An EO framework for soil Carbon Sequestration Monitoring

At the COP21 in Paris an agreement was reached to combat climate change and to work to a sustainable low carbon (C) future. The Paris agreement requires that all countries need to make a national action plan, which includes regularly reporting on their emissions and on their implementation efforts.

For example, the  Netherlands  wants to reduce the greenhouse gas (GHG) emissions by 55% by 2030, compared to 1990 levels, and a 95% reduction by 2050. To achieve these goals, the Dutch government made agreements with the various national participating sectors and their contribution. In the Dutch Climate Agreement a specific target for carbon sequestration in mineral agricultural soils was included, which aims for 0.5 Mton CO ₂  additional sequestration per year by 2023 ( source ).

Mineral soils are reported under the land use, land use change and forestry (LULUCF) sector in the national inventory report on the greenhouse gas emissions. Van Baren et al. (2024) provides a detailed description of the methodologies, activity data and emission factors that were used ( current state ).

This EO4CSM project specifically focusses on the monitoring of the carbon stock changes in mineral soils of cropland and grassland. The aim is to modify the current methodology and improve the national monitoring of carbon sequestration of agricultural soils for the Netherlands, mainly by using more Remote Sensing (RS) data and products (see  goal  and  objectives  EO4CSM).

Good monitoring is important to achieve the goal of sequestering an additional amount of 0.5 Mt CO ₂  per year in Dutch mineral agricultural soils. Monitoring of soil carbon stocks in the Netherlands is done in two ways ( Slim Landgebruik ):

1. Measurements of the soil carbon stock
2. Model calculations of the changes in soil carbon stock

In 2018-2019 the soil carbon stock has been measured in agricultural soils within the CC-NL network (Tol-Leenders et al., 2019). The CC-NL dataset (1152 locations) serves as a baseline measurement to observe future changes in soil carbon stocks, but is also used as input for model calculations.

To observe a trend and see the effect of the measures, a large number of measurements over a long period of time are required. A next campaign will take place in autumn  2024 .

The soil stores about two times as much carbon compared to the atmosphere and tree times as much in vegetation. Due to the large carbon pool in our soil, it is difficult to measure changes in the soil carbon stock over a short time. Therefore, models are also used to simulate changes in carbon stock ( Slim Landgebruik ). One of the models currently used in the Netherlands is RothC (and is also used for the Dutch LULUCF reporting). This model not only calculates the carbon stock changes under current soil management, but it can also be used to determine the potential carbon sequestration when certain measures are applied.

 RothC  is a model for the turnover of organic carbon in non-waterlogged topsoil that allows for the effects of soil type, temperature, moisture content and plant cover on the turnover process (Coleman and Jenkinson, 2014). RothC is scientifically acknowledged and applied worldwide at various scales and has been extensively calibrated and validated in long-term experiments. The model is relatively simple, easy to use, and requires little input data (Lesschen et al. 2020).

The RothC model distinguishes five carbon pools: Resistant Plant Material (RPM), Decomposable Plant Material (DPM), Microbial Biomass (BIO), Humified Organic Matter (HUM), and Inert Organic Matter (IOM) (Coleman et al., 2024). Each of these compartments has its own specific decomposition coefficient (the decomposition is a fraction of the amount present), except for the IOM compartment in which organic matter is no longer broken down. The decomposition coefficient for each compartment is influenced by soil texture, temperature, moisture and soil cover.

At this moment, RothC is operated at 4-digit zip code level (PC4, ~3400 units) for the Dutch National Inventory (LULUCF) Reporting (Van Baren et al., 2024).

The necessary input data is collected from different sources and available at various levels (national/regional):

• Monthly soil cover
• Crop areas of grass & arable land [ha]
• Annual carbon inputs from crop or grass residues, green manure crops, compost and organic manure [ton C ha ^-1 ]
• Soil data including the initial soil organic carbon content (SOC ₂₀₁₈  ; Tol-Leenders et al., 2019), clay content (%) (Helfenstein et al., 2024), soil type (De Vries and Onderstal, 2008) and soil depth (cm).
• Climate data (monthly rainfall, evaporation and air temperature) ( KNMI )

The RothC results (at PC4 level) on the right show clear differences between the Soil Organic Carbon (SOC) balance in clay soils (dominantly positive balance) and in sandy soils (dominantly negative balance). Grassland has, in comparison to most arable crops, a more positive effect on the SOC balance, because of the relatively high carbon inputs from (root) residues and higher manure inputs. Besides, it has an all-year-round soil coverage (unless it is renewed), which results in less soil disturbance (and thus less CO ₂  emissions).

The model is quite sensitive to climate conditions, because soil moisture and temperature influence the decomposition rate (Lesschen et al., 2020). All years except 2021 were relatively dry, especially during the growing season. This resulted in a lower mineralization of the soil organic matter, while the carbon inputs were more or less the same. Therefore, the overall carbon balance was mostly negative in 2021.

In the current situation, RothC uses monthly vegetation cover values [binary: bare soil or fully covered], which are based on a national crop calendar that contains general start and end for each crop grown in the Netherlands. Within EO4CSM, we will replace these with EO-based monthly crop cover values at parcel level, based on Sentinel-2 (S2).

The S2-data not only introduces much more spatial detail, but it also allows variation between parcels with the same crop, and annual variations (dry-wet years), as is demonstrated in the figure on the right.

In the current situation, RothC uses cover crop (also known as 2 ^nd  crop or catch crop) information that is based on information provided by farmers via the parcel registration system (LPIS, known as BRP in the Netherlands), and which is aggregated per crop type to PC4 level.

Within EO4CSM the presence of a cover crop will be based on S2-data (and LPIS). Furthermore, the start (SOS) and end (EOS) of the cover crop season is determined (see figure).

Like vegetation cover, the S2-based cover crop information introduces spatial variation between parcels. Furthermore, the EO-data contains more information than what is available through LPIS.

In the current situation, annual regional statistics of crop production (or crop yields, from  CBS ), is used to estimate the input of organic matter from crop residues.

There are so-called dry matter productivity (DMP) maps available based on EO-data, e.g. the 10-daily 300m DMP product of the Copernicus Land Monitoring Service ( CLMS ). This DMP product is based on a Light Use Efficiency (LUE) model, driven by meteorological data (radiation and temperature), land cover information, fAPAR, and several other parameters. fAPAR corresponds to the fraction of photosynthetically active radiation absorbed by the green elements of the canopy and can be derived from satellite observations such as Sentinel-3. However, the CLMS-DMP product is too coarse to be used at parcel level.

In EO4CSM we will follow a similar approach as the CLMS-DMP product, however we will use fAPAR data at parcel level, and estimate crop production of both the main crop and cover crop by using the estimated start (SOS) and end (EOS) of the crop seasons (limited to CC-NL parcels). The crop simulation model used for generating the biomass estimates is related to the LINTUL crop simulation model (Light INTensity UtiLisation), which describes the daily increase in crop biomass through a simple light use efficiency approach (Shibu et al. 2010, Berghuijs et al. 2023). LINTUL has been applied for many different crops including cereals, potato and grasslands. However, the algorithm has been adapted to allow for a hybrid model that combines the satellite inputs with a daily time-step crop simulation model. The hybrid model used to generate the biomass estimates is called CSSF (Crop Simulation by Satellite-derived fAPAR).

The figure below shows the comparison between CBS statistics of crop yield (dry weight in kg/ha) at provincial level and parcel-level estimates of crop yield based on the CSSF model, for the CC-NL parcels and years 2018 to 2022.

The figure on the right and the map below show the first results of the improved monitoring framework, where we have downscaled the calculations of the National Inventory from PC4 level (~3400 units) to field-scale calculations (~520.000 parcels) for the years 2018 to 2022.

The soil data (clay content and soil organic matter content) were assessed at field level using national soil and SOC maps. The downscaling also allowed the crop and cover crop combination per field. Although quite some inputs could be downscaled using LPIS and EO data, some assumptions were still needed. Regional average manure application rates were used as annual manure input, and the fraction of crop residues that are left in the field are based on national values.

Based on these changes, we were able to assess the carbon balance per field. While the overall pattern look similar to the PC4 level results, the more detail confirms the (in general) more negative SOC balance of arable fields. In 2018 the carbon balance is more positive compared to the other years. In this year, a severe drought took place during the growing season. As mentioned before, the drought caused a lower mineralization during the summer period. The positive balance might be overestimated for grassland as the grass carbon input is currently a fixed value, independent of yield/weather.

The next step is to integrate parcel level EO-information information as described above, which include monthly vegetation cover, the presence of cover crops and grassland renewal activities. Finally, we will also integrate parcel level crop production estimates (using the Light Use Efficiency-model) of main and cover crops (this is limited to the ~500 CC-NL parcels).

Viewer: SOC balance 2022 at parcel level.

Berghuijs, H.N., Silva, J.V., Rijk, H.C.A., van Ittersum, M.K., van Evert, F.K. and Reidsma, P. (2023) Catching-up with genetic progress: Simulation of potential production for modern wheat cultivars in the Netherlands. Field Crops Research, 296, p.108891.

Coleman, K. and Jenkinson, D.S. (2014) RothC-26.3 – A model for the turnover of carbon in soil. Evaluation of Soil Organic Matter Models: 237-246.

Coleman, K., Prout, J.M. and Milne A.E. (2024) RothC - A model for the turnover of carbon in soil - Model Description ( link )

Hendriks, C., Lesschen, J. P., Timmermans, B., Hanegraaf, M., Dijkman, W., Rougoor, C., Cruijsen J. and Schepens, J. (2023) Description and development Soil Carbon Tool. ( link )

Lesschen, J. P., Hendriks, C., Linden, A. van de, Timmermans, B., Keuskamp, J., Keuper, D., Hanegraaf, M., Conijn, S., and Slier, T. (2020) Ontwikkeling praktijktool voor bodem C.  https://doi.org/10.18174/517746 

Schepens, J. A. B., Timmermans, B., Moinet, G. Y., & Koopmans, C. (2023). Evaluating carbon sequestration of different alternative management practices in the Netherlands. In 2023 Book of Abstracts: Wageningen Soil Conference: Working together on solutions for a sustainable world (pp. 60-60). Wageningen University & Research.

Shibu, M.E., Leffelaar, P.A., Van Keulen, H. and Aggarwal, P.K. (2010) LINTUL3, a simulation model for nitrogen-limited situations: Application to rice. European Journal of Agronomy, 32(4), pp.255-271.

van Baren, S. A., Arets, E. J. M. M., Hendriks, C. M. J., Kramer, H., Lesschen, J. P., & Schelhaas, M. J. (2024). Greenhouse gas reporting of the LULUCF sector in the Netherlands: Methodological background, update 2024 (No. 255). WOT Natuur & Milieu.  https://doi.org/10.18174/648278 

van Tol-Leenders, D., Knotters, M., de Groot, W., Gerritsen, P., Reijneveld, A., van Egmond, F., Wösten, H., and Kuikman, P. (2019) Koolstofvoorraad in de bodem van Nederland (1998-2018).  https://doi.org/10.18174/509781 

Slim Landgebruik ( https://slimlandgebruik.nl/ )