*4.4. Modeling the N Mineralization Rate*

Evaluation of six process-based models of European cropping systems led Yin et al. [76] to conclude that all of the models had difficulties predicting both the mean of and variance in soil N mineralization. It is clear that predicting Mn remains a challenge, due to the complexity of N cycling, which demonstrates why even the best models can explain only 65–80% of the variance [12,77]. Model 2, calibrated with our dataset, lies in this range. We also observed that a model parameterized only with soil properties and the I\_Sys indicator was less accurate than the "soil" model of Clivot et al. [12] (R<sup>2</sup> = 0.61). It confirms, in the present study, the utility of using mineralization indicators as input variables of models, obtained either by measuring EON [69] or by incubation measurements [5,77].

SMB is rarely used to predict mineralization because it is relatively labor intensive, and few laboratories measure it routinely. Clivot et al. [12] concluded that SMB improved prediction of mineralization little, if SON had already been included as an input variable; this can be explained by the weaker correlation between the mineralization rate and SMB in their study than that with SON (r = 0.54 vs. 0.62, respectively), even though both were highly significant. In contrast, Vnmean and SMB were more strongly correlated in our dataset than Vnmean and SON (r = 0.48 vs. 0.31, respectively), thus replacing SON with SMB increased the accuracy of the GAMs.

The I\_Sys indicator significantly increased the accuracy of mineralization predictions and represents an original contribution of this study. The selection of I\_Sys in the final GAM confirms the influence, in the medium term, of crop rotations and the application of organic waste on mineralization, while also providing information that complements that provided by SMB and EON. The lack of correlation between I\_Sys and these two variables leads to the hypothesis that EON is relatively insensitive to the cropping history over the medium term but is strongly influenced by the geochemical background and the cropping history over the long term, as shown by its strong correlation with SON (r = 0.64).

#### **5. Conclusions**

The experimental design of this study, based on repeatedly measuring Mn for three consecutive years after two years of unfertilized maize, ultimately placed it in the best possible conditions for estimating mineralization of SON. The measurements confirmed the high variability in mineralization, which lay near the top of the range reported in the literature. It can be explained by the combined effects of soil types, with variable but generally high SON content; cropping systems representative of livestock-production regions, with regular application of organic waste and the frequent presence of grassland in rotations; and the semi-oceanic climate, which favors mineralization.

The drivers of mineralization were those identified in the literature: SON, POM-N, SMB, EON, and soil texture. Original results of this study include (i) experimental evidence of the influence of cropping history in the medium term on mineralization, and (ii) the fact that the I\_Sys indicator provides information complementary to EON. This result strengthens the hypothesis of Ros et al. [31] that mineralizable N cannot be predicted from a single soil test alone, but instead requires a combination of components including EON and site-specific information, such as land-use and soil properties.

The modeling approach identified the most influential measured variables and showed that the proportion of variance explained by a model based only on basic soil properties and the I\_Sys indicator (R2 = 0.47) was not sufficient to consider the model operational. The accuracy of the model increased greatly when SMB and EON were included, but the proportion of the variance explained by model 2 remained relatively moderate, despite the supplemental information provided by the model's covariates: basic soil properties, chemical and biological indicators, and a land-use indicator.

**Author Contributions:** Conceptualization, T.M. and Y.L.; methodology, T.M., Y.L., L.B., B.M. and N.B.; software, L.B.; validation, T.M., Y.L. and L.B.; formal analysis, T.M., P.G., B.L. and L.B.; investigation, T.M. and L.B.; resources, T.M. and L.B.; data curation, T.M., Y.L., L.B., B.M. and N.B.; writing—original draft preparation, T.M. and L.B.; writing—review and editing, T.M.; visualization, T.M. and L.B.; supervision, T.M.; project administration, T.M. and Y.L.; funding acquisition, T.M. and Y.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was financed by the Loire Bretagne Water Agency; the Regional Council of Brittany; the Departmental Councils of Côtes d'Armor, Finistère, and Morbihan; and the French government (DRAAF).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data analyzed in this article were published by *Data in Brief*: https://doi.org/10.1016/j.dib.2021.106795 (accessed on 1 April 2021). Raw data are deposited in the public repository Data INRAE. Data identification number: 10.15454/VYEYBK. Direct URL to data: https://doi.org/10.15454/VYEYBK (accessed on 19 January 2021).

**Acknowledgments:** This study was supported by partnerships with the SEMSE, AUREA, and Galys laboratories. The authors thank Michael Corson for proofreading the manuscript's English.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**

