Since complexity is a trait linked to the context investigated (e.g. low complexity might mean different things in modeling global vegetation distribution and modeling chloroplast activity at the leaf level) a relative definition of complexity was chosen for this analysis. I selected a set of ten different models from the general domain of application in focus here. These reference models include detailed individual-based competition models and gap models (SORTIE, Pacala et al. 1996; KiWi, Berger and Hildebrandt 2000; ForClim v2.9.3, Bugmann and Solomon 2000, Risch et al. 2005), process-based models of different level of aggregation (3-PG, Landsberg and Waring 1997; BiomeBGC, Running and Coughlan 1988, Thornton et al. 2002) as well as models simulating forest dynamics at larger scales (LANDIS, He and Mladenoff 1999, Mladenoff 2004; ED, Moorcroft et al. 2001). I also included models integrating different approaches in modular, hierarchical and hybrid model approaches (FireBGC, Keane et al. 1996, 1999; LANDIS-II, Scheller et al. 2007, Scheller and Mladenoff 2004; PICUS v1.4, Seidl et al. 2005, Lexer and Hoenninger 2001). This selection of reference models intends to span a variety of widely applied modeling approaches. It has to be noted, however, that it is not destined to be an exhaustive review of forest ecosystem models. Furthermore, where different model versions and modifications exist one indicative representation was selected (see references above), with the exception of LANDIS. LANDIS is a highly flexible modeling platform and two different realizations of the model, spanning a wide complexity gradient, were included in the set of references models: The original LANDIS version (cf. He and Mladenoff 1999, Mladenoff 2004), representing a widely applied example of a classical landscape model and a LANDIS-II rendering employing extensions using PnET physiology and Century soil processes (R.M. Scheller, personal communication). It has to be acknowledged that the relative complexity scoring is essentially an expert assessment, i.e. entails a certain degree of fuzziness and is no accurate quantitative metric. Thus to facilitate transparency in this assessment a descriptive account of the complexity assessment for the selected models is given below with regard to aspects of structural complexity (Tables 2-6), for functional complexity (Tables 7-11), and spatial complexity (Tables 12-14) in addition to the relative scoring.
The reference models - Methods and material - iLand model complexity and its niche in the landscape of models