![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
How certain are coniferous forest SOC estimates? A comparison of model simulations and measurements at regional scales |
VerfasserIn |
Carina Ortiz, Erik Karltun, Jari Liski, Göran Ågren |
Konferenz |
EGU General Assembly 2011
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 13 (2011) |
Datensatznummer |
250052679
|
|
|
|
Zusammenfassung |
Soil organic carbon (SOC) in boreal forests is a considerable carbon pool and small changes
in this pool have the potential to affect the overall national carbon balances for countries with
large forested areas. Sweden report annual SOC changes in forest soils to the UNFCCC and
the Kyoto protocol using data from a repeated soil inventory. Another method of predicting
change in SOC is process-based models. Determination of a small change in a large pool over
a long time span is associated with the risk that the observed change is either non-significant
due to random errors or erroneous due to systematic errors. This leads to difficulties
in reporting whether forest soils are sinks or sources of CO2 to the atmosphere.
Process-based models are often calibrated to few intensively studied sites which leads
to uncertainties in SOC change estimates when up-scaling model predictions. In
addition, the driving variables of the models are only available at large scales and
applying them at small scales is associated with uncertainties, which lead to additional
uncertainty in SOC change predictions. Here we present an analysis of uncertainty
sources in SOC stock estimations and how the variability in litter input and climate
affect the SOC changes in models. Two models, Yasso07 and Q, were used in the
comparison with the inventory data between 1994 and 2000. The analysis included model
calibration and validation. The Q model was calibrated with Generalized Likelihood
Uncertainty Estimation (GLUE) at county scale and both Q and Yasso07 were validated
regionally.
Both model and inventory estimates result in quantitatively substantial, but comparable,
uncertainties in SOC change estimations aggregated at the national level. From the
inventory the average change in SOC between 1994 and 2000 was estimated to 2
(±41)
Tg yr-1. The corresponding estimate with the Q model was 4 (+9;
-92)
Tg yr-1, and with the Yasso07 model 1(±81) Tg yr-1. However, the uncertainties in models
and inventory SOC estimates are conceptually not comparable since the sources of the
uncertainties differ. The major sources of uncertainty in modeled SOC estimates arise from
litter input estimations and parameter uncertainties. Uncertainties in inventory estimates are
aggregated from several sources and the main uncertainty is caused by spatial variation in
SOC. Both simulated and inventory SOC changes result in considerable inter-annual
variation. Inventory SOC change estimates vary between 0.5 and 3 Tg between years, Q
model estimates between 1 and 5 and Yasso07 between 0 and 9 Tg C. The inter-annual
variations also originate from different sources depending on method used. Inter-annual
variations in our simulations are generated mostly by climate variability while the
inter-annual variability for measurements could be due to several different reasons.
Inter-annual variation and confidence intervals for inventory based estimates become smaller
the larger scale and sample size used. The calibrated counties with small sample size resulted
in uncertainty bounds1 of SOC stocks of 100 Mg ha-1, while the counties with
larger sample size had uncertainty bounds of 20 Mg ha-1 in 2002. The modeled
estimates also become more certain at larger scales mainly due to more accurate area
estimates. When going from county to regional scale the uncertainties of the SOC
stocks estimated by the Q model decreased with 30%, equivalent to 30 Mg ha-1. |
|
|
|
|
|