Interim report of the repeated German agricultural soil inventory

Poeplau, Christopher et al. Working Paper Interim report of the repeated German agricultural soil inventory Thünen Working Paper, No. 277a Provided in Cooperation with: Johann Heinrich von Thünen Institute, Federal Research Institute for Rural Areas, Forestry and Fisheries Suggested Citation: Poeplau, Christopher et al. (2025) : Interim report of the repeated German agricultural soil inventory, Thünen Working Paper, No. 277a, Johann Heinrich von Thünen-Institut, Braunschweig, https://doi.org/10.3220/253-2025-235 This Version is available at: https://hdl.handle.net/10419/333952 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Thünen Working Paper 277a Interim report of the repeated German Agricultural Soil Inventory Christopher Poeplau, Laura Sofi e Harbo, Florian Schneider, Marcus Schiedung, Axel Don, Stefan Heilek, Rene Dechow, Elli Vasylyeva, Arne Heidkamp, Roland Prietz, Heinz Flessa Christopher Poeplau, Laura Sofie Harbo, Florian Schneider, Marcus Schiedung, Axel Don, Stefan Heilek, Rene Dechow, Elli Vasylyeva, Arne Heidkamp, Roland Prietz, Heinz Flessa Thünen Institute of Climate-Smart Agriculture Bundesallee 65 38116 Braunschweig Contact: PD Dr. Christopher Poeplau Phone: +49 531 2570 1239 E-Mail: C[email protected] Thünen Working Paper 277a ©2025 the author, Thünen Institute. This is an open access publication distributed under the terms and conditions of the Creative Commons Attribution 4.0 International (CC BY 4.0) license Braunschweig/Germany, Dezember 2025 I Summary The dynamics of organic soil carbon (SOC) play an important role in atmospheric CO2 concentrations and are therefore included in national greenhouse gas inventories. SOC is also essential for soil fertility. On behalf of the Federal Ministry of Food and Agriculture, the Thünen Institute of Climate-Smart Agriculture therefore conducted the first nationwide representative inventory of agricultural soils (BZE-LW) between 2010 and 2018. A total of 3,104 arable, grassland and permanent crop sites were sampled in an 8x8 km grid and analysed for SOC contents and stocks, as well as other parameters, down to a depth of 1 m. The BZELW repeat inventory project began in 2022, and since the beginning of 2023, resampling of the sites identified at that time has been in full swing. In addition to soil sampling, annual management data is also being collected. The main objective of this project is to quantify and explain potential changes in SOC contents and stocks over the past decade. This interim report presents the initial results of the ongoing repeat inventory. Compared to the initial BZE-LW, there were some deviations in the implementation of the repeat inventory. Only the top 50 cm are sampled and organic soils are not resampled. Instead of a central profile pit and eight additional core drillings, four small pits are now excavated for sampling. Some parameters are not recorded again (e.g. soil type, grain size distribution, stone content), while others have been added (e.g. aggregate stability, air capacity, cation exchange capacity). By October 2025, approximately 1,350 sites in eight federal states had been resampled and almost 1,000 had been analysed for bulk density and SOC content in order to calculate mass-corrected SOC stock changes. During the initial evaluation, it was noticed that the initial SOC content of the topsoil from the profile pit was systematically slightly too high, leading to an overestimation of SOC losses. For this reason, the SOC contents from additional core samples of the initial BZE-LW were used instead. However, analysis of those has not been completed yet, which is why only 587 sites have been included in the evaluation at this stage. Slight changes in SOC content have been observed in arable soils over the past decade. While a slightly positive trend in SOC content was observed on average in 0-10 and 10-30 cm, the change in SOC stocks in 0-30 cm (-1.6%) and 0-50 cm (-2.7%) was significantly negative due to a slight decrease in bulk density despite mass correction. At a depth of 0-10 cm, however, the change in SOC stocks was also slightly positive (0.9%), which can possibly be explained by a nationwide decline in tillage intensity and the resulting redistribution of SOC in the soil profile. However, the evaluation of the management data from the questionnaire was so far focused on one parameter: the frequency of cover cropping. In this respect, the BZE-LW data correspond well with national data, which show approximately a doubling of the annual cover crop area in the period under review. However, this gradual increase in SOC input into German arable soils was apparently not sufficient to compensate for potential negative influences on SOC. For grassland, there was a more pronounced decrease in SOC stocks, which was most pronounced at a depth of 0-10 cm (-8.1%). At depths of 0-30 cm and 0-50 cm, the significant relative decreases were -5.9% and -5.1%. Negative trends were also observed on average at all depth levels for the 14 permanent crop sites to date. The investigation of the causes of these SOC losses is still ongoing. According to the hypotheses developed here, it is land use history and soil genesis, rapidly advancing climate change, and recent changes in cultivation practices that are affecting the SOC dynamics currently being observed: The historically wet and often SOC-rich sandy soils of north-western Germany tend to suffer particularly severe SOC losses under current land use, which is consistent with the results of long-term soil observations in Lower Saxony and the neighboring Netherlands. Over the last 50 years, there has been an average air temperature increase of 2.1°C at the BZE-LW sites, about half of which has occurred in the last 1-2 decades. According to modelling and experimental work, warming alone is sufficient to explain the II magnitude of average SOC losses. Finally, national statistics clearly indicate a reduction in livestock farming and nitrogen fertilisation, which is also likely to have a negative impact on SOC stocks. Separating the various factors influencing SOC stocks requires complex methodology and will be a central part of the next project phase, alongside the completion of resampling and the remeasurement of the initial core samples. Another challenge will be the preparation and implementation of new reporting requirements, in particular the EU Soil Monitoring Law. The trends observed to date only apply to part of the Federal Republic of Germany. The average rates of change presented here should therefore not be extrapolated. Keywords: Agricultural soils, soil monitoring, soil organic matter, soil carbon, greenhouse gas reporting III Table of contents Summary I 1 Introduction 7 2 Materials and methods 9 2.1 Sampling and field data collection 9 2.1.1 General characteristics of the BZE-LW sampling design 9 2.1.2 General characteristics of the first repeat inventory 9 2.1.3 Sampling and field data collection for the repeat inventory 12 2.2 Sample preparation, laboratory analyses, and calculations of relevant parameters 12 2.2.1 Bulk density of fine soil and gravimetric content of coarse soil and water 12 2.2.2 Organic and inorganic carbon content and total nitrogen content 13 2.2.3 Soil organic carbon stock 14 2.2.4 pH value and electrical conductivity 15 2.2.5 Additional parameters 15 2.2.6 Archiving of soil samples 16 2.3 Collection of management data 16 2.4 External data 17 2.4.1 Land use history 17 2.4.2 Climate data 17 2.5 Statistics 17 3 Comparison of SOC results from the initial profile pit and soil cores 19 4 Results 22 4.1 Temporal dynamics of organic soil carbon content 22 4.2 Temporal dynamics of organic soil carbon stocks 25 4.3 Changes in key environmental and management influences 29 4.3.1 Changes in climatic conditions 29 4.3.2 Cover cropping 29 5 Discussion 31 5.1 Soil carbon dynamics in agricultural soils in Germany 31 5.2 An unscheduled course correction in the agricultural soil condition survey 35 6 Outlook 37 IV References 38 Acknowledgements 42 Appendix 43 V Table of Figures Figure 1 Sampling scheme for up to eight repeat inventories (marked by numbers 2-9) and drone image of a joint sampling of a site with four simultaneously opened profile pits. No. 1 marks the location of the central soil profile of the initial BZE-LW. ...........................................................10 Figure 2 Location of sites sampled in the repeat inventory by October 31, 2025, by land use, as well as sites still to be sampled. .............................................................................................................11 Figure 3 Density function of z-transformed SOC contents (uniformly scaled per site and depth level, with 0 as the mean value of the site across all available measurements) for the data from core drilling and profile pits from the initial inventory, as well as the values from the repeat inventory. ...19 Figure 4 Change in SOC stock based on the core drilling plotted against the change in SOC stock based on the initial profile pit with linear regressions and corresponding correlation coefficients (R²). The dashed diagonal line represents the 1:1 line. The section selected for better visualization (-20 to 20 Mg ha⁻¹) led to the exclusion of 15 sites in 0-10 cm, 51 sites in 10-30 cm, and 66 sites in 30-50 cm. The regressions are not affected by this. ..............................................................20 Figure 5 Location of sites that have already been resampled for which data from the initial core drilling and the relevant analysis results from the repeat inventory are available. ..............................21 Figure 6 Distribution of SOC contents from the initial inventory and repeat inventory by depth level and land use with median and mean values. ...................................................................................22 Figure 7 Distribution of changes in SOC content by depth level and land use with mean and median. The black dotted line represents no change (zero line). The section selected for better visualization (-20 to 20 g kg(-1) )led to the exclusion of 18 sites in 0-10 cm, 6 sites in 10-30 cm, and 4 sites in 30-50 cm. The representation of the median and mean values are not affected by this. ........23 Figure 8 Changes in SOC content sorted by size (colored dots) for all cropland and grassland sites at three depth levels with minimum detectable difference (MDD) for the respective site (gray bars). The MDD is a positive value that, depending on the variability at the site, indicates how large a change (amount) in SOC content must be in order to assume a change with statistical certainty. For the figure, the MDD was mirrored into negative values to enable a direct comparison with the losses in SOC content. Negative changes in SOC content are colored red, positive changes blue, and values greater than the MDD are shown in dark colors. Changes within the MDD, and therefore not significantly significant, are shown in light colors. The section -25 to 25 g kg-1 selected for better visualization led to the exclusion of 4 grassland sites in 0-10 cm, one cropland and one grassland site in 10-30 cm, and 3 cropland and 2 grassland sites in 30-50 cm. .......................................................................................................................24 Figure 9 Box plots (with 1st, 2nd, and 3rd quartiles) of SOC stock changes in the three cumulative depth levels with number of observations (n) for all land uses. ..........................................................26 Figure 10 Spatial distribution of SOC stock changes at a depth of 0-30 cm for all sites resampled to date for which core drilling data were available (n=578). Areas marked with an asterisk have deviations of more than 25 Mg ha-1) (n=21, 6 croplands, 15 grasslands). ................................27 Figure 11 Changes in SOC stocks at a depth of 0-30 cm as a function of the mean SOC stock of the site (mean of both inventories), sand content, and initial C:N ratio for all three land use classes with correlation coefficients (R²). ......................................................................................................28 Figure 12 Box plots (with 1st, 2nd, and 3rd quartiles) of the change in SOC stock at a depth of 0-30 cm in cropland (brown) and grassland sites (green), grouped according to land use history (cropland = long-term cropland, grassland to cropland = cropland with grassland history, cropland to grassland = grassland with cropland history, grassland = long-term grassland). Only those sites are shown that had no confirmed grassland rotation in the 10 years prior to the initial sampling, VI or were characterized by other previous uses (e.g., moor, heath, forest). The land use change is based on the last 136 years according to Emde et al. (2024). ...............................................28 Figure 13 Distribution of mean annual temperature, mean monthly temperature, annual potential evaporation, and annual climatic water balance of all 3104 BZE-LW sites for the last six calendar decades. .....................................................................................................................................30 Figure 14 Time series showing the proportion of sites with intercropping for each year in the period 20012022. Blue = initial inventory questionnaire, green = repeat inventory questionnaire. Where the two questionnaires overlapped (possible in the years 2011-2015), the information from the repeat inventory questionnaire was used for the sake of simplicity.........................................30 Figure 15 Annual rates of change in average national SOC stocks in various European countries and Germany for cropland and grassland in the respective inventory periods, differentiated by topsoil, subsoil (defined slightly differently depending on the study) and the entire soil profile (if relevant). The data are taken from a previously unpublished literature review and are presented here in modified form (Harbo et al., submitted). .....................................................32 Materials and methods 13 straw), and roots. All four components are weighed individually. Anthropogenic material is given special consideration, as it could provide clues about the history of the site and the general degree of material admixtures in agriculturally used soils. The roots are not washed, but weighed directly after sieving and sorting. Furthermore, no distinction is made between dead and living roots. It is therefore a very rough indicator of the actual root biomass. The bulk density of the entire soil BDtotal (g cm-3) and the bulk density of the fine soil BDfine (g cm-3) are calculated as follows: 𝐵𝐷𝑡𝑜𝑡𝑎𝑙=𝑀𝑡𝑜𝑡𝑎𝑙 𝑉𝑜𝑙𝑡𝑜𝑡𝑎𝑙 𝐵𝐷𝑓𝑖𝑛𝑒=𝑀𝑡𝑜𝑡𝑎𝑙-𝑀𝑟𝑜𝑐𝑘𝑠 𝑉𝑜𝑙𝑡𝑜𝑡𝑎𝑙-𝑀𝑟𝑜𝑐𝑘𝑠 𝜌𝑟𝑜𝑐𝑘𝑠 Here, Mtotal denotes the total mass of the sample (g), Voltotal the total volume of the sample (cm³), while Mfine denotes the mass of the fine soil <2 mm (g), Mrocks denotes the mass of the rock fragment fraction >2 mm (g), and ρrocks denotes the bulk density of the rock fragment fraction (g cm-3) in the sample. The latter was set to 2.65 g cm-3 (bulk density of quartz) as standard if there were no indications of significant deviations from this value. For certain sediments or volcanic rocks, the bulk density of the skeleton may deviate significantly from this value. In such cases, that bulk density was analyzed in the initial BZE-LW in order to adjust the calculation of the BD(fine) accordingly (Jacobs et al., 2018) . These specific bulk densities were also used in the repeat inventory. 2.2.2 Organic and inorganic carbon content and total nitrogen content The samples for chemical analysis are dried at 40°C until their weight remains constant and sieved to 2 mm. For organic and inorganic carbon (SOC, SIC) and total nitrogen (Ntot), the samples from each individual pit and depth level are measured separately. This serves to identify small-scale fluctuations and possible outliers. It also provides a measure of uncertainty, which helps in the classification and statistical analysis of observed changes. In addition, it opens up the possibility of calculating the SOC reserve for each individual pit and thus more accurately on average, as the bulk density per pit and depth level is also determined. Prior to elemental analysis, aliquots of approximately 20 g are ground to <63 µm to homogenize the sample. Of these homogenized samples, 300-800 mg are used for elemental analysis. Carbonate-free soils are analyzed for carbon and nitrogen using a CN analyzer (vario MAX cube, Elementar, Langenselbold), while carbonate soils (as identified in the field or via pH measurement) are analyzed using a TOC analyzer (soli TOC cube, Elementar, Langenselbold) with differentiated temperature increase. The threshold between SOC and SIC is set at 550°C according to VD-LUFA. The Ntot content in carbonate soils is subsequently measured using a nitrogen analyzer (rapid N exceed, Elementar, Langenselbold). Since there was a change in the elemental analyzers between the initial inventory of the BZE-LW and the repeat inventory (from LECO to Elementar), a comparison between the different instruments was carried out on a systematically selected sample set by measuring archived samples from the initial BZE-LW again with the new elemental analyzers. A total of 50 carbonate-free and 50 carbonate-containing samples between 0 and 100 g kg-1 SOC were selected and measured again. No systematic deviations for SOC and Ntot could be detected. During the initial evaluation of the results, several sites stood out where mean SOC contents varied greatly between the two inventories. Very high spatial variability was identified as a major reason for the significant changes after a decade. Within the four profiles of the repeat inventory, ranges in SOC content of up to 100 g kg-1 or 10 percentage points were observed. These were often sites with organic horizons or burials in 14 Materials and methods deeper soil areas. It can be assumed that the sampling design of the BZE-LW, especially with the peculiarity of the non-replicated initial profile pit, reaches its limits at such and other sites with extreme, small-scale variability, or does not allow for a meaningful evaluation of SOC content changes. Based on the distribution of these ranges, threshold values were defined for each land use and depth level, above which a site was excluded from the evaluation in this interim report. In cropland and permanent crop soils, these ranges were 20 g kg-1 in 0-10 cm and 10-30 cm, and 40 g kg-1 in 30-50 cm, and 40 g kg-1 in all depth levels in grassland soils. The variation patterns in Ntot content were very similar to those in SOC content, which is why the reduction of sites was limited to filtering based on SOC content. Any site where an exceedance was found at any depth level was excluded from further evaluation. A total of 21 of the resampled sites were excluded from this interim report. 2.2.3 Soil organic carbon stock The key parameter of the BZE-LW is the SOC stock (Mg C ha-1 and given depth). This is determined for each individual profile pit and depth level using the following formulas and then cumulated for the depth levels 0-30 and 0-50 cm: 𝐹𝑆𝑆 = 𝐵𝐷𝑓𝑖𝑛𝑒×𝑡ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠×(1-𝑟𝑜𝑐𝑘𝑠 100)×100 SOC stock = 𝐹𝑆𝑆 ×𝑆𝑂𝐶 𝑐𝑜𝑛𝑡𝑒𝑛𝑡 1000 FSS is the fine soil stock (Mg ha-1), which can be calculated from the BDfine, the thickness of the respective depth level (cm), the volumetric rock fragment fraction (rocks, vol. %) and a conversion factor of 100 (Poeplau et al., 2017). The volumetric rock fragment fraction of the soil was taken from the initial BZE-LW, as this can be considered approximately constant over a 10-year interval and would otherwise cause excessive noise with regard to SOC stock changes (Munera-Echeverri et al., 2025). The SOC stock (Mg ha-1) at any depth level can be obtained by multiplying the FBV by the SOC content (g kg-1) and a conversion factor of 1000. Changes in the bulk density of the fine soil lead to a change in the soil mass or the fine soil stock at the respective depth level. However, a comparison of SOC stocks between two points in time should be based on identical fine soil stocks in order to track actual changes in the SOC stock in a defined amount of soil (Ellert and Bettany, 1995; von Haden et al., 2020). Since sampling identical fine soil stocks in the field is practically impossible, a mathematical correction must be made. In this case, the fine soil stock in a given soil package (0-10, 0-30, and 0-50 cm) was specified by the initial inventory. For this purpose, first determined the cumulative fine soil stock for the soil package to be corrected for each of the four soil profiles in the repeat inventory. The procedure for mass correction is that soil packages with an excessively high fine soil stock (compared to the reference) are reduced mathematically at the lower end to simulate a correspondingly shallower sampling, while soil packages with an excessively low mass are extrapolated to reach the reference mass. This is done separately for all three soil packages (0-10, 0-30, and 0-50 cm) and each individual profile, always using the respective fine soil stock of the initial inventory as a reference. The reduction of a soil package and the subsequent correction of the corresponding SOC stock of a profile in the repeat inventory is carried out as follows: ∆𝐹𝑆𝑆 = 𝐹𝑆𝑆𝑟𝑒𝑝𝑒𝑎𝑡-𝐹𝑆𝑆𝑖𝑛𝑖𝑡𝑖𝑎𝑙 Materials and methods 15 𝑆𝑂𝐶 𝑠𝑡𝑜𝑐𝑘𝑠ℎ𝑜𝑟𝑡𝑒𝑛𝑒𝑑 =SOC stock -∆𝐹𝑆𝑆× 𝑆𝑂𝐶 𝑐𝑜𝑛𝑡𝑒𝑛𝑡𝑙𝑜𝑤𝑒𝑠𝑡 𝑖𝑛𝑐𝑟𝑒𝑚𝑒𝑛𝑡 1000 Here, FSSrepeat denotes the fine soil stock of a profile pit in the repeat inventory down to the depth of the respective soil package (Mg ha-1), while FSSinitial denotes the fine soil stock of the respective soil package of the initial profile (Mg ha-1 ). The difference between these two fine soil stocks is then used to determine a SOC stock, which is subtracted from the previously determined SOC stock of the soil package. To do this, ΔFSS is multiplied by the SOC content (g kg-1) of the lowest depth level of the soil package to be corrected. The extrapolating correction of a soil package is carried out accordingly for each individual soil profile of the repeat inventory as follows: 𝑆𝑂𝐶 𝑠𝑡𝑜𝑐𝑘𝑠ℎ𝑜𝑟𝑡𝑒𝑛𝑒𝑑 = SOC stock -∆𝐹𝑆𝑆×𝑆𝑂𝐶 𝑐𝑜𝑛𝑡𝑒𝑛𝑡𝑢𝑛𝑑𝑒𝑟𝑙𝑦𝑖𝑛𝑔 𝑖𝑛𝑐𝑟𝑒𝑚𝑒𝑛𝑡 1000 In comparison to the preceding formula, the SOC content of the underlying increment is now used. For the soil package 0-10 cm, this is the SOC content from 10-30 cm, and for the soil package 0-30 cm, it is the SOC content from 30-50 cm. For the soil package 0-50 cm, extrapolation with SOC content from the repeat inventory is not possible, as no underlying increment is sampled below 50 cm. If extrapolation is nevertheless necessary, the SOC content of the underlying increment (50-70 cm) from the initial inventory is used. However, in order to perform the extrapolation with as little "foreign" data as possible that was not measured in the respective profile, a two-step approach is used for extrapolation below 50 cm: First, the SOC stock of each profile is adjusted to the heaviest mass of the four profiles in the repeat inventory. For this purpose, the SOC content of the 30-50 cm depth level from the respective profile is used. Only the difference between the fine soil stock of the heaviest profile of the repeat inventory and the profile of the initial inventory is extrapolated using the SOC content below 50 cm from the initial inventory. 2.2.4 pH value and electrical conductivity The pH value is determined both in distilled water (pHH2O) and in calcium chloride (pHCaCl2). For the analysis of pH and electrical conductivity, pool samples are first created for each depth increment of a site. To do so, approximately equal proportions of the individual samples from the four profile pits are mixed into one sample after sieving in order to reduce the analytical effort and the number of archive samples. Only for the selected focus sites (core sites) are all individual samples analyzed in order to obtain a measure of the smallscale variability of the respective parameters. For the analysis, five ml of dried fine soil is shaken upside down for 20 minutes with 25 ml of distilled water or a 0.01 molar CaCl2 solution. After subsequent 10minute centrifugation, the measurement is carried out using a pH-meter (ProLab 4000, SI-Analytics, Mainz) in a measuring robot (SP2000, Skalar, Breda). 2.2.5 Additional parameters Spectra were also recorded in the laboratory in the mid-infrared range (2500-25000 nm; 4000-400 cm-1) using diffuse reflection infrared Fourier transform spectroscopy (DRIFT-MIR, Nicolet iS50 with Collector II, Thermo Fisher, Waltham). A total of 7000 spectra have been collected out of a planned 28000 spectra. Together with the spectra in the near-infrared range (900-3400 nm; 11000-3000 cm-1 (FT-NIRS MPA, Bruker, Billerica) with around 14000 spectra (Jaconi et al., 2017; Vos et al., 2018), this will result in the "German Agricultural Soil Spectral Library" (GASSL). Infrared spectroscopy enables the estimation of various soil parameters, such as organic carbon, its distribution in fractions of varying stability, but also texture and pH, as well as other parameters (Sanderman et al., 2020). The DRIFT-MIR spectra in particular contain direct information about the chemical and 16 Materials and methods functional composition of the organic soil substance as well as about the properties of the mineral phase (Margenot et al., 2023; Schiedung et al., 2025). The advantage here is that spectroscopy is far less complex than conventional methods and can therefore be used to perform qualitative and quantitative assessments for all samples, not just a selection, using the GASSL in the future. At the same time, several parameters can be estimated from a single spectrum. Thus, in the further course of the BZE-LW, models for the quantitative estimation of various soil parameters and also qualitative parameters will be integrated into the evaluation. However, GASSL will not be discussed further in this report. In addition, the samples from the repeat inventory are analyzed for cation exchange capacity and base saturation (all depth levels) as indicators of nutrient supply, as well as aggregate stability (topsoil, 0-10 cm) and air capacity (subsoil, 30-40 cm and 40-50 cm) as indicators of structural stability and subsoil compaction. In addition, topsoil samples (0-10 cm) were selected from 300 cropland sites to characterize the microbial community and investigate various functional microbial properties. An analysis of the proportion of pyrogenic carbon in the soils of the core sites is also planned. Previous results of these additional parameters are not covered in this report. 2.2.6 Archiving of soil samples The finely ground aliquot of each sample (approx. 20 g of dry soil), which was also used for elemental analysis, is archived. In addition, 2 kg of each mixed sample from the individual depth levels of the four profiles is set aside for future analysis, if available. No mixed samples are kept from the core sites; instead, all individual samples are retained. The archive is dark (UV-protected), cool, and dry to ensure minimal change and maximum shelf life of the soil samples and containers. 2.3 Collection of management data In order to interpret changes in soil properties, fieldand farm-specific management information since the initial inventory is of great importance. This information is collected using a questionnaire, which is available in both digital and analog formats. The following data is requested, which is limited to the central question of the BZE-LW: Type of farm, farming method (conventional/organic, since when), land area of the farm, average purchase and sale quantities of organic fertilizers or substrates, cropland and grassland area of the sampled parcel, size of the sampled field, and distance of the sampling point from the central farm location. Annual data is collected on the type and yield of the main crop, removal of straw and by-products, type, sowing date, incorporation date and use of any catch crops, type of permanent grassland use, number of cuts in grassland, date of 1st and 2nd cut, type and depth of grassland renewal, type, number and duration of livestock farming on the sampled plot, type and amount of fertilization, type and amount of liming, type and depth of grassland renewal, type, number and duration of livestock farming on the sampled plot. cut, type and depth of grassland renewal, type, number and duration of livestock farming on the sampled plot, type and amount of fertilization, type and amount of liming, type and depth of tillage, additional measures such as drainage, irrigation, deep loosening, soil or plant additives, or alternative land use. The questionnaires received are quality assured and, if information is missing or too inaccurate, the farmers are contacted by telephone. This is also the case if questionnaires are not received even after multiple reminders. Since participation in the BZE-LW is voluntary and completing the questionnaire requires a considerable amount of effort, some farmers are unwilling to participate. As a result, the response rate for the questionnaires is not optimal, but this can only be influenced to a limited extent. To date, 1,756 questionnaires have been sent out and 889 have been completed and returned (51%). Quality assurance and processing of the management information are ongoing. Harmonizing the data from both questionnaires into a consistent time series is a particularly challenging task. For this reason, the focus Materials and methods 17 at this stage has been on a single parameter that is central to SOC dynamics in cropland soils: the proportion of catch crops in crop rotation. In order to calculate this trend for all questionnaires received to date and to create a general time series, 1) only those sites that reported for at least 5 years in both time periods (before the initial sampling and between samplings) were used, 2) in the case of temporal overlaps between the two questionnaires, only information from the second questionnaire was used, 3) the last two years (2023 and 2024) were ignored due to the significantly lower sample size. This resulted in a slightly different number of questionnaires (223-368) for each year between 2001 and 2022 for which information on the cultivation of catch crops was available. From this, the proportion of locations where catch crops were cultivated in a specific year was calculated. 2.4 External data 2.4.1 Land use history Previous land use or land cover can have a strong influence on soil properties and their current temporal dynamics (Emde et al., 2024) . Data on land use history at the sampling points come from various sources. One source was the initial inventory questionnaire, which already asked whether information on the historical use or cover of the area was available. A distinction was made between cropland, grassland, permanent crops, forest, moorland, heathland, and other uses. Since only some of this data is available to today's farmers, the questionnaire data set contained large gaps and did not go back further than four decades on average. Therefore, various archives and repositories of historical maps and orthophotos were consulted in order to obtain as complete a data set as possible. The aim was to have land use information available for every location for at least 100 years, every 30 years. Continuous time series were generated from these incomplete time series by interpolation, from which approximate dates of land use changes can be derived. More detailed information on the generation of this data set is described by Emde et al. (2024). This report uses only part of this dataset to enable meaningful evaluation: long-term cropland and longterm grassland (no land use changes in the past 136 years), as well as cropland with a grassland history and grassland with an arable history in the past 136 years. Locations with frequent changes between cropland and grassland were excluded, as were those with other historical land uses. 2.4.2 Climate data Climate conditions are highly relevant for the development of SOC stocks, as they directly influence microbial activity and thus the degradation kinetics of organic matter in the soil on the one hand, and biomass growth and thus the amount of crop and root residues that can be returned to the soil on the other. The weather data used here comprise monthly average temperatures (DWD, 2025a) and monthly totals for precipitation (DWD, 2025b) and sunshine duration (DWD, 2025c). The data were obtained as grid maps with a resolution of 1x1 km and for the period 1970 to 2024 from the Open Data Center of the German Weather Service. Time series for the BZE-LW locations were then extracted from these maps. Global radiation was calculated from sunshine duration according to (Allen et al., 1998) and potential evaporation was estimated according to Turc (Wendling et al., 1991). 2.5 Statistics Average changes in SOC content, BDfine and SOC stocks were analyzed using a bootstrapping method to enable error estimation and determine significance. For this purpose, the change in a parameter between the initial inventory and the resampling was determined for all available locations. From these n changes, a 18 Materials and methods new sample was generated by drawing n times with replacement, and the mean value was calculated. This process was repeated a total of 5000 times, resulting in 5000 bootstrap mean values of the change. The interval between the 2.5% and 97.5% percentiles of the bootstrap means was defined as the 95% confidence interval (CI95) of the mean change. If this interval of change was completely beyond zero, i.e., exclusively in the positive or negative range, the mean change was interpreted as significant (Ho et al., 2019). The bootstrapping method was chosen because it is nonparametric, does not require normal distribution of the data, and, in addition to significance, confidence intervals of the changes could also be calculated (Çetinkaya-Rundel and Hardin, 2024). By switching from the initial profile pit to the surrounding core drillings (see Chapter 3), the spatial variability of the SOC contents could be determined for both time points. With this uncertainty, a minimum detectable difference (MDD) could be calculated for each site, which is directly proportional to the small-scale variability for a given sample size and allows a statement to be made about the statistical significance of the measured SOC content change. The formula according to (Valk et al., 2000) was used for this purpose: 𝑀𝐷𝐷=√(𝑍𝛼+𝑍𝛽)²×𝜎² 𝑛 Where α expresses the statistical significance level (here α=0.05), β is a measure of the probability that a statistical effect can be found (1 - statistical power, here 0.8), and the Z-value indicates how many standard deviations a value is from the mean of a standard normal distribution. With an α of 0.05, Zα is the 95% percentile of a standard normal distribution and thus 1.96 standard deviations from the mean, while Zβ is 0.84 standard deviations from the mean. The standard deviation of the measured values is included in the formula as σ, and n describes the number of measurement repetitions. An MDD was determined for both the initial BZE (8 soil cores) and the repeat inventory (n=4). The change in SOC content was considered significant if it was greater than both MDD values. To test whether land use history had a significant effect on the recent change in SOC stocks in 0-30 cm, an analysis of variance (ANOVA) of the changes was performed. The aim was to specifically test whether SOC stocks from long-term cropland or long-term grassland have a significantly different dynamic than those with known land use change (from cropland or grassland) over the past 136 years. The normal distribution of the residuals was checked using QQ plots. Due to the non-normal distribution of the ANOVA residuals, a non-parametric Kruskal-Wallis test was performed with Wilcoxon as a post hoc test. Due to the reduced size of the data set (see Chapter 3), the limited spatial representation of the sites that have been resampled and analyzed to date, and the fact that only partial management information is available, no global statistical model was calculated in this interim report to explain changes in SOC content and reserves, as well as other soil properties. Only simple correlation analyses (linear regressions) with individual soil properties and changes in climate variables were performed. Results 19 3 Comparison of SOC results from the initial profile pit and soil cores An initial analysis of the changes in SOC content and stocks from the eight federal states sampled to date revealed significant decreases after the first resampling compared to the initial inventory. This was observed regardless of land use. To verify the plausibility of the supposed trends, additional ram core sampling (hereinafter referred to as soil cores) from the initial inventory was used where available. It turned out that i) the deviation between the mean value of the soil cores and the profile pit strongly determined the difference between the initial profile pit and the resampling, and ii) the difference between the resampling and the initial core sampling was on average significantly smaller than the difference between both and the initial profile pit (Fig. 3). Figure 3 Density function of z-transformed SOC contents (uniformly scaled per site and depth level, with 0 as the mean value of the site across all available measurements) for the data from core drilling and profile pits from the initial inventory, as well as the values from the repeat inventory. Source: Thünen Institute In addition, the difference between the initial profile pit and the resampling was most pronounced at precisely those depth levels that had already been identified as problematic during the initial inventory (1030 cm in cropland land and 0-10 cm in grassland) (Fig. 3). These depth levels are those in which strong gradients in SOC content often occur or which are underlain by significantly SOC-poor depth levels. It seems likely that sampling of the initial profile pit systematically led to slightly lower SOC contents and SOC stocks. One possible explanation is that areas poorer in SOC (appearing lighter in the profile) tended to be omitted or sampled disproportionately. 20 Comparison of SOC results from the initial profile pit and soil cores Furthermore, there is a surprisingly low degree of agreement between the SOC changes determined on the basis of the results of the initial profile pit and the initial ram core drilling (Fig. 4). This also leads to the conclusion that the SOC contents of the initial profile pit are a poorer reference than the results from the drill cores. It was therefore decided to remeasure the SOC and Ntot contents on some of the drill cores that had not yet been analyzed (four of eight drill cores, three depth levels) and to use the existing results from the core drilling as the initial values for the SOC contents for this report. SOC changes are now reported accordingly for 578 locations for which core drilling from the initial inventory and resampling are currently available (Fig. 4). It is unclear whether or to what extent other chemical soil properties show similar systematic differences between the sampling methods (profile pit or drill cores). Changes in pH values are therefore not addressed in this report, as they have yet to be determined for the drill cores. One advantage of this change is that rates of change can now also be checked for significance on a site-specific basis or assigned an uncertainty value, which allows for a better classification of differences between the two inventory rounds. Figure 4 Change in SOC stock based on the core drilling plotted against the change in SOC stock based on the initial profile pit with linear regressions and corresponding correlation coefficients (R²). The dashed diagonal line represents the 1:1 line. The section selected for better visualization (-20 to 20 Mg ha⁻¹) led to the exclusion of 15 sites in 0-10 cm, 51 sites in 10-30 cm, and 66 sites in 30-50 cm. The regressions are not affected by this. Source: Thünen Institute Results 21 Figure 5 Location of sites that have already been resampled for which data from the initial core drilling and the relevant analysis results from the repeat inventory are available. Source: Thünen Institute 22 Results 4 Results 4.1 Temporal dynamics of organic soil carbon content The conversion to SOC contents from the core drilling samples initially resulted in a significantly reduced sample size for the SOC parameter. Of the original 1,000 or so resampled and analyzed sites, an evaluation of SOC content and stock changes could be carried out for a total of 578 sites for which core drilling data had already been collected (see Chapter 3). The fundamentally right-skewed distributions of SOC contents in the three land uses and depth levels and for both inventories are shown in Fig. 6. The change in SOC contents is shown in Fig. 7 and Table A1. These are approximately normally distributed. While the SOC content in the topsoil of cropland (0-10 and 10-30 cm) increased slightly (significantly in 0-10 cm), significant decreases were observed on average in the subsoil (30-50 cm). The most significant overall change in SOC content was observed in the topsoil of grassland soils, especially at a depth of 0-10 cm. A significant decrease in SOC content was observed at this depth for the 140 grassland soils sampled to date. Decreases were also measured on average at the depths below. This was also true for permanent crops, where there was also a tendency toward decreases on average at all depths. However, due to the small sample size of permanent crops, this trend could not be statistically verified. Figure 6 Distribution of SOC contents from the initial inventory and repeat inventory by depth level and land use with median and mean values. Source: Thünen Institute Results 29 A preliminary evaluation of the land use history showed that previous use as cropland or grassland had a significant influence on the recent dynamics of SOC stocks at a depth of 0-30 cm (Fig. 12). While cropland with a grassland history tended to show SOC losses, the SOC stocks of long-term cropland (no land use change in the last 136 years) tended to be in equilibrium. Conversely, a clear negative trend in SOC stocks was observed for long-term grassland, which differed significantly from grassland with a history of arable use. 4.3 Changes in key environmental and management influences 4.3.1 Changes in climatic conditions Since the 1970s, the mean annual temperature of the air near the ground (2 m above ground level) has risen significantly at the 3104 BZE-LW sites (Fig. 13). While the mean annual temperature in the 1970s was 8.4°C, it has been 10.5°C in the past 5 years. A significant increase in the annual mean temperature has been observed, particularly in the last 15 years, representing a non-linear increase over the past 50 years (Fig. A1). Accordingly, potential evapotranspiration (evaporation) has also increased significantly with each decade. Annual precipitation, on the other hand, has changed little overall, causing the annual climatic water balance to become more negative or smaller. The proportion of potentially arid sites (more potential evaporation than precipitation on average over the calendar decade) has increased from 11% in the 1970s to 23% in the past five years. Cropland sites (26%) are more affected than grassland sites (14%). The spatial distribution of warming is much more homogeneous than the change in the climatic water balance (Fig. A2). The already dry east of Germany has become even drier over the past 50 years. Almost all BZE-LW sites in Brandenburg and Saxony-Anhalt, as well as parts of Mecklenburg-Western Pomerania and Saxony (Fig. A3), have been arid on average over the past 15 years. The proportion of arid sites has also increased significantly in Rhineland-Palatinate, Hesse, and Bavaria. In contrast, the climatic water balance in the far north-west of Germany has become more positive over the past 15 years. It can therefore be assumed that there will be region-specific effects of changed climatic conditions on soil properties. 4.3.2 Cover cropping According to an initial evaluation of the two questionnaire data sets, the proportion of cropland sites with catch crops has increased significantly over the past 10 years (Fig. 14). While between 2000 and 2013 the average was around 12% of sites, the data from the repeat inventory questionnaire shows a clear upward trend. The past two years (2023, 2024) were not included in the time series due to insufficient data, but by 2022 the proportion of sites with a cover crop had risen to just under 20%. 30 Results Figure 13 Distribution of mean annual temperature, mean monthly temperature, annual potential evaporation, and annual climatic water balance of all 3104 BZE-LW sites for the last six calendar decades. Source: Thünen Institute with data from the Deutscher Wetterdienst (DWD 2025a, b, c) Figure 14 Time series showing the proportion of sites with intercropping for each year in the period 2001-2022. Blue = initial inventory questionnaire, green = repeat inventory questionnaire. Where the two questionnaires overlapped (possible in the years 2011-2015), the information from the repeat inventory questionnaire was used for the sake of simplicity. Discussion 31 5 Discussion 5.1 Soil carbon dynamics in agricultural soils in Germany In the BZE-LW cropland sites analyzed repeatedly, only small, albeit in some cases significant changes in SOC content and stocks have been observed. While there was a slight surplus of sites with a positive trend (SOC increase) in the topsoil (0-10 and 10-30 cm), slight losses were observed in the subsoil (30-50 cm). Analysis of the minimum detectable difference (MDD) for SOC content revealed that a large proportion (60%) of the changes observed at site level were below the detection limit (Fig. 8). This means that no reliable statements can be made about changes at these sites. For those sites with very small changes in particular, this can be interpreted as meaning that the positive or negative trend was more likely to be random. Nevertheless, a significant increase in SOC content was observed in the 0-10 cm depth of the cropland soils for 24% of the sites, while only 16% of the sites showed significant SOC losses (Table 1). Even though this was not reflected in the average cumulative SOC stock of cropland soils (Fig. 9), it could be an indication that both increased intercropping and more conservation-oriented, and thus shallower, tillage are having a certain positive effect. According to the Federal Statistical Office, conventional tillage using plows has declined from 53% to 40% in Germany over the past seven years (Destatis, 2025b). Conservation tillage, or no-till farming, ensures at least a redistribution of SOC within the soil profile, i.e., an enrichment in the area of the highest C inputs (close to the surface) and a tendency toward a decrease in the area below (abandoned topsoil) (Meurer et al., 2018). Intercropping has increased significantly over the past decade at the BZE sites already sampled (Fig. 14). This is consistent with statistics on a national scale: according to the Federal Statistical Office, intercropping has increased from just under 1.2 million hectares to just under 2.2 million hectares since 2010 (Destatis, 2024). This increase can be explained primarily by changes in agricultural subsidies and new regulations. With the reform of the Common Agricultural Policy in 2013, the creation of ecological priority areas, including the cultivation of cover crops, was specifically promoted for the first time. In addition, with the entry into force of the new Fertilizer Ordinance (2020), the cultivation of cover crops between winter and summer crops has become mandatory in nitrate-polluted, so-called red areas. The effects of these positive developments in soil management on SOC contents and stocks can therefore only be observed to a limited extent so far. Despite the slightly positive trends in SOC content and stocks at a depth of 0-10 cm in cropland soils, there have been significant decreases in SOC stocks in all land use classes considered in the depth range relevant for greenhouse gas reporting (0-30 cm) over the past decade. This is comparable to the results from other regions in Europe. Various national inventories (Fig. 15) and the European Commission's Europe-wide soil inventory (LUCAS Soil) currently report negative trends in SOC contents and stocks (De Rosa et al., 2024). This also applies in part to the long-term soil monitoring of the federal states (Höper and Meesenburg, 2021; Wiesmeier et al., 2025). The lack of measurable positive effects of improved soil management on average SOC contents and stocks in cropland soils, as well as the even more pronounced loss of SOC from grassland soils and those under permanent cultivation, can have various causes. 32 Discussion Figure 15 Annual rates of change in average national SOC stocks in various European countries and Germany for cropland and grassland in the respective inventory periods, differentiated by topsoil, subsoil (defined slightly differently depending on the study) and the entire soil profile (if relevant). The data are taken from a previously unpublished literature review and are presented here in modified form (Harbo et al., submitted). Source: Thünen Institute 1) Increasing cover crop cultivation from 10% to 20% of the annual cultivated area has a relatively small effect on the SOCreserves of all cropland soils. A share of 10% means that, mathematically speaking, every field has a cover crop once every 10 years on average. Doubling this cultivation area is not only a positive development for SOCin the soil, but is also relevant for many soil and ecosystem functions (Shackelford et al., 2019) . Doubling the annual cultivation area means that, on average, a cover crop is currently cultivated twice in 10 years. The average effect of green manure from cover crops on the SOC stock is approximately 0.3 Mg ha-1 yr-1 with annual cultivation (Poeplau and Don, 2015). However, if only one more cover crop was grown in the past decade, then the expected effect is very small compared to other possible effects on the SOC stock (see following points) and compared to the magnitude of the random sampling error. The random sampling error was estimated at an average of about 3 Mg ha⁻¹ when sampling and resampling cropland soils using three soil profiles on the same day (Poeplau et al., 2022). In addition, the proportion of catch crops has not increased sharply over the last decade, but rather gradually. It is therefore likely that the effect of an overall doubling of the area under catch crop cultivation on SOC stocks is still too small to be detectable as such in the repeat inventory of the BZE-LW on a national scale. However, if intercropping is included in a larger statistical model in the future, it is likely that part of the variability in SOC stock changes will be explained by this trend. 2) Climate change is highly likely to have a negative impact on global SOC stocks (García-Palacios et al., 2021). In recent decades, and especially in the current decade, Central Europe has warmed significantly. An increase in the annual mean temperature of 2°C since the 1970s has now been exceeded at all BZE-LW sites Discussion 33 (Fig. 13 and Fig. A1). In the period between the initial and repeat inventories alone, a warming of the air near the ground of about 1°C was measured. It has also been shown that the soil temperature in cropland soils rises more strongly than the air temperature (Dorau et al., 2022). This has consequences for microbial activity and thus the turnover of organic matter in the soil. According to model calculations and warming experiments in different climate zones, the magnitude of the relative loss of SOC due to warming is approximately 3-5% per °C increase in near-surface air temperature (Peplau et al., 2021; Poeplau and Dechow, 2023; Verbrigghe et al., 2022). With an average initial SOC stock of 65 Mg ha-1 initial mean value in 0-30 cm of the cropland sites sampled to date would correspond to approximately 1.9-3.3 Mg ha⁻¹; for grasslands (initial mean value of 89 Mg ha⁻1 this would already be 2.74.5 Mg ha⁻¹. In addition, higher evaporation and an accumulation of dry years in many places are leading to a trend toward drier conditions (Fig. A2), which can have a negative impact on yield formation and thus on C inputs. Although the average yields of the most important crops have continued to rise slightly over the past 20 years, interannual yield variability has also increased significantly, and the size of irrigated cropland in Germany has risen from 370,000 hectares in 2009 to around 500,000 hectares in 2019 (Destatis, 2023). In grasslands, weather conditions have a particularly strong influence on biomass development (Liu et al., 2023), and so the dry years of 2018 and 2022 resulted in two years of low yields in grasslands between the two sampling periods (Destatis, 2025a). In rather dry locations, the permanent decrease in soil moisture can have a negative effect on the mineralization of SOC (Kuka et al., 2025), but in rather wet locations, the opposite can also be the case (Smith et al., 2007; Van Wesemael et al., 2010), as drying leads to increased aeration and thus stimulation of microorganisms. The effects of rapid climate change are complex, and at least the temperature increase is omnipresent to a similar extent. For this reason, the temperature increase, or other climate variables, could not explain the variability of SOCchanges in simple regression analyses (data not shown). This will be similar in more complex statistical models, which is why only the use of process models can isolate the potential climate change signal in SOC dynamics from other influences. However, it is important to carefully check whether these models can correctly represent climate change effects (Hararuk et al., 2015). The fact is that, in many places, improved soil management in the face of advancing climate change can no longer be about enriching SOC, but merely about limiting losses (Don et al., 2024b; Riggers et al., 2021). 3) Changes in the SOC content of the soil occur over long periods of time, which means that past conditions can persist or have a lasting effect for a long time, thereby also influencing current trends (Fig. 12). After extensive research into land use history, it was shown for the initial BZE data set that grassland or arable land use can have an effect lasting several decades (Emde et al., 2024). The new steady state of the SOC stock in cropland soils after grassland use was estimated at around 180 years. In line with this, the present report also showed that fields that had been used as grassland in the past 136 years tended to lose SOC on average, whereas this was not the case for long-term croplands. A reverse trend was observed for grassland sites, which indicates the influence of land use history. Fundamentally, and viewed over a very long period of time, all of today's cropland soils are highly likely to have a history of higher SOC contents (Sanderman et al., 2017), which may continue to have an impact to this day. For example, losses of SOC in Finnish cropland soils have been linked to deforestation that took place decades ago (Heikkinen et al., 2013). In contrast, there are long-term grasslands that have presumably not been used as cropland due to site characteristics. These site characteristics include, for example, waterlogging and low groundwater levels, which do not allow cultivation as cropland. In fact, permanent grasslands with high losses contain an above-average number of marshes and gley soils (data not shown), which also have elevated SOC stocks due to high groundwater levels (Poeplau et al., 2020). Figure 12 of the land use history shown here should therefore not be misinterpreted to mean that plowing up grassland can lead to a reduction in SOC loss; the opposite is true. Rather, it shows that site characteristics and land use 34 Discussion history are closely intertwined and that both have an impact on recent SOC dynamics. SOC dynamics. In northwestern Germany in particular, there were massive reductions in the groundwater level in the last century in order to convert formerly marshy areas and moors to agricultural use. Even though many of these soils are now classified as mineral soils, some of them are very SOC rich, and their SOC stocks are certainly not yet in equilibrium and therefore tend to decline. These losses are unlikely to be offset by cultivation. Many of these soils are also characterized by wide C:N ratios and a sandy texture and were already highlighted as black sands in the initial BZE-LW (Poeplau et al., 2021; Vos et al., 2018). Accordingly, changes in SOC stocks in all three land use classes tended to correlate negatively with the initial C:N ratio, the mean SOC stock, and the sand content (tendency of SOC to decrease at sites with a wide C:N ratio, high SOC stock, and high sand content). These initial trends are consistent with the results of long-term observations in Lower Saxony, where cropland and grassland sites close to groundwater (sandy gley soils) showed the greatest SOC losses compared to less sandy soils further away from groundwater (Höper and Meesenburg, 2021). The highest losses of SOC in grassland soils shown in Fig. 15 also conceal the SOC -rich soils of the Netherlands, a region bordering Lower Saxony with comparable soil genesis. However, the high SOC contents of the predominantly northwestern German black humus sands may also have other causes and thus have different effects on the current dynamics. For example, sod cutting was very widespread in Lower Saxony, enriching the very nutrient-poor sandy soils with organic matter (Blume and Leinweber, 2004) . In addition, some of today's cropland was once covered by heathland, which may have left behind very stable organic matter in the soil (Springob and Kirchmann, 2010). A more detailed evaluation of the land use history, together with a characterization of the organic matter, will provide more information about the potential SOC dynamics of these soils in the future. 4) In addition to declining livestock numbers, mineral fertilization in Germany has also been gradually reduced over the past decades. Nitrogen surpluses in the soil have thus fallen from 177 kg per hectare of agricultural land in the 1990s to around 77 kg in the early 2020s (Federal Environment Agency, 2024). Overall, nitrogen inputs into agricultural systems and soils have also been reduced nationwide. These trends can also be reflected in the dynamics of SOC contents and stocks in cropland and grassland soils if the reduced N inputs mean lower yields and thus less biomass input into the soil. Various fertilization experiments on both grassland and cropland soils have shown that the SOC stock is linearly correlated with the amount of nitrogen fertilization and that approximately one kilogram of SOC is built up in the soil per kilogram of mineral nitrogen fertilizer (Kätterer et al., 2012; Poeplau et al., 2018) . Conversely, extensive agricultural production also carries the risk of a shrinking SOC reserve in favor of other positive environmental effects. A preliminary evaluation of 45 grassland questionnaires from the repeat inventory showed an average decrease in nitrogen fertilization (organic and mineral) of 32 kg N (-19%) compared to the initial inventory (data not shown). However, whether a reduction in N fertilization has an effect on humus depends crucially on its yield effectiveness. In the case of the reduction of high N surpluses that are harmful to the environment and climate, it can be assumed that the yield and organic matter effectiveness is low. The SOC dynamics of a soil are therefore influenced simultaneously by recent soil management, individual previous use or history, and changes in abiotic site characteristics (Heikkinen et al., 2013). A more in-depth analysis of the factors controlling the SOC dynamics of the BZE-LW sites was not possible at this point in time, or rather, it did not make much sense due to the still fragmented data situation. In the currently sampled population, the soils of Lower Saxony play a major role in terms of quantity, which, as described, are strongly influenced by their sometimes very specific history in terms of hydrology and land use. Fig. 10 clearly shows that the most extreme SOC changes also occurred in this region. It therefore remains to be seen whether the results obtained so far will be confirmed for the whole of Germany. The mean values Discussion 35 reported here are therefore of limited significance and should not be extrapolated as such. However, the fact that a negative trend in SOC stocks was found for all land uses and across different regions suggests that global warming is already having a negative impact on SOC stocks in agricultural soils in Germany. This has recently been counteracted in cropland soils, primarily through the increased cultivation of catch crops, while in grassland soils, changes in fertilization intensity and organic fertilization may have had an additional negative impact on SOC stocks (Poeplau et al., 2018). 5.2 An unscheduled course correction in the agricultural soil condition survey A very fundamental difficulty of large-scale soil monitoring over long periods of time is that many soil properties change relatively slowly and therefore relatively little per unit of time. The framework conditions for soil inventories, on the other hand, can change significantly from one iteration to the next. In order to be able to detect the small changes in soil properties with certainty, an exact and, in the best case, unchanging procedure in the field and laboratory is an important theoretical prerequisite, but rarely a reality. Changes in political, financial, organizational, content-related, or analytical conditions all too often lead to changes that can have a more or less significant impact on the quality of time series. Classic examples include changes in analytical methods, contract laboratories, or even individual devices (Even et al., 2025; Wollmann et al., 2025), the not entirely accurate relocation of sites (Heikkinen et al., 2020), variation in the number of sites (Poeplau et al., 2015), or even changes in sampling depth (Jones et al., 2024). Such serious changes should be avoided if possible or, if necessary, mitigated by corrective functions. While the latter is established practice for analytical methods, a directional sampling error can hardly be corrected. Even though the BZE-LW has always placed great emphasis on consistency and continuity, it is not free from systematic errors. During the evaluation of the initial BZE-LW, it was already established that the two methods used to determine SOC -contents and SOC stocks (profile pits, core drilling) led to different results. This is not surprising. It has already been shown elsewhere that the type of sampling alone can have a significant influence on the result of the analysis (Del Duca et al., 2025; Walter et al., 2016). However, the evaluation of the initial inventory did not reveal which of the two methods is less prone to error, as both approaches are known to have strengths and weaknesses. In addition, only some of the core drilling samples could be processed and measured due to resource constraints. Furthermore, core drilling is only of limited suitability for determining bulk density and stone content, ii) causes relatively extensive damage to standing crops due to the use of large equipment, iii) requires a greater amount of manpower, and iv) all other parameters were also determined from the profile pit samples, the repeat inventory was also carried out using profile pits. From the outset, care was taken to ensure that a representative sample was taken from each small test pit at the respective depth level. This was achieved by taking a uniformly thick slice of soil material from the profile wall over the entire length of the pit. In the initial inventory, this was also theoretically the case, but with one important restriction: at that time, an attempt was made to carry out coupled sampling of depth levels and diagnostic horizons so that both variants could be evaluated. However, additional samples were only taken at so-called intermediate depth levels when there was a distance of five cm or more between the horizon and depth level boundaries. If, for example, a plow horizon ended at 27 cm and the depth levels of the plow horizon to be sampled were 0-10 and 10-30 cm, the official task was to sample the lower three centimeters of the depth level to the same extent and to mix it with the 10-30 cm sample. The greatest deviations between the profile pit and core drilling in the field were indeed found precisely in the 10-30 cm depth level. In grassland, it is the 0-10 cm depth level where there is also a strong vertical gradient in the SOC content to the underlying soil material. Under certain circumstances, the inclusion of 36 Discussion horizon boundaries in depth-specific sampling may have led to unrepresentative depth samples in some cases. Such an error is ruled out in sampling with drill cores, as sampling was carried out precisely according to depth levels and no selection can be made during drilling as to where exactly the sample material is taken. Due to the fixed positions of the eight drill cores, this is a systematically random sampling (Brus and Saby, 2016) . The evaluation of the repeat inventory now suggests that sampling with drill cores does indeed provide the more stable and accurate SOC contents for the sampled plots. Even though the absolute directed deviation between profile pits and ram core sampling in SOC content in the most affected depth levels was only 1-2 g kg-1 i.e., about 0.1-0.2 percentage points SOC (Jacobs et al., 2018), this has a relatively strong impact on the SOC stock and its change. Therefore, this systematic deviation cannot be ignored and must result in an adjustment of the methodology. In the case of the evaluation presented here, it was decided to combine the advantages of both sampling systems of the initial soil inventory (soil profile, drill cores) in order to determine the stocks of organic soil matter. The drill cores from the core drilling were used to record the SOC contents at defined depth levels, while the dry bulk density was recorded at the central soil profile. As a direct consequence, the number of sites that could be evaluated for this interim report (sites with already analyzed drill cores) decreased, and it became necessary to initiate remeasurements of the samples from the core drilling. As shown in Fig. 5, in which there are only very weak correlations between the SOC changes based on profile pits and core drilling, there are also advantages to taking this time-consuming step . A single profile pit, even in a central position, cannot adequately represent a plot (now 12x12 m) due to known small-scale variability and, per se, introduces a high degree of random uncertainty into the data set. Even a slight positive deviation of the SOC content in the pit from the plot mean already increases the probability of a negative trend in resampling and vice versa (Slessarev et al., 2023). Spatial replication is therefore important even in the smallest plots (Poeplau et al., 2022). Spatial replication with subsequent analysis of all individual samples has the great advantage over a mixed sample in that the small-scale variability at the specific location is known and thus an uncertainty can be specified for each inventory run. This can be used to statistically validate measured changes at the site and thus support the interpretation of observations. It can even be used to develop specific sampling strategies. For example, the present evaluation has shown that the minimum detectable difference (MDD) in grassland soils tends to be higher than in cropland soils. One could conclude from this that the number of samples in grassland soils should be higher than in cropland soils. However, since the differences in MDD between individual sites within each land use are even more extreme than the differences between land uses, this is not necessarily expedient. Discussion 37 6 Outlook At this stage of the project, no representative picture of changes in soil properties for agricultural soils in Germany has yet emerged. Initial trends have been observed and possible causes identified. In addition to expanding and improving the data set, a main objective in the coming project phase will be to delve deeper into researching the causes and explaining changes in key soil properties. The separation of the influences of recent soil management, climate change, and site history on observed changes in soil properties plays a central role. The systematic use of the collected management data, as well as other external data (e.g., remote sensing products) in process models and statistical approaches will play an important role in this. Only by using a broad methodological spectrum can complex and overlapping spatial patterns be resolved in order to ultimately classify observed changes and incorporate them into political decision-making processes. The second phase of the BZE-LW repeat inventory (until 2030) also presents some particular challenges: This report has focused heavily on the parameter SOC. This parameter remains of central importance for current reporting requirements. However, a number of other parameters are also being collected that are more or less closely related to SOC and will provide additional insights relevant to practice, policy, and science on the development of agricultural soils. These include pH values, cation exchange capacity, and base saturation, which provide information about the nutrient supply and liming requirements of soils; air capacity of the subsoil and aggregate stability of the topsoil as structural parameters; and the quality of organic matter. The general decline in dry bulk density reported here (Table A4) must also be investigated and understood in more detail. A similar trend was observed for cropland and grassland soils in the French soil inventory (RMQS), but this has also not yet been explained (Munera-Echeverri et al., 2025). Due to the implications for SOC stock calculations, the aforementioned study suggested working with unchanged dry bulk density values in this case. In the next phase of the BZE-LW project, a comparison of different methods for calculating SOC stocks could provide information on the influence of variations in the handling of dry bulk densities on changes in SOC stocks. The Soil Monitoring Law is an important environmental policy innovation of the European Union, which aims to harmonize soil protection on a continental scale and strengthen it in a legally binding manner. At this point in time, it is uncertain how the Soil Monitoring Law will affect soil monitoring in Germany as a whole, and in particular the repeat inventory of the BZE-LW. Redensification and expansion of the grid, shortening of the sampling interval to six years, new parameters, and land use classes will increase the cost of soil monitoring in Germany, with existing systems in Germany forming an important basis. In addition to the Soil Monitoring Law, two further reporting obligations have recently been introduced for which the BZE-LW can and will provide data: the first is the German adaptation strategy to climate change, which has set concrete, measurable targets for the first time since 2024. One of the targets is to prevent SOC losses from German agricultural soils. An even more ambitious target has been set by the Nature Restoration Law: here, cropland soils should show an upward trend in SOC contents on average. 38 Outlook References Ad-Hoc-Ag Boden, 2005. Bodenkundliche Kartieranleitung, 5. E.Schweizerbart'sche Verlagsbuchhandlung, Hannover. Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop evapotranspiration-Guidelines for computing crop water requirements-FAO Irrigation and drainage paper 56. Fao, Rome 300(9), D05109. Çetinkaya-Rundel, M., Hardin, J., 2024. Bootstrap confidence interval for the mean paired difference., Introduction to modern statistics. OpenIntro. De Rosa, D., Ballabio, C., Lugato, E., Fasiolo, M., Jones, A., Panagos, P., 2024. Soil organic carbon stocks in European croplands and grasslands: How much have we lost in the past decade? Global Change Biology 30(1), e16992. Del Duca, S., Tondini, E., Vitali, F., Lumini, E., Garlato, A., Vinci, I., Tagliaferri, E., Brenna, S., Motta, S., Maillet, E., Bispo, A., Orgiazzi, A., Jones, A., Mocali, S., Fantappiè, M., 2025. Comparison of LUCAS and Italian Sampling Procedures for Harmonising Physicochemical and Biological Soil Health Indicators. European Journal of Soil Science 76(3), e70108. Destatis, 2023. Agrarstrukturerhebung Statistischer Bericht. Destatis, 2024. Landwirtschaftliche Betriebe mit Zwischenfruchtanbau im Zeitvergleich, https://www.destatis.de/DE/Themen/Branchen-Unternehmen/Landwirtschaft-ForstwirtschaftFischerei/Feldfruechte-Gruenland/Tabellen/zwischenfruechte.html. Destatis, 2025a. Grünland 2025: Erntemengen um 4 % niedriger als 2024, https://www.destatis.de/DE/Themen/Branchen-Unternehmen/Landwirtschaft-ForstwirtschaftFischerei/Feldfruechte-Gruenland/aktuell-gruenland1.html. Destatis, 2025b. Pflugeinsatz verliert in der Landwirtschaft an Bedeutung, https://www.destatis.de/DE/Presse/Pressemitteilungen/2024/05/PD24_212_41.html. Don, A., Drexler, S., Poeplau, C., 2024a. Indikatoren zur Bewertung von Humusgehalten für die Bodengesundheit. Bodenschutz (1), 8-15. Don, A., Seidel, F., Leifeld, J., Kätterer, T., Martin, M., Pellerin, S., Emde, D., Seitz, D., Chenu, C., 2024b. Carbon sequestration in soils and climate change mitigation—Definitions and pitfalls. Global Change Biology 30(1), e16983. Dorau, K., Bamminger, C., Koch, D., Mansfeldt, T., 2022. Evidences of soil warming from long-term trends (1951– 2018) in North Rhine-Westphalia, Germany. Climatic Change 170(1), 9. Drexler, S., Broll, G., Flessa, H., Don, A., 2022. Benchmarking soil organic carbon to support agricultural carbon management: A German case study#. Journal of Plant Nutrition and Soil Science 185(3), 427-440. DWD, C.D.C., 2025a. Raster der Monatsmittel der Lufttemperatur (2m) für Deutschland, Version v1.0., https://opendata.dwd.de/climate_environment/CDC/grids_germany/monthly/air_temperature_mean/. DWD, C.D.C., 2025b. Raster der Monatssumme der Niederschlagshöhe für Deutschland, Version v1.0. , https://opendata.dwd.de/climate_environment/CDC/grids_germany/monthly/precipitation/. DWD, C.D.C., 2025c. Raster der Monatssumme der Sonnenscheindauer für Deutschland, Version v1.0. , https://opendata.dwd.de/climate_environment/CDC/grids_germany/monthly/sunshine_duration/. Ellert, B.H., Bettany, J.R., 1995. Calculation of organic matter and nutrients stored in soils under contrasting management regimes. Canadian Journal of Soil Science 75(4), 529-538. Emde, D., Poeplau, C., Don, A., Heilek, S., Schneider, F., 2024. The centennial legacy of land-use change on organic carbon stocks of German agricultural soils. Global Change Biology 30(8), e17444. Even, R.J., Machmuller, M.B., Lavallee, J.M., Zelikova, T.J., Cotrufo, M.F., 2025. Large errors in soil carbon measurements attributed to inconsistent sample processing. SOIL 11(1), 17-34. Appendix 45 Figure A1: A) Difference in average annual temperature between the periods 1970-79 and 2020-24 for all BZE-LW locations; B) Linear slope of average annual temperature over the last six calendar decades. Source: Thünen Institute with data from the Deutscher Wetterdienst (DWD 2025a) 46 Appendix Figure A2: A) Difference in average annual climatic water balance between the periods 1970-79 and 2020-24 for all BZE-LW sites; B) Linear slope of annual mean temperature over the last six calendar decades. Source: Thünen Institute with data from the Deutscher Wetterdienst (DWD 2025a, b, c) Appendix 47 Figure A3: Classification of all sites as potentially arid (negative annual climatic water balance) and humid (positive annual climatic water balance) for the last six calendar decades. Source: Thünen Institute with data from the Deutscher Wetterdienst (DWD 2025a, b, c) Bibliografische Information: Die Deutsche Nationalbibliothek verzeichnet diese Publikationen in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet unter www.dnb.de abrufbar. Bibliographic information: The Deutsche Nationalbibliothek (German National Library) lists this publication in the German National Bibliographie; detailed bibliographic data is available on the Internet at www.dnb.de Bereits in dieser Reihe erschienene Bände finden Sie im Internet unter www.thuenen.de Volumes already published in this series are available on the Internet at www.thuenen.de Zitationsvorschlag – Suggested source citation: Poeplau, C., Harbo, L.S., Schneider, F., Schiedung, M., Don, A., Heilek, S., Dechow, R., Vasylyeva, E., Heidkamp, A., Prietz, R., Flessa, H. (2025) Interim report of the repeated German Agricultural Soil Inventory. Thünen Working Paper 277a. Johann Heinrich von Thünen-Institut, Braunschweig. https://doi.org/10.3220/253-2025-235 Die Verantwortung für die Inhalte liegt bei den jeweiligen Verfassern bzw. Verfasserinnen. The respective authors are responsible for the content of their publications. Thünen Working Paper 277a Herausgeber/Redaktionsanschrift – Editor/address Johann Heinrich von Thünen-Institut Bundesallee 50 38116 Braunschweig Germany [email protected] www.thuenen.de DOI: 10.3220/253-2025-235 urn: nbn:de:gbv:253-2025-000219-5 Title photo: Thünen-Institut/Ulf Schneidewind