Methodological Standards to Detect Forest Soil Carbon Stocks and Stock Changes at Landscape Scales.
Rainer Baritz1, Dietmar Zirlewagen2, Eric Van Ranst3, Mats Olsson4, Robert Jandl5, Pere Rovira6, Juan Romanya6, Christian Wirth7, Maria Erlandsson8, Zoltan Somogyi9, Carly Green10, Mike Starr11, and Pekka Tamminen11. (1) Federal Institute for Geosciences and Natural Resources, Stilleweg 2, Hannover, Germany, (2) INTERRA, Kenzingen, Greece, (3) Ghent University, Ghent, Belgium, (4) SLU, Uppsala, Sweden, (5) BFW, Vienna, Austria, (6) University of Barcelona, Barcelona, Spain, (7) MPI BGC, Jena, Germany, (8) Genth University, Ghent, Bermuda, (9) JRC, Ispra, Italy, (10) University College Dublin, Dublin, Ireland, (11) METLA, Helsinki, Finland
For the first time, the main results of a European research project on soil carbon inventories (CarboInvent, work package 3, http://www.joanneum.at/CarboInvent/) which is currently being finalized will be presented. The objective was to assess the capacity of existing national and regional soil inventory schemes to provide baseline soil C assessments and to detect soil C changes. In order to fulfil these objectives plot level errors and errors related to the upscaling of plot data were carefully investigated. A thorough analysis of representativity, and the variability received for different grid densities have been assessed. A sampling campaign has been conducted to fill remaining representativity gaps. All steps have been accompanied by tracking all uncertainty related aspects. Suggestions were made, and guidance will be provided to assess not only the commonly applied random statistical error (spatial variability and plot density), but also systematic errors (inventory “quality”). The upscaling methods tested include geomatching combined with class matching, and regression supplemented by spatial analysis (kriging). Both approaches were compared. The data base for the geo-/class matching consists of 16x16 km inventory data from the ICP Forests Forest Soil Condition Monitoring (Europe (N=5,269 plots; and for selected test countries, Ireland = 22 Finland = 442, Spain = 452, Germany = 414 Austria = 131, Sweden = 1,249 plots). For the regression kriging, test countries were selected with condensed data sets (mostly ca. 8x8 km: Germany = 1,800, Austria = 513, in the case of Sweden: N = 8,581 plots), and one test region (Thuringia = 250; roughly 4x4 km). The quality of upscaling can be greatly improved using geomorphographic landscape analyses of digital elevation models (50m in the test area, 250 m for the test countries) combined with various site information such as land cover, climate, and soils. The method allows tracking the performance of the most influential predictors responsible for the accumulation of soil carbon, which is regionally different. The evaluations were conducted separately for the forest floor and the mineral soil at varying total depths depending on data availability in the inventories. Different approaches to stratification have greatly improved the predictive value of the regional models. The results are highly relevant for (a) improving the assessment of C sequestration at landscape scale (e.g. by focussing on hot spots), (b) determine the capacity of the existing inventory schemes to detect change, (c) identify data gaps related to carbon change assessments and model input, and (d) identify research needs in the area of data availability for modelling and to better related to forest ecosystem level investigations (carbon cycle; forest management). Since forest vegetation pattern and liming maps (test area) have been considered in the evaluations, the relevance of the developed approaches allow the projection of management effects on soil carbon.