Silicon (Si) is increasingly recognized as a beneficial element for crop growth, particularly under biotic and abiotic stress. Its bioavailability in soil, however, is not well understood, especially in temperate climates. The uptake of Si by plants, primarily in the form of silicic acid, influences crop yields, particularly for Si-accumulating plants like rice, sugarcane, and wheat. While chemical weathering and leaching processes can deplete soil Si, agricultural practices, including crop harvesting and residue management, also modify the Si pool. This study aimed to quantify the spatial variability of plant-available silicon (PAS) in France, a region with diverse soil types representative of many temperate regions globally. The research focused on identifying the soil characteristics controlling PAS in both cultivated and non-cultivated soils, and on determining the prevalence of potential Si deficiency in wheat-growing areas of France. Understanding the influence of agriculture on PAS is critical for optimizing crop production and potentially mitigating yield stagnation in Si-demanding crops. The researchers utilized a spatial statistical approach, employing an extensive dataset from the French Soil Quality Monitoring Network to analyze the relationship between PAS, soil properties, and land use. The study's geographical scope and data richness offer valuable insights into the impact of agricultural practices on Si bioavailability in temperate soils, with implications for sustainable agriculture and crop improvement.

Existing research demonstrates the beneficial effects of silicon (Si) on crop yields and stress tolerance (Coskun et al., 2019; Datnoff et al., 2001; Epstein, 1994; Liang et al., 2007; Meunier et al., 2017). However, the role of Si remains debated (Exley & Guerriero, 2019). Soil Si bioavailability is a key factor influencing plant Si uptake and subsequent benefits (Liang et al., 2015; Guntzer et al., 2012; Hodson et al., 2005; Orazem, 2015). Several studies highlight the potential for Si fertilization to enhance yields in Si-depleted soils (Guntzer et al., 2012; Yan et al., 2018). The sources of plant-available Si include the dissolution of primary and secondary silicate minerals and the recycling of phytoliths from plant residues (Alexandre et al., 1997; Derry & Kurtz, 2005). Factors influencing Si bioavailability include mineral solubility and kinetics (Berner & Berner, 1996; Hellmann, 2005), plant uptake, microbial uptake (Riotto et al., 2020), and Si adsorption onto mineral surfaces (Saueur et al., 2007). Measuring plant-available Si remains a challenge (Hiemstra et al., 2014), with various extraction methods used as proxies (Tubau et al., 2007; Amphorn, 2015). Previous research has shown positive correlations between PAS and soil properties such as pH, <2 µm fraction, organic matter, and iron oxides (Lieutier, 2020; Meunier et al., 2018; Manson et al., 2014; Phonda et al., 2014; Yan et al., 2016; Landis et al., 2020; Narayan et al., 2019; Babu et al., 2016; Kordomskiy et al., 2015). Conversely, negative correlations with total Si have also been reported, reflecting the influence of low-solubility minerals like quartz (Landis et al., 2020). Critical PAS levels have been identified for some crops in tropical climates (Haynes et al., 1973; Clymans et al., 2011; Despagne et al., 2006; Datnoff et al., 2006; Meunier et al., 2008; Guntzer et al., 2012; Watanabe et al., 2007; Barrio et al., 2020). Long-term intensive cultivation has been shown to decrease PAS (Darmanto et al., 2006; Savant et al., 1997).

This research utilized data from the French Soil Quality Monitoring Network (RMQS), comprising over 2200 sites across France. The dataset included various soil properties, including plant-available silicon (PAS) measured using the 0.01 M CaCl₂ extraction method (Sicaciz), soil pH, <2 µm fraction content, cation exchange capacity (CEC), organic carbon content, and iron oxide content. Land use (cultivated vs. non-cultivated) and parent material were also recorded. The study employed a two-pronged approach. First, a digital soil mapping (DSM) technique was applied using a regression kriging (RK) model within the scorpan framework (McBratney et al., 2003). This involved selecting relevant spatial covariates (soil type, parent material, climate, land use, vegetation indices derived from NDVI data) using the Boruta algorithm (Kursa & Rudnicki, 2010) to predict the spatial distribution of Sicaciz. The model's accuracy was assessed via 30-fold cross-validation, using metrics such as R², RMSE, concordance, and bias. Second, statistical analyses were performed to explore the relationships between Sicaciz and soil properties in both cultivated and non-cultivated soils. Correlations were calculated, and linear models were used to determine the variance explained by land use. The impact of pH on Sicaciz was investigated in relation to the <2 µm fraction content. The CEC of the <2 µm fraction was used as a proxy for clay mineral composition to evaluate its influence on Sicaciz. Finally, potential Sicaciz deficiency for wheat was evaluated using critical levels established for sugarcane and rice (Haysom & Chapman, 1973; Babu et al., 1999) as reference values for temperate soils. A subset of the data (1986 points) was used for the primary analyses due to data completeness requirements. Additional data processing involved imputation of missing values for total Si concentration based on a cubist model (Landré et al., 2018). The uncertainty associated with the Sicaciz measurements was also estimated (Equation 3).

The study revealed a substantial range of Sicaciz concentrations in French topsoil (2.3–134 mg kg⁻¹), with a median of 17 mg kg⁻¹. The spatial distribution of Sicaciz was successfully mapped using the DSM approach (R² = 0.43), with parent material being the most important predictor variable. In non-cultivated soils, Sicaciz showed strong positive correlations with pH (r = 0.46) and the <2 µm fraction (r = 0.59), with weaker correlations with organic carbon and iron oxides. However, the relationship between Sicaciz and pH was dependent on the <2 µm fraction content. A clear positive relationship between Sicaciz and pH was observed only for soils with <2 µm fractions ranging from 50 to 325 g kg⁻¹. For soils with <2 µm fractions below 50 g kg⁻¹ or above 325 g kg⁻¹, the relationship weakened or disappeared. Analysis also indicated a positive correlation between Sicaciz and the <2 µm CEC, suggesting an influence of clay mineralogy. Importantly, cultivated soils showed significantly higher Sicaciz concentrations than non-cultivated soils for those developed from sediment parent material (73% of French soils), but not for soils derived from metamorphic or igneous intrusive rocks. This increase in Sicaciz under cultivation is attributed to the increase in pH resulting from liming practices, particularly for non-carbonated soils on sediments. The study indicates that approximately 4% of French soils used for wheat cultivation might exhibit Sicaciz levels below the lower critical threshold established for sugarcane (20 mg kg⁻¹), suggesting potential Si deficiency in these areas. This estimate was based on the critical values determined for sugarcane and rice, highlighting the need for further research to establish definitive critical levels for wheat in temperate climates. The observed changes in clay mineral composition (increase in smectite/vermiculite content under cultivation) also support the influence of agricultural practices on soil Si bioavailability.

The findings of this study demonstrate that agricultural practices significantly influence the bioavailability of silicon in temperate soils. The observed increase in Sicaciz in cultivated soils developed on sediments, primarily due to liming and consequent pH increase, suggests that soil management practices can enhance Si availability for crops. However, the lack of a significant effect of cultivation on Sicaciz in soils derived from metamorphic or igneous rocks points to the importance of considering parent material in assessing Si bioavailability. The finding that only a small percentage (4%) of wheat-growing soils appear deficient in Sicaciz, based on the critical levels for other crops, needs further investigation to confirm the applicability of those critical values to wheat in temperate systems. Future research should focus on establishing precise critical Sicaciz levels for wheat in temperate climates and on exploring the long-term impacts of different soil management practices on Si dynamics. Additionally, the underlying mechanisms of Si release and transformations in various soil types and under different cultivation practices warrant further exploration. Further research incorporating kinetic studies and chronosequences would significantly enhance our understanding of these complex interactions. The study highlights the importance of integrating soil properties, land use, and parent material into models predicting Si bioavailability, especially in regions with diverse soil types and varying agricultural practices.

This study provides valuable insights into the spatial variability and determinants of plant-available silicon (PAS) in temperate soils. Agriculture, specifically liming practices, was shown to enhance PAS in soils developed on sediments. However, the influence of cultivation was less pronounced in soils with different parent materials. A small percentage of wheat-growing soils may be Si-deficient, but further research is necessary to determine precise critical levels for wheat. Future research directions should focus on determining accurate critical levels for wheat, investigating long-term impacts of cultivation, and clarifying the complex interplay of soil properties, management practices, and Si dynamics.

The study relied on data from a single sampling campaign (2000–2009) from the RMQS network; temporal variability in PAS was not directly assessed. The critical levels for Si used in the deficiency assessment were derived from studies on rice and sugarcane, not wheat, under temperate conditions. Therefore, the estimate of Si-deficient soils for wheat might not be perfectly accurate. Some data imputation was necessary for total Si concentration, which might have introduced some uncertainty. Finally, the study did not directly address the kinetic aspects of Si transformations in the soil.