How does Soil Organic Matter affect Potato Productivity on Sandy Soil? Evidence from a Greenhouse Study

Soil organic matter (SOM) can contribute to crop productivity through nutrient availability and soil improvement. However, SOM is typically low (< 2%) in the coarse-textured soils used for potato production. The potato cropping system is irrigated, frequently tilled, and fertilized causing potential nitrate leaching and contamination. The effect of varying levels of SOM on potato productivity in sandy soil is unclear. This study aimed to estimate the effect of varying SOM on potato productivity in sandy soils and to understand if nitrogen (N) mineralization was a primary mediator of this effect. Soil from nine �elds in Wisconsin, USA (SOM range of 1.1 to 3.8%) were collected for a greenhouse study. Red Norland was the variety planted and vine, tuber, and total biomass and N uptake were measured. In-situ ion exchange resin strips and potentially mineralizable N (PMN) measured at harvest were used as proxies for N mineralization. We found that SOM had a positive effect on plant productivity. The effect was statistically signi�cant for four productivity metrics (fresh matter whole biomass, dry matter vine biomass, and total N uptake in the vines and whole biomass) and marginally signi�cant for four more. We found that N mineralization (as PMN) was a partial mediator of SOM effect on productivity doing a formal mediation test. Hence, it is likely that SOM improved plant productivity through mechanisms beyond just N acquisition by plants. Our results suggest future efforts should explore other mechanisms through which SOM can affect productivity.


Introduction
Soil organic matter (SOM) can be a source of fertility for crop plants by releasing nutrients via mineralization of organically bound plant nutrients into plant usable forms (Henis 1986; Manlay et al. 2007).SOM can also contribute to crop productivity through other mechanisms such as nutrient retention via cation exchange capacity (CEC), aggregate formation and increased moisture retention and aeration (Krull et al. 2004;Murphy 2015).SOM also provides the substrate and habitat for soil organisms which can bene t other soil ecological properties, like soil-borne disease suppression (Brady and Weil 2008;Cooperband 2002;Fageria 2012).Thus, SOM can increase crop productivity through N availability or by improving other soil physio-chemical and biological characteristics (Daly et al. 2021; Kästner et al. 2021;Lehmann and Schroth 2002).A trend towards production practices that rely on internal biological cycling of nutrients instead of chemical inputs has generated an increasing interest in the topic (Bouwman et al. 2013; Goulding et al. 2008; House and Brust 1989; Swift et al. 2004).Hence, increasing SOM in cropping systems has the potentil to reduce the reliance on external N fertilizer inputs (Foley et al. 2011).
Potato production necessitates substantial soil disturbance due to the need for planting, hilling, and harvesting of this below-ground crop (D' Haene et al. 2008;Rasmussen et al. 1999).In addition, potatoes in Wisconsin (WI) are mostly grown under irrigation on sandy soils which are low in SOM.Similarly, the potato-cropping system in WI is frequently tilled, fertilized, and fumigated.Potato is an annual plant with less root biomass compared to most other annual crops commonly grown in the US (Lesczynski and Tanner 1976); in general, less root biomass can lead to a reduction in overall SOM and may reduce nutrient availability over time (Grace et al. 1995;Larson et al. 1972).Furthermore, crop residues decompose faster in sandy soils compared to ner-textured soils.Hence, N released from the residue after harvest may be leached before it can be utilized by the subsequent crop, resulting in a decrease in available N for the crops (Wolkowski et al. 1995).Thus, it is unclear if the inherently low amounts of SOM can provide the necessary agronomic bene ts (yield increase and soil health bene ts) on sandy soils as seen in richer soils.
Previous global and regional studies on maize (Old eld et al. 2019), wheat (Old eld et al. 2020), and corn and soybean (Kravchenko and Bullock 2000) have shown that small differences in SOM at low levels (≤ 4%) have a greater impact on crop productivity compared to large differences at high levels (≥ 4%).However, there is a lack of quantitative evidence on the effect of SOM on crop yields on sandy soils worldwide (Loveland and Webb 2003).Although many studies make reference to the positive effects of SOM on crop yields (Bauer and Black 1994), there is almost no quantitative information on the effect of the increase in SOM on potato productivity on sandy soil.Such data is essential because sandy soils cover a vast area (900 million ha) worldwide and are under extensive cultivation (Yost and Hartemink 2019).Our investigation aims to provide insight into the importance of SOM on sandy soil for potato production.
Fields vary in many ways that can confound the effects of SOM on productivity, including varying climate and management practices.Thus, evaluating the effects of varying SOM on crop productivity requires an experimental approach that isolates the effects of SOM, such as through a controlled greenhouse experiment.The interactions of SOM with other soil properties are complex (Grigal and Vance 2000;Luo et al. 2017;Oades et al. 1989), but greenhouse experiments allow for the control of non-soil environmental factors and other eld characteristics.The use of natural soils collected from potato elds that vary in SOM allowed for investigation of the effects of naturally produced SOM on productivity, without the potential caveats of manipulating SOM levels using soil organic inputs in the greenhouse.
The rst objective of the study was to investigate the relationship between SOM on sandy soil and potato productivity with the hypothesis that plant productivity will increase with an increase in SOM.The second objective was to determine if the effect of SOM on potato productivity was mediated primarily through N mineralization.We applied the conceptual mediation testing approach proposed by Baron and Kenny (1986) to identify net effects (objective 1) and to distinguish between the direct and indirect effects (objective 2) of SOM on productivity.

Experimental design
To quantify the SOM -potato productivity relationship, we collected soils from nine elds equally distributed among three farms across WI.The elds were managed under similar environmental and climatic conditions, all with a history of potato production but with various levels of SOM (Table 1).The elds were selected based on prior experimental studies and the SOM level was con rmed later through laboratory analysis.In each eld, the soils were collected using a shovel from 0-30 cm depth from multiple randomly chosen locations to cover the whole eld.Soils were transported back to our greenhouse complex at the University of Wisconsin-Madison, where they were homogenized and hand sorted to remove large stones (> 5 mm diameter), macro fauna, and large roots (> 2 mm width) (Fig. 1).All soil Table 1 Locations (latitude/longitude) of nine elds in the study used to collect the soil samples from, their measured values of selected soil characteristics: soil organic matter (SOM (%), pH, sand, silt, and clay (%), and soil nutrients: calcium (Ca), magnesium (Mg), potassium (K), phosphorus (P), sodium (Na), copper (Cu), zinc (Zn), iron (Fe), and manganese (Mn) We had one pot per eld within each experimental block.As we had soils from nine elds and three blocks, we had a total of 27 experimental pots arranged in a randomized complete block design.All the pots were fertilized with phosphorus (P), potassium (K), sulfur (S), and calcium (Ca) to avoid the limitation of any of these essential nutrients to the plants.N  respectively.The temperature inside the greenhouse was controlled using vents, fan coil units (FCU), sensor fan, and exterior fan.The lights inside the greenhouse had sensors that turns them off when the available sunlight reached the threshold.The photon threshold was detected by the sensor on the roof of the greenhouse.

Experimental measures
Before planting, we measured SOM content via loss on ignition (Nelson and Sommers 1996) for soil from each eld.Nineteen additional soil physical, chemical, and biological properties were measured either pre-planting or at harvest at the Kansas State Soil Testing laboratory and the University of Wisconsin Soil and Forage laboratory to control for the potential confounding soil properties in the study (Table 2).
Both NH 4 -N and NO 3 -N (extracted from ion exchange resin strips four times during the greenhouse study) and potentially mineralizable nitrogen (PMN) (measured at harvest) were used as the proxies for nitrogen mineralization.NH 4 -N and NO 3 -N values were measured during 10-17 days after planting (dap), 32-39 dap, 60-67 dap, and 109-116 dap using 2 cm X 6 cm ion exchange resin strips following the method by Qian and Schoenau (1996).The strips were cleaned with acid (5% HCl) and then charged with 1 M NaCl before deploying them in the pots.After extracting the strips out of the pots, NO 3 -N and NH 4 -N were extracted using 1 M KCl from anion and cation strips respectively and were analyzed using a ow injection  (Knepel 2003).PMN (mg kg − 1 ) was measured using 7-d anaerobic lab incubation method (Waring and Bremner 1964) and was calculated as an average of the two replications of NH 4 -N obtained after seven days incubation minus initial NH 4 -N at time zero times the extraction ratio (Goodale and Aber 2001).The extraction ratio was calculated as the ratio of the volume of extract solution (50 ml) divided by the measured weight of soil samples (5.0 ± /0.01 g).
Vine biomass was harvested on 7 Jun 2022 and the tubers were left in the pots for 12 more days before the nal harvest.After measuring fresh weights (g), all the plant biomass was dried to constant mass at 70°C, and then weighed to achieve dry matter biomass (g).The dry matter was ground and was analyzed for total N (%) and total C (%) analysis at the Kansas State Soil Testing Laboratory using a dry combustion-based Dumas method (Bremner 1965).Total N uptake was calculated by multiplying dry matter biomass and total N concentrations for each plant tissue.Whole plant biomass was calculated by adding vine, tuber, and root biomass altogether.Fresh as well as dry matter vine, tuber, and whole biomass, total N contents and total N uptake in the dry matter vine, tuber, and whole biomass, and tuber numbers were used separately as distinct metrics of plant productivity.

Statistical analysis
Our statistical approach occured across four steps.The rst step (Step 1) was conducted as part of Objective 1 and Steps 2-4 were conducted as part of Objective 2.

Objective 1: investigate the relationship between SOM and potato productivity
We used linear mixed effect models to investigate the effect of SOM on productivity.Since we had a small sample size and many independent soil variables, it was not possible to include all the independent soil variables in one model when testing their effect on productivity.Firstly, to avoid over-tting our models, a few soil variables among all were removed from any further analysis based on prior understanding of the relationship between SOM and those soil variables.Secondly, to avoid the issue of multicollinearity while selecting the soil variables in the model, we determined the Pearson or Spearman rank correlation between SOM and measured soil variables and among the measured soil variables using the cor function in the Hmisc package in R (Fig. 2).Finally, after considering these correlations, we were left with a set of uncorrelated soil variables to use in the linear mixed effect models.We then subjected these full models (one for each plant productivity metric) to a stepwise backwards selection based on the Akaike's Information Criterion (AIC) using the AIC function in the car package in R (Burnham and Anderson 2004; Frey 2019) (Fig. 2).
Step 1 Once an optimal linear mixed effect model was identi ed for each plant productivity metric, we performed a regression analysis estimating the effect of SOM on productivity with SOM as a continuous xed effect and eld identity as a random effect while controlling for other potentially confounding soil properties.We used eld identity as a random effect to account for the non-independence of the three replicate pots given the same soil source, and to account for differences in potato productivity among the nine soil sources that were not related to linear gradients in the measured soil variables.If SOM and any of the productivity metrics were positively associated with each other, then we concluded that SOM had a positive effect on potato growth and development.

Objective 2: determine if N mineralization was a mediator of the effects of SOM on productivity
We investigated the potential for N mineralization to mediate the effect of SOM on plant productivity through an approach which tests the signi cance of the coe cients in a series of simple and multiple regression analyses conducted in several steps (Judd and Kenny 1981;Wright 1921).Given that our study had a limited sample size (n = 27), we chose to perform a simpler mediation test to determine the indirect effect of SOM on productivity via N minealization (Pearl 2009;Westland 2010).
If the effect of SOM on any of the productivity metrics (step 1) was found to be signi cant at p < 0.05 level, subsequent steps (steps 2-4) were taken to investigate the role of N mineralization as a mediator of the effect (Fig. 3).We used the same model covariates and random effects as the models described above.
Step 2 We performed regression analysis testing the relationship between N mineralization and SOM.The test was conducted separately for all the proxies of N mineralization.If SOM promoted productivity primarily through N mineralization, then there must be a substantial positive relationship (p < 0.05) between SOM and at least one of the proxies of N mineralization (NH 4 and NO 3 or PMN).
Step 3. We performed linear regression for each productivity metrics that were positively affected by SOM from step 1 (as dependent variables) with each N mineralization proxy that had a substantial Fig. 3A owchart for determining if N mineralization mediates the effects of soil organic matter (SOM) rates on potato productivity positive relationship with SOM from step 2 (as independent variables).To mediate the effect of SOM on productivity, at least one of the proxies of N mineralization must have a direct effect on at least one of the productivity metrics.If there was a signi cantly positive (p < 0.05) relationship between any of these plant productivity metrics and any of these proxies of N mineralization then step 4 was investigated.
Step 4 Finally, we conducted a multiple regression analysis testing the effect of both SOM and N mineralization on plant productivity.The change in the estimated coe cient for the SOM effect on productivity in this multiple regression, compared to the coe cient of the SOM effect in model without N mineralization (e.g.step 1), provides a quantative measure of the degree to which the SOM effect is mediated by N mineralization.For instance, if the proxy of N mineralization had a signi cant positive effect whereas the SOM effect on the productivity metric was greatly reduced (near zero) and statistically non-signi cant in the multiple regression, then N mineralization was considered the primary mediator of SOM to productivity.If the SOM effect was reduced, but not eliminated, in the multiple regression, this indicated that N mineralization was a partial mediator of the SOM effect.Whereas, if the SOM term retained a signi cant positive effect (p < 0.05) and was similar in magnitude in the multiple regression compared to its effect in the original model, then we concluded that the effect of SOM on productivity occurred primarily through mechanisms independent of N mineralization.

Objective 1: investigate the relationship between SOM and potato productivity
SOM content ranged from 1.1-3.8%and soil pH ranged from 4.3 to 6.3 (Table 1).The observed soil texture were loamy sands and sandy loam.The summary statistics of the samples in the study for measured plant productivity (biomass and N uptake) are presented in Table 3.The soil variables that were initially removed from the analysis based on prior studies were soil total N, total OC, and CEC measured pre-planting.Similarly, the soil macro and micronutrients that had a strong correlation with SOM were Fe, Ca, Mg, P, and K (Table 4).Hence, these soil variables were also not used in the models with SOM.Furthermore, the variables that had a strong negative correlation with one another were % sand and % silt, soil Na and pH; and % sand and soil Cu (Table 4), whereas soil Zn and Mn were positively correlated with each other and negatively correlated with soil pH (Table 4).Based on these intercorrelations between variables, we established our initial, full model as including three soil variables: SOM, soil pH, and % sand.By dropping each of the covariates (% sand and soil pH) one at a time, the optimal model based on lowest AIC included just SOM and soil pH.The same optimal model was selected for each of all metrics of plant productivity.
SOM had a positive effect on all the 10 plant productivity metrics; this effect was statistically signi cant (p < 0.05) for four of the metrics (fresh matter whole biomass, dry matter vine biomass, and total N uptake in the vines and whole biomass) (Fig. 4), marginally signi cant (p < 0.10) for four more (fresh matter tuber and vine biomass, dry matter whole biomass and tuber numbers), and neither statistically nor marginally signi cant for two of the metrics (dry matter tuber biomass, total N uptake in the tuber biomass) (Table 5).Our models indicate that fresh matter whole biomass and dry matter vine biomass were 1.7 and 1.6 times greater at the maximum SOM (3.8%) compared to the minimum SOM (1.1%) at a soil pH of 6.5.Similarly, total N uptake in the whole and vine biomass were 1.7 and 2.1 times higher at maximum SOM (3.8%) than minimum SOM (1.1%) at a soil pH of 6.5.Soil pH had a strong negative effect on all plant productivity metrics except tuber numbers (Table 5).

Objective 2: determine if N mineralization was a mediator of the effects of SOM on productivity
Since four measures of productivity (fresh matter whole biomass, dry matter vine biomass, and total N uptake in the vine and whole biomass) had a statistically signi cant net positive relationship with SOM (Table 5), we proceeded with the additional steps of mediation analyses to determine if N mineralization was a full or partial mediator of the SOM effect.In step 2, SOM had a positive relationship with PMN (Fig. 5); however, it did not substantially affect Fig. 5 The relationship between soil organic matter (%) and potentially mineralizable N (PMN) (mg kg − 1 ).PMN was used as a proxy for N mineralization in the study soil NO 3 -N and NH 4 -N measured four times throughout the growing season using in-situ ion-exchange resin strips.In step 3, PMN had a positive relationship with total N uptake in the vine biomass (Fig. 6); however, it did not substantially affect the other plant productivity metrics (Table 6).Multiple regression analysis between SOM and PMN on productivity (step 4) found a non-signi cant effect of both SOM and PMN on productivity when included in the same model (Table 7).The estimates predicting the effects of SOM on productivity decreased with PMN compared to those from models without PMN for N uptake in the vines (Table 6 versus Table 7).This indicated that PMN was a Fig. 6 The relationship between potentially mineralizable N (PMN) (mg kg -1 ) and total N uptake in the vine biomass (g N plant -1 ).PMN was used as a proxy for N mineralization and N uptake in the biomass was one of the metrics of plant productivity in the study partial mediator of SOM effects on productivity measured as the total N uptake in the vine biomass.But, for the majority of our productivity measures, including PMN in models with SOM did not substantially reduce the estimated SOM effect on productivity (Table 4 versus Table 5).Our study did not show that SOM had a signi cant effect on tuber yields or N uptake in the tubers (p < 0.1) which suggests that the increase in OM in sandy soils within the range (1.1% − 3.8%) increased aboveground more than belowground potato production.Our potato plants were grown in a limited space (11.4 L) pot in the greenhouse which may have hindered the growth and development of belowground tubers.Another explanation could be that N available from SOM could only contribute to vine growth during the reproductive phase.In addition to the overall positive relationship between SOM and productivity, we found a positive association between SOM and N mineralization (measured as PMN), which was consistent with previous research by Old eld et al. (2020).However, in-situ measurements of N mineralization (measured as NH 4 -N and NO 3 -N concentrations) using ion exchange resin strips did not show a positive association with SOM.This may be because the regular watering of the experimental pots with plants and the limited surface area of the strips (2 cm X 6 cm) in the pots allowed for plant uptake and leaching of N to occur, making it di cult to accurately capture the rate of N mineralization.
We found a positive relationship between PMN and only one of the productivity metrics (measured as N uptake in the vines).In contrast to most claims that N mineralization increases productivity, as reported by Haase et al. (2007) on potatoes and Abbasi and Khizar (2012) on maize.It is also possible that ion exchange resin strips and PMN as the a 7-day anaerobic incubation may not be the best proxies for net N mineralization from the soil.We would suggest expanding the suite of N mineralizaton analysis in future work.
Although we found a positive relationship between PMN and N uptake in the vines, plants in soils with higher PMN did not tend to have higher vine, tuber, or whole biomass.One explanation could be that the increased N supply to the aboveground biomass from SOM did not translate to greater N below ground, which is subsequently translocated from the vines later in the growing season.Hence, emphasizing on the importance of fertilizing the plants with mineral N supply beyond what SOM can provide particularly towards later in the growing season for Red Norland potato variety.
The simple regression analysis showed that the direct effect of SOM on fresh vine biomass and N uptake in the vines was marginally or fully signi cant, respectively (Table 5).However, when PMN was added to the multiple regression model with SOM, the effect size of SOM on fresh vine biomass and N uptake in the vine biomass weakened by 77% and 25%, respectively, and became non-signi cant (Table 5, Table 7).This suggests that SOM affects productivity partially through PMN, but that much of the effect of SOM could not be explained via our measures of N mineralization.Moreover, for the other metrics of productivity, controlling for PMN in the multiple regression model with SOM did not result in a substantial weakening of the estimated SOM effect.Hence, our results indicate that SOM affects productivity through mechansims beyond just supplying N, including those involving soil physical, chemical, and biological characteristics, as suggested by Carter et al. (2004) and Gomiero et al. (2011).Therefore, future research should focus on identifying other speci c processes that underlie this effect.Figure 7A owchart of independent and dependent variables, model outputs, and signi cance in linear mixed effect models investigating the relationship between soil organic matter rates, nitrogen mineralization, and potato productivity

Conclusions
Our greenhouse study supports the importance of SOM within the range of 0-4% in sandy soils for potato productivity.Our study evaluated the plethora of variables acting together that could affect potato productivity or could confound the effects of SOM on productivity while controlling for the confounding variables in the models.Hence, the results contributed to a clearer and overall understanding of the importance of OM in the small range on potato productivity in sandy soils.Additionally, our study suggests that SOM improved plant productivity through mechanisms beyond just N acquisition by plants.However, long term global and regional studies are needed to explore the potential effect of organic inputs and residue management towards building SOM in sandy soils, and any consequences for fertility management in these systems.Similarly, further research into the soil physical, chemical, and biological properties of SOM in sandy soils should be conducted to get a better understanding of the drivers of SOM to productivity.This will provide more evidence through experimental studies towards building soil health for agronomic outcomes and overall agricultural sustainability.

Declarations Figures
Page 20/26  A owchart for determining if N mineralization mediates the effects of soil organic matter (SOM) rates on potato productivity The effects of soil organic matter (%) on fresh matter whole biomass (a), dry matter vine biomass (b), TN uptake in the vine biomass (c), and TN uptake in the whole biomass (d).Soil organic matter (%) plotted against the residual output of the regression of plant productivity metrics with soil pH The relationship between potentially mineralizable N (PMN) (mg kg -1 ) and total N uptake in the vine biomass (g N plant -1 ).PMN was used as a proxy for N mineralization and N uptake in the biomass was one of the metrics of plant productivity in the study

Figure 3
Figure 3 fertilizer was not added to any of the experimental pots.Red Norland potato seed minitubers were planted on 22 February 2022 and harvested after 147 days on 19 July 2022.Minitubers were provided by the Wisconsin Seed Potato Certi cation Program.All the plants were watered with equal volume (0.4 L) every two to three days.To promote emergence and tuberization, the plants were grown under a photon threshold of 600 µmol from 22 Feb 2022 to 31 Mar 2022 and under 200 µmol from 1 Apr 2022 to 19 July 2022 and at a day and a night temperature of 24 o C and 21 o C

Table 2
Measured soil physical, chemical, and biological properties and their respective methods of analysis, units, and references for the methods in the study

Table 3
General summary of fresh and dry matter biomass, tuber numbers, and total N uptake of the plant productivity metrics measured at harvest

Table 4
Correlation coe cients between the measured soil properties ranging from 1 to -1 where positive and negative values represent the positive and negative correlations respectively.Ca = calcium, Cu = copper, Fe = iron, K = potassium, Mg = magnesium, Mn = manganese, Na = sodium, P = phosphorus, and Zn = zinc Correlation coe cients in bold indicates signi cant correlations (p<0.05)

Table 5
Summary of the models estimating the effect sizes of soil organic matter (SOM) and pH on plant productivity metrics with independent variables, their estimates, standard errors (std.errs), degrees of freedom (df), and p-values Model variables in bold and italics indicate signi cant effect sizes at p<0.05 and p<0.1 respectively

Table 6
Summary of the models estimating the effect size of potentially mineralizable nitrogen (PMN) and pH on plant productivity metrics with independent variables, their estimates, standard errors (std.errs),degrees of freedom (df), and p-valuesOur study support the claims of previous regional and global analyses conducted in both controlled greenhouse and eld conditions that showed a positive association between SOM and crop productivity (Old eld et al., 2019, 2020, 2022; Wade et al., 2020).A previous greenhouse experiment with SOM ranging from 1 to 9% found that increasing OM in soils up to a threshold of 5% led to greater aboveground biomass for spring wheat (Old eld et al., 2020), supporting the claim of our study.Our study nds a substantial increase in aboveground plant productivity as SOM levels increased from 1.1 to 3.8% suggesting the importance of SOM on productivity in this low range.
Model variables in bold indicate signi cant effect sizes at p<0.05

Table 7
Summary of the models estimating the effect size of potentially mineralizable nitrogen (PMN), soil organic matter (SOM), and pH on plant productivity metrics with independent variables, their estimates, standard errors (std.errs), degrees of freedom (df), and p-values