Les modèles doivent être simples, mais pas simplistes!
Le texte sera disponible en français prochainement.
The first part of Dr. Papaik’s presentation addressed several general questions concerning model complexity in support of sustainable forest management objectives. He used a saying attributed to Albert Einstein, “Models should be as simple as possible, but no simpler”, and a definition of modeling given by Paul Charboneau at the Max Planck Institute as the foundation for explaining two approaches that SFM modelers can use to meet Einstein’s dictum, and introduced the notion of models as tools for sustainable forest management to underscore the importance and challenges of comparing different models.
The second part of the talk was an application of the above, where Dr. Papaik presented results of a paper that is currently under review involving comparison of two models, SORTIE and LANDIS being used to support SFM objectives. SORTIE, is a spatially-explicit, individual tree model based on species-specific growth, mortality, and dispersal, and has been used primarily for studying stand-level dynamics. The SORTIE model has been parameterized for six tree species in the region of the Lake Duparquet Research and Teaching Forest in western Quebec. ScaledSORTIE is a “scaled-up” application of SORTIE which, by reducing the amount of data processing, facilitates use of the model for studying landscape-level dynamics. In the work presented, ScaledSORTIE was used to create 300 SORTIE stand development simulations for different initial conditions, site types, and harvest regimes on a 11, 000 ha landscape. ScaledSORTIE was compared to LANDIS, a landscape-scale model which has had considerable use in modeling long-term changes of landscapes under managed and natural disturbance regimes. One of the apparent weaknesses of LANDIS is the use of a strict shade-tolerance rule which has a major influence on recruitment, particularly of shade-intolerant species, after partial or gap disturbance. Both models were set to the same initial conditions and histories in order to enable comparisons of simulation results of succession dynamics under different disturbance scenarios: three different harvest regimes (clear cuts, gap cuts, and uniform thinning) and a non-harvest regime.
Simulation results of SORTIE at the stand scale under no-harvest scenario showed variation in relative species abundance among four different land types. The variation persisting for the first 200 years of succession was replaced by the dominance of fir and birch on dry sites and spruce and fir on all the other soil types. In the absence of harvesting, ScaledSORTIE and LANDIS generated similar initial mature tree dynamics, differing only in the timing of the onset of the shade intolerant species decline. Similarly, under the scenario of clear cuts with 80-year rotation period, both ScaledSORTIE and LANDIS simulated very similar patterns of mature tree dynamics in which all the species were present for the whole span of simulation. In the case of gap cuts, mature tree dynamics of ScaledSORTIE did not differ significantly from the non-harvest scenario. In contrast, the mature tree dynamics of the LANDIS gap cut simulation differed greatly from the LANDIS non-harvest regime in the percentage of landscape covered with adult cohorts and in the decline of all species except for balsam fir. Uniform thinning in ScaledSORTIE was similar to the non-harvest regime, whereas, these two regimes differed dramatically in LANDIS.
While the two models forecasted fairly similar mature tree dynamicsas described above , important differences were observed in the understory dynamics under different harvest scenarios. LANDIS understory dynamics were highly simplified as a result of the strict shade tolerance rules. Although validation of models that generate long-term projections is virtually impossible in the short term, comparison of the results of ScaledSORTIE and LANDIS suggest that including more detail on variation in stand dynamics and relaxing the stringency of the shade tolerance rule may improve the capacity of LANDIS to more closely capture these fine-scale dynamics that appear to influence landscape-scale dynamics.
Because of the lack of long-term, broad-scale empirical data, much of what we know about forest ecosystem dynamics is based on fine-scale knowledge acquired from stand- or tree-level studies. Scaling up fine-scale knowledge to the landscape level does not necessarily incorporate dynamics at the broader level and invariably simplifies what is happening in the real world. For example, certain landscape factors affect seed dispersal and influence recruitment at both the local and landscape levels. Disregard for these factors may lead to the loss of diversity in structure and species composition at the landscape level. Predictions of the succession dynamics generated by models that integrate both stand and landscape scales may prove to be a useful tool for planning sustainable forest management that has to accommodate multiple forest services.
Étudiante au Doctorat en Sciences de l’environnement
Université du Québec en Abitibi-Témiscamingue
Courriel : firstname.lastname@example.org