BIOME-BGC is a bio-geochemical model capable of estimating the water, carbon and nitrogen fluxes and storages of terrestrial ecosystems. Previous research demonstrated that, after proper calibration of its ecophysiological parameters, the model can reproduce the main processes of Mediterranean forest types. The same investigations, however, indicated a model tendency to overestimate woody biomass accumulation. The current paper aims at modifying BIOME-BGC ecophysiological settings to improve the simulation of the woody compartment in Mediterranean forests. The modified ecophysiological parameter is the whole-plant mortality fraction (WPMF), which directly affects the amount of woody biomass stored. The optimal WPMFs of six main forest types in Tuscany are identified by forcing the model to reproduce the maximum standing volumes found in regional and local forest inventories. The effects of this operation are evaluated by comparing the model outputs produced using the original and modified settings to independent measurements from national forest inventories. The results obtained demonstrate the effectiveness of the modifications introduced and consolidate the methodological basis for extending the use of the modeling strategy to other Mediterranean areas.
In the last decades, bio-geochemical models have been used to quantify forest carbon dynamics and monitor the main fluxes and stocks within terrestrial ecosystems (
These problems are partly overcome by the BIOME-BGC model (
Recent investigations of our research group showed that BIOME-BGC can be efficiently integrated within a more general modeling strategy for the effective simulation of both water and carbon fluxes and storages in Tuscany forests (
The current paper presents an approach to overcome this limitation based on the modification of the model parameter settings which regulate these processes. Specifically, a model parameter which controls biomass storage in woody tissues is adjusted to reproduce the maximum standing volumes measured by regional and local forest inventories. The results of this modification are evaluated by comparison with measurements of forest cover (FC) and current annual increments (CAI) derived from independent inventory sources.
The paper first provides a description of the study region (Tuscany, Central Italy) and of the input data-layers. The methodology section then introduces the modeling strategy based on BIOME-BGC and the procedure applied to improve the model parameter settings. The paper is concluded by the presentation of the results achieved and by a discussion section.
Tuscany extends over about 9°-12° E long., 42°-44° N lat. The topography of the region ranges from flat areas near the coast-line and along the principal river valleys, to hilly and mountainous zones towards the Apennines chain. Approximately 2/3 of the region is covered by hilly areas, 1/5 by mountains and only 1/10 by plains and valleys. From a climatic viewpoint, the climate ranges from typically Mediterranean to temperate warm or cool according to the altitudinal and latitudinal gradients and the distance from the sea (
Forests cover approximately half of the regional surface and are mainly located over hilly and mountain areas (
A regional forest inventory (IFT) was carried out in Tuscany during the years 1990-1998 (
A more recent national forest inventory (INFC), carried out during the years 2000-2008, provided similar data aggregated on a regional basis for all main forest types in Italy (
Additional information was derived from CONECOFOR, a project carried out all over Italy aiming at monitoring the status of forests (
Daily meteorological data were derived from the Tuscany weather network for the years 1996-2008. Distinctively, daily maximum and minimum temperatures and daily total precipitation were collected from 139 and 179 stations spread all over the regional territory, respectively.
A Digital Terrain Model (DTM) of the study region with 1-km spatial resolution was derived from the Regional Cartographic Service of Tuscany. The same Service provided the digital forest map produced by
A regional map of stem volume was produced by
Soil information was derived from the Soil Map of Tuscany (1:250.000), produced by the Regional Government during the years 2000-2006 (http://sit.lamma.rete.toscana.it/websuoli/). This map provided soil depth and texture expressed in sand, silt and clay percentages.
BIOME-BGC is a bio-geochemical model developed at the University of Montana (
The model requires daily meteorological data (minimum and maximum temperature, precipitation, humidity and solar radiation), and a description of site characteristics (
Unlike its precursor FOREST-BGC (
BIOME-BGC works through a spin-up run which enables to find a quasi-equilibrium condition (
To address this issue,
where
where
where
As can be easily understood, the maximum stem C simulated by BIOME-BGC directly affects the estimated NVA and FCA and, consequently, the predicted NPPA. Hence, its correct estimation for each FT is crucial for an accurate prediction of actual forest structure and functions, and, particularly, for the simulation of net carbon fluxes.
The model parameter settings provided by
To address this issue the BIOME-BGC parameter settings of the six Tuscany forest types were further modified. In particular, attention was focused on the whole annual plant mortality fraction (WPMF), which represents the fraction of the above- and below-ground ecosystem carbon pools that are removed and sent to the litter compartments over the course of each simulation year. By this mechanism woody material (live and dead) does not accumulate within the tree pools and becomes available for the decomposition process. This mortality rate includes both natural tree mortality and mortality due to natural disturbances such as wind-throw. According to the BIOME-BGC logic, this parameter influences the amount of woody biomass that is accumulated yearly (
The application of the modeling strategy to Tuscany forests required the use of daily meteorological data (minimum and maximum temperatures, precipitation and solar radiation) for each 1-km2 pixel. To this aim, the daily temperature and rainfall data were extended from the weather stations to the regional surface by means of the DAYMET algorithm (
After this first run, the maximum standing volumes which are reasonable for the six forest types in Tuscany were identified using the point measurements of IFT and of the other two local forest inventories (San Rossore and Vallombrosa). In particular, the maximum volumes found by these inventories for each FT were conservatively considered to be 90-95% of the volumes which could be potentially sustained by the corresponding sites. The WPMFs of the six forest types were therefore iteratively modified until the standing volumes predicted by
The effect of changing the BIOME-BGC WPMF settings was evaluated by comparing the accuracy of the actual forest cover (FCA) and current annual increment (CAI) estimates before and after the modification against independent regional measurements.
As regards the first test, FCA estimates were computed by feeding
For the second test, the NPPA estimates of the six FTs were calculated by feeding
where
Both comparisons were made using the available aggregated data for the six forest types of Tuscany. The overall estimation accuracy over the six FTs was summarized by means of common statistics (
The last column of
As regards the carbon stocks estimates, they only partly follow these GPP patterns, due to the different respiration and allocation of the six forest types (
In almost all cases the new WPMFs identified by the current investigation (
The mean FCA and CAI values of the six forest types measured and estimated before and after the WPMF modification are shown in
The comparison between measured and estimated CAIs provides similar results. The model with the original WPMFs strongly underestimates INFC CAIs for FT 3, 4 and 6. This problem is partly corrected by the use of the new WPMFs, with the highest improvement obtained for FT 3. As a consequence, the new WPMF settings lead to a substantial reduction of the mean errors: RMSEold = 2.24 m3 ha-1 year-1 and MBEold = -1.47 m3 ha-1 year-1 against RMSEnew = 1.65 m3 ha-1 year-1 and MBEnew = -0.17 m3 ha-1 year-1.
BIOME-BGC has been shown to be applicable to simulate the behaviour of a wide variety of forest ecosystems, both in terms of eco-physiological processes and accumulated carbon pools (
The current paper proposes a more sound approach to address this problem. This approach is based on the modification of a major parameter setting which controls carbon accumulation in the vegetation compartment, WPMF (
The application of the current method yields new WPMFs which are not descriptive of the ages reachable by the six forest types in all conditions. In fact, the WPMFs identified are related to the average turn-over times which would characterize the existing forests in the absence of disturbing factors. Consequently, these WPMFs are dependent on local environmental factors (fertility, climate, growth form, etc.), and are not directly exportable to all other cases.
As expected, the use of new WPMFs does not modify the original LAI, GPP and respiration estimates but leads to substantial changes in accumulated stem carbon. The new volume averages of the six forest types are mostly lower than the original values and more consistent than these with the estimates provided by different sources (
The WPMF modification results in a marked improvement in the reproduction of the carbon stocks stored in the woody ecosystem compartment. This improvement is decisive in enhancing the assessment of NVA and FCA. As a consequence, the modification implies a substantial enhancement in the estimation of net production processes (
These results are particularly relevant when considering that all model simulations have been performed over forests which are very heterogeneous due to the variable climate and site factors and to the different management practices which characterize Tuscany environments. It can therefore be concluded that the use of the new BIOME-BGC versions can significantly enhance the capacity of the modeling strategy to simulate carbon stocks and fluxes across a variety of Mediterranean forest ecosystems.
The work was partially carried out under the C_FORSAT project “Modelling the carbon sink in Italian forest ecosystems using ancillary data, remote sensing data and productivity models” (national coordinator: Prof. G. Chirici), funded by the FIRB2008 program of the Italian Ministry of University and Research (RBFR08LM04).
The authors thank Dr. M. Moriondo, Dr. L. Fibbi and Prof. M. Bindi for their assistance in the calibration and application of BIOME-BGC in Tuscany, and Prof. P. Corona for his precious comments on the subject of the paper.
Map of Tuscany forests grouped into 6 main forest types (FTs - FT 1: evergreen oaks; FT 2: deciduous oaks; FT 3: chestnut; FT 4: beech; FT 5: lowland conifers; FT 6: mountain conifers), superimposed on a digital illumination model.
Simplified scheme of the modeling steps followed to obtain CAI estimates starting from the input data.
Comparison between the reference volumes of the 6 FTs and the volumes estimated using the original and the new WPMFs.
Comparison between actual forest cover (FCA) measured and estimated by the old and the new versions of BIOME-BGC. (*): significant correlation, P< 0.05.
Comparison between current annual increments (CAI) measured and estimated by the old and the new versions of BIOME-BGC. (*): significant correlation, P< 0.05.
List of the six forest types (FTs) considered with indication of the biome types and of the relevant GPP averages simulated by BIOME-BGC in Tuscany.
ForestType | Dominantforest species | Biome type( |
GPP(g C/m2/year) |
---|---|---|---|
1 | Evergreen oaks | Evergreen broadleaf forest | 1238 |
2 | Deciduous oaks | Deciduous broadleaf forest | 1288 |
3 | Chestnut | Deciduous broadleaf forest | 1170 |
4 | Beech | Deciduous broadleaf forest | 1016 |
5 | Lowland conifers | Evergreen needleleaf forest | 1378 |
6 | Mountain conifers | Evergreen needleleaf forest | 1200 |
Definitive BIOME-BGC parameter settings. Only the model parameters changed with respect to default values are reported. The parameters controlling conductance reduction (leaf water potential and vapour pressure deficit) were reduced to 90% of the default values for the first three ecosystems and were left unchanged for the others. The last column reports the new WPMFs, which replace the default settings (0.005 1 yr-1) proposed by
Ecosystemtype | Maximum stomatalconductance (m s-1) | Fraction of leaf N in Rubisco | New WPMF(1 yr-1) |
---|---|---|---|
1 | 0.0016 | 0.029 | 0.00625 |
2 | 0.0020 | 0.090 | 0.01 |
3 | 0.0023 | 0.078 | 0.0125 |
4 | 0.0045 | 0.090 | 0.01 |
5 | 0.0024 | 0.022 | 0.005 |
6 | 0.0032 | 0.027 | 0.01 |