). We test the hypotheses that BNF is more sensitive to changing rainfall patterns in summit vs. toeslope positions and in till vs. no-till consistent with patterns of soil texture and organic matter. Methods At a site in the


Introduction
Cropping systems receive nitrogen (N) additions primarily from organic amendments, manufactured fertilizers, and biological N 2 xation (BNF). These inputs, most notably the use of fertilizer, account for dramatic increases in reactive N (Nr) on Earth over the last 100 years (Vitousek et al. 2013). When not taken up by plants or soil microorganisms, N r can be lost from cropping systems and become pollutants in water and the atmosphere. BNF contributes ~ 25% to N r inputs to the biosphere (Vitousek et al. 2013) (Rao et al. 1994); soybeans (Glycine max L.) in particular contribute ~ 10.4 Tg N r yr − 1 , representing ~ 18% of total global BNF inputs (Gelfand and Robertson 2015). Soybean BNF can substitute for N fertilizer application and has a lower environmental cost since systems with BNF as a major source of N r require less N fertilizer and tend to have lower hydrologic N r losses (Blesh and Drinkwater 2013; Syswerda et al. BNF has been shown to vary with soil properties. In southwestern Victoria, Australia, for example, in a survey of 71 dairy pasture sites BNF in white clover (Trifolium repens) ranged from 0 to 100% of total plant N across different soil textures with rates 7% higher on light-textured soils (Riffkin et al. 1999). In a Denmark pea (Pisum sativum L.) eld sampled at 56 points, BNF ranged from 26-81% (Hauggaard-Nielsen et al. 2010). And at sites in Central New York, USA, soybean BNF differed by soil type (Schipanski et al. 2010).
Climate change, particularly changes in rainfall intensity and amount, may also affect BNF. The US Midwest, responsible for > 80% of US soybean production (USDA NASS 2018), has experienced increasingly more intense and frequent heavy rainfall over the past few decades (Pryor et al. 2014), resulting in longer dry periods between rainfall events. Global circulation models predict that both the increasing length of dry intervals between precipitation events and the amount of precipitation falling in single events will further increase (Pryor et al. 2014). In Michigan, both the number of extreme precipitation events and observed annual precipitation amounts are increasing (Frankson and Kunkel 2017). Because changes in rainfall will likely be accompanied by changes in the timing of Nmineralization and other N transformations (Bowles et al. 2018; Robertson et al. 2013), potentially promoting or suppressing BNF.
Given these changes and the importance of BNF to legume crop productivity and as well to soil fertility and the N nutrition of subsequent crops, it seems prudent to examine potential impacts of future climates on BNF. Since BNF appears related to soil organic matter (OM) and texture, two primary determinants of soil moisture and N availability, one might expect the strength of BNF responses to changing rainfall patterns to vary across elds with as OM and texture, particularly along toposequences. For example, impacts at summits may be greater than at toeslopes, where greater OM, clay, and water holding capacity and may buffer against rainfall variability.
In general, BNF rates are lower under drought conditions due to nodule sensitivity to decreased phloem ow (Serraj et al. 1999). Although most studies investigating drought effects on BNF have been performed in greenhouse settings, where plants are protected from normal environmental conditions and the rhizobia-plant symbiosis does not re ect the impact of rainfall changes. Furthermore, arti cial environments alter nodules, as well as nodule depth and distribution (Pueppke 1986). Field experiments are needed to better understand BNF response to natural variability and that imposed by climate change.
Particularly needed is an understanding of soil by rainfall interactions that may become more pronounced with climate change. Incorporating this knowledge into quantitative models will allow for better predictions of BNF and its global consequences.
Here we examine the hypotheses that BNF is vulnerable to changing rainfall regimes in summit vs. toeslope landscape positions and, separately, under till vs. no-till management. In both cases we expect BNF responses to changes in rainfall to be attenuated where OM and texture favor water availability. We address three speci c questions: The experiments were conducted on conventionally tilled rainfed elds used for corn, soybean, and wheat planted to soybeans during the years of each experiment (Fig. 1). These elds were managed as per regional practice, including fertilization with potassium (KCl), phosphorus (potash), and lime as recommended by soil tests. All elds were planted at 150,000 seeds ha − 1 in 15-inch rows. Glyphosate was used to control weeds during soybean years. For all experiments we quanti ed soybean BNF by 15 N natural abundance using nodulating and non-nodulating isolines, as described below.
Toposequence Experiment (Question 1) For the toposequence experiment, we used three elds of 6-8 ha each (KBS elds 30 − 1, 38, and 79 − 8 south; Fig. 1). Paired 1 ⋅ 1 m plots were placed along four transects in three landscape positions (summit, midslope, and toeslope) in each eld (Table 1). In one plot of each pair we planted in 2015 ve nonnodulating soybean plants (described below) at the same density as in the rest of the eld. At physiological maturity plants were harvested both above and below ground and separated into aboveground vegetative biomass, belowground biomass, and seeds as detailed below.

Rainfall Intervals by Tillage Experiment (Question 3)
We conducted the rainfall intervals by tillage experiment in four replicate 1 ha plots in each of the conventional and no-till treatments of the KBS LTER main cropping system experiment (MCSE; Robertson and Hamilton, 2015). In 2015, in one 4 x 4 m subplot per replicate we imposed rainfall intervals of two weeks for the period between soybean planting and harvest; adjacent control subplots received ambient rainfall. Rainfall was excluded by complete-exclusion rainfall manipulation shelters (5 × 5 × 1.5 m high as

Biological Nitrogen Fixation
We quanti ed soybean BNF by using 15 N natural abundance in nodulating (Pioneer P22T69R) and nonnodulating (PI 547695, seed source: 04U-3266; Horosoy variety) isolines obtained from the USDA Soybean Germplasm Collection (USDA-ARS, Urbana, IL). To plant the non-nodulating isolines, we removed nodulating seeds from the soil immediately following planting and replaced them with nonnodulating seeds. Seeds were planted in late May and early June and plants harvested in late September and early October at the R6.5 stage (physiological maturity).

Soil Analyses
Soil was sampled in all experiments by compositing four 2.5 cm diameter × 25 cm depth push-probe soil cores on each sample date in each subplot replicate. Soils were passed through a 4 mm mesh screen and analyzed for texture, inorganic N, N mineralization, organic matter, gravimetric moisture, and pH at the time of peak N xation, mid-season.
Soil texture was measured using the hydrometer method (https://lter.kbs.msu.edu/protocols/108). Forty g of air-dried soil were shaken for 24 hours in 200 mL jars using sodium hexametaphosphate as a dispersant. The slurry was then put in 1 L cylinders and supplemented with water. Hydrometer and temperature readings were taken over 8 hours. Since sand falls out of solution too quickly to accurately record density changes, to supplement hydrometer readings sand the remainder of the sample was sieved out with a 53 µm mesh.
Soil pH was measured for two duplicate subsamples. A slurry of 15 g of eld moist soil and 30 mL deionized water was shaken by hand for 10 seconds than allowed to settle for 30 minutes before measuring pH (VWR International, Randor, PA) (https://lter.kbs.msu.edu/protocols/163) Fe, K, and P contents were analyzed on composite soil samples by atomic absorption spectroscopy following extraction with 0. Potential N mineralization was measured via a 28-day laboratory incubation where 10 g of soil were held at 60% WFPS in a 25°C incubator. Inorganic N was measured by extracting 10 g of soil in a 100 mL solution of 1 M KC, followed by shaking and ltration.
Soil organic matter (SOM) was measured by the Cornell Nutrient Analysis Laboratory (Ithaca, NY) via loss on ignition whereby soil was weighed into crucibles and placed in a 500° C oven for two hours. Weights were recorded before and after combustion and adjusted for soil moisture.

Plant Analysis
Whole plants were harvested at physiological maturity (R6.5) to determine total nitrogen xed and soybean biomass. For root harvest in 2015 (toposequence experiment), coarse and ne roots were collected within a 25 × 25 × 25 cm soil volume. Soil was then carefully shaken from roots in the eld and roots were examined for the presence or absence of nodules. Roots were then placed on a 0.1 mm screen and washed gently with water to remove adhering soil. All plants from the same replicate subplot were composited and dried to a stable weight in a 60°C forced air oven. Plant material was divided into grain, aboveground vegetative biomass (stem, leaves, and pods, less grain), and, when sampled, belowground biomass (roots including nodules).
Biomass was weighed and ground to pass through a 1 mm sieve and 3-5 mg of homogenized plant material were weighed into tins. Packed tins were analyzed for 15 where N ref represents tissue from non-N 2 -xing soybean isolines, N x represents tissue from N 2 -xing soybean isolines, and N b represents tissue from N 2 -xing soybean isolines grown with atmospheric N 2 as the only N source.
We used the δ 15 N b value determined by Gelfand and Robertson (2015), who used the same commercial variety grown in N-free sand culture in a KBS greenhouse with N-free Hoagland's solution (0-7-5 NPK with micronutrients; GreenCare Fertilizers, Chicago, IL, USA).

Statistical Analysis
For Question 1, all statistical analyses were performed using R software version 3.6.1 (R Development Core Team 2019) and with a signi cance value of P < 0.05. We t the BNF data with a linear mixed model using the "lme4" package with landscape positions as a xed factor, and elds and transects as random factors. To compare BNF differences among landscape positions, pairwise T-tests were conducted with the "lsmeans" package.
For Question 2, the statistical model included two rainfall treatments and two landscape positions and the interactions between them were considered xed factors. Fields were considered a random factor. Landscape position was speci ed as the whole plot factor, the interaction between elds and landscape position were considered a random factor, and this interaction was used to test landscape position effects. Analysis of variance was used by considering landscape positions as a whole plot factor, and rainfall treatments as a subplot factor.
For Question 3, the statistical model included two rainfall treatments and two tillage treatments and the interactions between them were considered xed factors. Blocks were considered a random factor. Tillage treatment was speci ed as the whole plot factor, the interaction between block and tillage treatment was considered a random factor, and this interaction was used to test tillage treatment effects. Analysis of variance was used by considering crops as a whole plot factor, and rainfall-interval treatments as a subplot factor.
For all questions, normality of residuals was visually checked by plotting residuals against tted values, and no violations of assumptions were found. Homogeneity of variance assumptions were examined by the "leveneTest" function in the "car" package and no heterogeneous variance was detected by Levene's test. Simple linear regressions were used to determine the relationship between %BNF and soil texture, OM, pH, and xation cofactors, holding %BNF as the dependent variable.

Results
Toposequence Experiment (Question 1) BNF contributed 77.1 ± 4.9% (standard error of the mean) to the N content of grain in summit positions and 36.5 ± 6.6% at toeslope positions (Fig. 2a). Backslope positions were intermediate with a BNF contribution of 62.6 ± 4.4%. Aboveground primary productivity was highest in toeslope positions (115.3 ± 13.5 g plant − 1 ) and lowest in summit positions (47.7 ± 8.1 g plant − 1 ) (Fig. 2b). We found no differences in BNF contributions to aboveground or belowground vegetative tissues (Supplementary table 1 and 2); only in seeds was there a BNF difference by landscape position effect, with a higher %N from BNF in summits. The %NPP from BNF showed no signi cant trend with landscape position (Fig. 2c).

Rainfall Intervals by Tillage Experiment (Question 3)
In 2015, in the ambient rainfall treatment BNF contributed from 66.1 ± 6.5% and 63.7 ± 2.9% to the grain N content in tilled and no-till plots, respectively. In soybeans experiencing 2-week rainfall intervals, BNF's contribution to the N content of grain was 68 .3 ± 4.4% and 68.9 ± 3.6% in tilled and no-till plots, respectively (Fig. 5).
In 2018, in the ambient rainfall treatment BNF contributed 81.0 ± 4.8% and 684 ± 5.1% to grain N content in tilled and no-till plots, respectively (Fig. 5). In soybeans experiencing 3-week rainfall intervals, BNF's contribution to the N content of grain was 66.8 ± 2.3% and 82.4 ± 1.0% in tilled plots and no-till plots, respectively (Fig. 5).. Soil Properties: Soil Texture, SOM, Mo, Fe, K, P, and pH The contribution of BNF to plant N content linearly increased with soil sand content and decreased with silt content (Fig. 6) and OM (Fig. 7). Percent sand was higher in summit positions and lower in toeslope positions (Fig. 6). Fe concentrations expressed the least range in concentrations from 0.8 to 5.1 mg kg − 1 .
Likewise, P differences ranged from 0.9 to 8.2 mg kg − 1 . K concentrations expressed the greatest range, from 26.3 to 141.4 mg kg − 1 . We found no relationship between BNF and soil Mo + 3 , Fe + 2 , P, or K concentrations or pH (Supplementary table 3).

Discussion
As hypothesized, the contribution of BNF to total plant N (% BNF) was highest in summit positions and lowest in toeslope positions, coincident with ner texture and organic matter contents at toeslope positions. Added rainfall at summit positions suppressed BNF. Tillage differences were less consistent: in 2015 neither tillage nor rainfall intervals affected %BNF, in 2018 %BNF was higher under conventional than no-till, and with longer rainfall intervals %BNF decreased under conventional tillage and increased under no-till.

BNF by Toposequence (Question 1)
How does soybean BNF vary by topographic position as affected by OM and its in uence on Nmineralization and water availability? Percent BNF along our toposequence ranged from 0-94%, re ecting ranges seen in the literature (Salvagiotti et al. 2008;Schipanski et al. 2010) and at a nearby site (Gelfand and Robertson, 2015). Nevertheless, we observed a signi cant effect of landscape position on BNF, correlated with soil texture and N-mineralization rates. %BNF was more than twice higher (77%) at summit positions where soils were coarser and N mineralization rates were lower (Fig. 2). Interestingly, however, when BNF contribution is scaled by grain yield there were no differences between landscape positions in the total amount of N r supplied through BNF (Fig. 2c).
The differential response of N 2 -xing and non-N 2 -xing soybeans to the fertility and textural gradients reveals the in uence of soil properties along toposequences on BNF across heterogeneous elds. Other studies have also revealed in situ soil properties that in uence BNF across individual elds. Riffkin et al. (1999) documented higher rates of BNF on sandier soils in Australia, as did Shipanski et al. (2010) in New York. Riffkin et al. (1999) found differences of 7% between light-and medium-textured soils.
We found no differences in %BNF in roots by landscape position, though this assumes that our 0-25 cm depth roots adequately represent roots throughout the pro le. Rooting depths at different landscape positions may have differed due either to water availably or low permeability soil layers caused by tillage or geology.
Whole plant %BNF re ects changes in all parts of the plant, though trends are especially prevalent in seeds since they are the biggest sink for N and have the highest concentrations (~ 6% N). Vegetative and belowground biomass have much lower N contents (~ 0.8% and ~ 1.3%, respectively) and %BNF (~ 4.5% and ~ 15%). Thus, the introduction of N r to the environment through soybean production is mostly associated with the grain.
Higher %BNF in summit positions is correlated with lower N mineralization rates (Fig. 3), congruent with other studies that note lower %BNF where mineralization rates are high (Schipanski et al. 2010). Soils with higher sand content tend to have less organic matter and available N in comparison to soils with more clay (Six et al. 2000), thus sandier soils would be expected to have lower N mineralization rates.
Comparing N assimilation between non-nodulating and N 2 -xing soybeans may indicate the ability of N 2xing soybeans to allocate carbon belowground and induce N mineralization in low fertility soils, such as those at summit positions (Schipanski 2010). We found soil N assimilation by both non-nodulating and N 2 -xing soybeans was similarly high in toeslope soils with their ner texture, evidenced by lower % BNF. George et al. (1993) found similar differences across an elevation and fertility gradient, with more soil N uptake in N 2 -xing soybeans compared to the non-nodulating plants at low soil N availability.
Roots obtain oxygen from pores in the bulk soil environment. Oxygen availability is an important regulator of nitrogenase activity; legume nodules can have four times the oxygen demand of an equal biomass of roots (Layzell and Hunt 1990). Soils with lower microbial respiration, then, may have more oxygen-rich environments capable of supporting a high number of nodules, which might promote more BNF. Thus, lower BNF at toeslope positions may also be due to lower oxygen availability insofar as soils with more clay and higher microbial activity have lower oxygen available (Layzell and Hunt 1990).

BNF with Changes in Rainfall Amounts by Landscape Position (Question 2)
Does BNF in summit and toeslope positions differ in response to added precipitation? In this experiment %BNF decreased ~ 30% with additional rainfall at summit positions, but not at toeslope positions. The most likely explanation for suppressed BNF with additional rainfall at summit positions is increased inorganic N supply. Inorganic N pools can suppress BNF (Schipanski et al. 2010), and it's likely that added water stimulated N mineralization, which in turn suppressed BNF. Toeslope positions, with their higher ambient water contents and N mineralization potentials (Fig. 3), may have likewise had BNF suppressed by soil mineral N pools.
Higher %BNF under drier (ambient) conditions in summit positions contrasts with studies that show nodule production, which is closely tied to BNF, generally decreases under drier conditions. Thus, one might have predicted additional rainfall to have increased rather than attenuated %BNF. N 2 xation is more sensitive to soil conditions than to plant stress (Abdelhamid et al. 2011); additionally by dry conditions that can lead to excess solutes in the root zone, restricting water availability to rhizobia (Walsh 1995). That BNF was not inhibited by drier ambient conditions is likely because summit soils were not su ciently dry: In an Illinois study, Gray et al. (2013) found that drought stress must be greater than 41% of the historical average to inhibit nodulation. Drought stress at KBS was ~ 15% in 2016 and nil in 2017 (https://lter.kbs.msu.edu/datatables/7).
The timing of dry conditions can also affect %BNF, which occurs differentially throughout plant stages (Gan et al. 2003). Furthermore, low nodulation response to rainfall variation could persist throughout the growing season despite temporarily improved soil moisture conditions and areas in the eld with greater water holding capacity. If dry conditions are severe enough to inhibit BNF stages where BNF rates are high, there will be lower total BNF values, thus scaling up BNF values from soybeans under drier conditions can potentially lead to an underestimation of BNF (Gelfand and Robertson 2015).

BNF with Changes in Rainfall Intervals under Different Tillage (Question 3)
Do changes in precipitation intensity in uence soybean BNF in tilled plots differently from no-tilled plots?
The same amounts of BNF occurred in tilled and no-till treatments under 3-day and 2-week rainfall interval treatments in 2015. However, when rainfall was excluded for three weeks in 2018, BNF decreased from 82 to 68 %BNF. This response fails to support the hypothesis that no-till management, with its higher organic matter content (Syswerda et al. 2011), will be better buffered against changes in rainfall intensity than will conventional tillage management. In fact, counter to expectations, BNF under conventional tillage increased under 3-week droughts from 68 to 80 %BNF, rather than decreased. Were the hypothesis supported, we would expect BNF under no-till to change little following drought and BNF in tilled soil to decrease.
There are several possible explanations for 1) BNF's being greater in no-till than in conventional till systems under ambient intervals; 2) decreasing %BNF in conventional till exposed to the longer dry interval; and 3) increasing %BNF in no-till soybeans exposed to the longer dry interval.
First, BNF could be higher in no-till than conventional due to higher N immobilization in no-till soils due to greater retention of organic matter, resulting in less available plant N following drought. This may be less of an issue in the ambient treatment because more consistent rainfall may have promoted more N mineralization; which is corroborated by past work of in situ assays in the same plots, which show higher net N mineralization in no till plots vs. conventionally tilled plots (Millar and Robertson 2015).
Second, biogeochemical processes crate a vertical gradient of 15 N through the soil pro le with a higher distribution of 15 N in the upper 10 cm than in the lower depths (Natelhoffer and Fry 1988). However, when soils are homogenized as they are in tilled plots, this disrupts the naturally occurring patterns of δ 15 N with soil depth in comparison to no-till plots. Since soybean roots are concentrated in the upper 15 cm (e.g., Böhm et al., 1977;Robertson et al., 1980), it is possible that soybeans in no-till systems may have different 15 N uptake patterns than soybeans in tilled plots because of associated δ 15 N patterns. If so, then this apparent difference would have been observed in 2015 as well, instead of just 2018.
Other explanations are also possible. These include the potential for differences in rhizobia populations ability to x N 2 as well as tillage-related compaction that may have limited rooting development in tilled plots. In tilled plots, compaction caused by repeated tillage could be di cult for roots to penetrate. Keisling et al. (1995) found soybean roots in no-till systems followed classical taproot trends, but when tillage pans were present, roots followed old root channels and pan fractures. There is also speculation that increased disturbance, like tillage, will decrease the presence of effective rhizobia, though this remains to be tested speci cally (Kiers et al. 2002). Trace element de ciencies could also contribute to lower %BNF, but there appear no trace element de ciencies in either conventional or no-till treatments (http://lter.kbs.msu.edu/datatables/354)

Validity of %BNF validity
Various soil and environmental factors can in uence the suitability of 15 N natural abundance to accurately re ect BNF rates, but three consistent outcomes in this study suggest that the values reported are robust. First, we found no signi cant relationships between %BNF and soil P, K, pH or soybean yield, which suggests that these factors were not in uencing BNF differentially in these elds during the experimental period. Second, calculated BNF values never exceeded 100%, which can occur when isolines do not appropriately represent N uptake under non-BNF conditions. And third, the N content in both nodulating and non-nodulating soybeans responded in the same direction, i.e., increases in the N content of nodulating soybeans were paralleled by increases in the N content of non-nodulating soybeans, further suggesting good correspondence between nodulating and non-nodulating isolines. These three observations warrant connecting the three studies and drawing overall conclusions.

Remaining Questions
Several lines of additional research could be useful for further understanding spatial patterns on BNF in the eld and responses to changing precipitation patterns. These include a better understanding of BNF with respect to 1) rhizobia strains; 2) other elements of global change than rainfall patterns; and 3) BNF rate differences in other legumes and throughout the growing season.
First, genetic variation among rhizobia strains may be sensitive to topographic position and tillage treatments. Rhizobia inoculants differ, and some are more effective at xing N 2 than others (Thilakarathna and Raizada 2017). We did not examine the distribution of strains, which are known to differ regionally (Batzli et al. 1992) and with management (Kiers et al. 2002). Weese et al. (2015), for example, found less-mutualistic rhizobia evolved in long-fertilized elds and likewise, there may be lessmutualistic rhizobia in toeslope positions where N mineralization is high. This may also be the case with different precipitation regimes, as we know that changes in regimes alter microbial communities (Zeglin et al. 2013). Certain rhizobia genes are needed to establish symbioses (Bottomley and Myrold 2015).
Unfortunately, it was not feasible to calculate separate B values for each eld or measure rhizobia strains or their effectiveness in different soils. Though we saw difference in spite of different strains, it would be valuable to understand differences in rhizobia strains at differing landscape positions, potentially justifying rhizobia inoculant for parts of elds.
Second, BNF appears sensitive to changes in both precipitation amounts and intervals. Understanding the full relationship between BNF and rainfall requires additional experimentation with different precipitation treatments, including both amounts and delivery patterns. For example, heavy rainfall events, particularly in the spring, can delay planting and lead to anerobic conditions that can limit rhizobia's ability to infect roots and thus lead to reduced nodulation (Layzell and Hunt 1990). Elevated

Implications And Conclusions
Changing precipitation patterns seem likely to in uence BNF in predictable ways depending on landscape position and tillage history. We found %BNF highest in summit positions and lowest in toeslope positions, coincident with coarser texture and less soil organic matter at summits. Increased precipitation diminished %BNF at summit positions, but not at toeslope positions. In no-till plots, subjected to longer (3-week) rainfall intervals, %BNF increased, while in conventionally tilled plants %BNF decreased. Results have several implications for global assessments of N xation and soybean N management.
First, results suggest that including information on landscape position and rainfall changes in calculations of eld-scale BNF rates could lead to more accurate estimates of landscape and regional contributions of BNF to N r . The creation of regional N r budgets for BNF are currently performed with values from small plots, likely more similar to toeslope positions than summits, insofar as most agricultural research is conducted on level ground of higher fertility (Robertson et al. 2007). This would imply that regional estimates of BNF from soybeans are likely underestimated.
Second, despite meta-analyses that show soybean responses to N fertilizer are rarely justi ed (Mourtzinis et al. 2017), N fertilizer use is promoted in some materials (e.g., Pioneer 2020). Results from rainfall and toposequence experiments suggest that soil organic matter is the best predictor of BNF in tilled elds (Fig. 6). With higher organic matter and more frequent rainfall in tilled elds, there are lower BNF rates.
Organic matter and irrigation can supply N through mineralization, thus managing soil organic matter and irrigation can provide N availably to soybeans, decreasing the necessity of applying fertilizer N to soybeans.
Third and nally, results suggest that management on a site-speci c basis would be helpful. Given increasing soybean acreage and yields in the US, in order to limit inputs, it will be practical to be e cient in inoculation, fertilizer, irrigation, and organic matter management. Understanding toposequence differences in soybean BNF provides the opportunity to manage inputs more e ciently by slope position.     Percent sand, silt and clay by %BNF. Error bars are omitted for clarity, n=4 replicate plots. Asterisks indicate signi cant effects (P<0.05) with soil texture