The main goal of Project MARVIC is to develop accurate, context-specific, and cost-effective monitoring systems. In this sense, our approach involves strategically combining various data sources, models, and remote sensing methodologies. This framework will be implemented in 26 test cases across 12 European countries, each linked to a specific Land Use and Soil Type (LUST) such as arable land on mineral soils, grassland on mineral soils, managed peatland, and agroforestry/woody crops. Specifically, the use of models requires calibration, validation, and accuracy testing, which can be conducted using data from benchmark sites where management practices are documented and soil and crop properties are measured.

Test cases have varying numbers of benchmark sites available, which are generally located within areas that share the same pedoclimatic conditions (i.e. climatic and geological) as the test case but can also belong to other regions, depending on the methodology used and the availability of suitable datasets. The benchmark sites differ in nature, as they may correspond to long-term experiments (LTEs), well-monitored sites at experimental farms, or networks of sites, such as national soil monitoring schemes and eddy covariance sites (e.g., ICOS). 

In this context, Project MARVIC has developed a metadata dataset of benchmark sites relevant to the calibration and validation of models used within the MARVIC test cases. Knowing which variables have been measured at each site is crucial for determining which models can be run on that dataset and which ones cannot due to missing input data. The dataset has been compiled mainly by assembling a comprehensive list of the data required to run and test the models with which the MARVIC consortium has experience, such as soil models (e.g., C-tool and Roth-C) and ecosystem models (e.g., Agricarbon-EO and Landscape DNDC). 

The dataset comprises nearly one hundred columns that contain features or metadata for each individual benchmark site. The information can be broadly categorized into three sections: i) an overview of the benchmark dataset and its characteristics, including a detailed summary of the pedo-climatic conditions, sampling strategy and data sharing possibilities of the benchmark dataset; ii) the carbon farming practices that have been implemented in the benchmark site, such as cover crops, crop rotation, paludiculture, and peatland rewetting; and iii) all the measured variables that can be used in the modelling process and whether they have been collected or not in the benchmark site dataset (e.g., soil organic carbon, pH, biomass, gas fluxes, etc.).

To date, MARVIC has identified 73 benchmark sites relevant to the calibration and validation of models used within the test cases, covering the four LUSTs (41 for arable land, 12 for grassland, 6 for peatland, and 14 for agroforestry) and fairly distributed among plot, field, farm, regional, and national sites. While the dataset has been created for internal use, it will be made publicly accessible once approval is received from the European Commission, and it will be updated throughout the project as additional benchmark sites are identified. Follow this link to find the full report.