2.1 LPJ-GUESS

LPJ-GUESS is a well-established, process-based ecosystem model designed for regional to global applications. Vegetation is modeled via plant functional types (PFTs) which represent the globally most abundant growth strategies. These can be distinguished in terms of e.g. growth form, phenology, life history strategy, allometry, photosynthetic pathway and a limited set of bioclimatic limits. Soil C and N dynamics are based on the CENTURY model [19] which contains eleven soil organic matter (SOM) and litter pools differing in their C to N (C:N) ratios and decay rates. A detailed description of the model is given by [16] while the N cycle implementation including the N allocation algorithm is described in [15]. Recent model developments include the incorporation of land-use change dynamics together with a crop module [14] which is based on approaches by [20] and [21]. [17] updated this version of the crop module by incorporating nitrogen dynamics and C-N interactions also for crops. Managed grasslands are represented in the standard version of LPJ-GUESS by removing 50% of the above-ground carbon [14]. This is meant to represent a 90% removal in intensively grazed pastures and a 50% re-entering of this carbon back to the litter pool as manure. The carbon allocation for grasses is done at the end of the year in the standard version of LPJ-GUESS.

To implement daily pasture management regimes and allow realistic feedbacks between management and vegetation, we incorporated the daily carbon allocation routine for natural C3 and C4 grasses developed by [22] into LPJ-GUESS (described above). The implemented functions and model modifications are based on theory from [23]. Carbon assimilated by photosynthesis on a daily time step is allocated dynamically to one root and four different shoot biomass compartments (growing leaves, first fully expanded leaves, second fully expanded leaves, senescing leaves), or is respired by autotrophic processes (see Fig 1). At the end of each simulation day, 10% of NPP is transferred into the reproduction pool while the rest is separated into leaf storage, roots and start storage. Thereby, the start storage enables growth when the grass has a low LAI but the conditions are favorable. Carbon moves from leaf storage to the four biomass compartments differing in leaf age classes based on a temperature dependent growth factor and daily phenology. The daily phenology is calculated as the minimum value of the ratio between water supply and water demand for full leaf cover, and the ratio between growing degree days above 5 degree Celsius (gdd5) and the lifeform specific gdd5 sum for full leaf cover. The movement between the four shoot biomass compartments is controlled by a PFT specific transfer rate and temperature.

G denotes the carbon flux from leaf storage while G_s denotes the carbon flux from the start storage.

After incorporating daily carbon allocation for pastures in the crop and land-use version of LPJ-GUESS, we added the management routines grazing, mowing and fertilization with both mineral N and manure on a daily basis. We assume that in a given grid cell, grasslands are either cut or grazed, and do not consider mixed management (mowing and grazing).

Mowing can be triggered either by providing specific harvest dates or is calculated dynamically as a function of LAI, and occurs every 30 days or more. After mowing, the LAI value is decreased to a fixed value of 0.5. Immediately after each of the first three cuts in a year, pastures are fertilized with mineral N and manure. We assume that mineral N fertilizer is distributed equally to the three mowing events (33% each). Manure is only applied after the first cut. 100% of harvest goes to the atmosphere while 10% of biomass not harvested goes to litter (biomass loss during harvest).

Grazing is simulated only at LAI values above 0.5. In case of LAI values below this threshold, the simulated grazing stops. Grazing resumes again when simulated LAI becomes greater than the threshold value. This approach is similar to [24] who used a threshold value of shoot biomass to determine the grazing period. For grazed pastures, a regular, relatively high daily grazing intensity of 2.5% of foliage was assumed. This means pastures are grazed close to their carrying capacity. More extensively grazed grasslands are not simulated in this study but this can easily be done by lowering the grazing intensity. Of each daily feed, 60% of the carbon (C) is assumed to be lost to the atmosphere directly, 15% C is incorporated into the animals body mass while the remaining 25% C return to the paddock in dung and urine (similar to [25, 26]). For N, 75% N is returned to the paddock as part of dung and urine [27–29] while the remaining 25% N is incorporated into the animals body mass. C and N incorporated into the animals bodies is moved to the slow carbon pool with a turnover rate of five years (representing both C and N stored in animal biomass during lifetime). Grazed pastures are fertilized three times per year, every two months with mineral N and manure.

Gridded mineral fertilizer and manure nitrogen application rates for European grasslands in the European Union (EU27) were estimated by the Common Agricultural Policy Regionalized Impact analysis (CAPRI) model (see [30, 31]) based on combined information from official and harmonized data sources such as EUROSTAT [32] and FAOstat [33]. It was spatially dis-aggregated using the methodology described in [34]. The data were estimated at a spatial resolution of 1 km and were re-aggregated here to a spatial resolution of 0.5°. For French regions, more detailed data from the French national statistics were used [35]. We used a set of rules to rebuild the temporal evolution of gridded nitrogen fertilization from 1901 to 2010. First, organic fertilizer was assumed to have remained constant over time (due to the lack of statistical data). Second, the application rate of mineral fertilizer evolved with time following the total mineral nitrogen fertilizer consumption of the European Union [36]. Third, mineral fertilizers were set to be applied since 1951, and application rates linearly increased from 0 to the observed level of 1961 during the period 1951-1960.

Manure is represented by an increase in the metabolic and structural soil organic matter (SOM) pool with a C to N ratio (C:N) of 30. This value has been chosen to represent the C and N content from sources ranging from poultry waste (C:N of ca. 15) to straw-rich manure from livestock (C:N of 40 or more). Since both metabolic and structural SOM pools have different turnover rates, N derived from manure becomes available for an extended period in the soil. Besides nitrogen fertilizer application, nitrogen deposition and nitrogen fixation by soil microorganisms were considered as nitrogen addition as well (see [15]). This means that for the grid cells with no fertilizer application, there were potentially still nitrogen inputs by deposition and fixation.

To account for changes in land-use intensity in the future, we used data from [37] who projected intensive and extensive use of pastures until 2040. The authors used grazing intensity of cattle, goats, and sheep [38] as a proxy for nitrogen inputs on pastures as suggested by [39]. This data were then disaggregated and reclassified into two classes, which were used as a proxy for low and high grassland intensity. More information can be found in [37]. We aggregated this data to 0.5° and assumed an increase in N application of 50 kg/ha (mean N application calculated over Europe) when a gridcell changed from extensive to intensive. We reduced N application by 50 kg/ha for the opposite change. For the missing years, we conducted a linear interpolation.