Environmental changes in biodiversity hotspot ecosystems of
South Ecuador: RESPonse and feedback effECTs (FOR2730)

Abstract:

Landslides are a hazard for humans and artificial structures. From an ecological point of view, they represent an important ecosystem disturbance, especially in tropical montane forests. Here, shallow translational landslides are a frequent natural phenomenon and one local determinant of high levels of biodiversity. In this paper, we apply weighted ensembles of advanced phenomenological models from statistics and machine learning to analyze the driving factors of natural landslides in a tropical montane forest in South Ecuador. We exclusively interpret terrain attributes, derived from a digital elevation model, as proxies to several driving factors of landslides and use them as predictors in our models which are trained on a set of five historical landslide inventories. We check the model generality by transferring them in time and use three common performance criteria (i.e. AUC, explained deviance and slope of model calibration curve) to, on the one hand, compare several state-of-the-art model approaches and on the other hand, to create weighted model ensembles. Our results suggest that it is important to consider more than one single performance criterion. Approaching our main question, we compare responses of weighted model ensembles that were trained on distinct functional units of landslides (i.e. initiation, transport and deposition zones). This way, we are able to show that it is quite possible to deduce driving factors of landslides, if the consistency between the training data and the processes is maintained. Opening the ?black box? of statistical models by interpreting univariate model response curves and relative importance of single predictors regarding their plausibility, we provide a means to verify this consistency. With the exception of classification tree analysis, all techniques performed comparably well in our case study while being outperformed by weighted model ensembles. Univariate response curves of models trained on distinct functional units of landslides exposed different shapes following our expectations. Our results indicate the occurrence of landslides to be mainly controlled by factors related to the general position along a slope (i.e. ridge, open slope or valley) while landslide initiation seems to be favored by small scale convexities on otherwise plain open slopes.

Abstract:

Soil-vegetation-atmosphere transfer (SVAT) is to be predicted for 2050/2100 for a study area in the southern Ecuadorian Andes. SVAT models require information on land use/ land cover (LUC) as lower boundary conditions. Since the study area suffers from high deforestation rates, LUC cannot be assumed as staying constant with time. A spatially explicit land use/land cover change (LUCC) model is therefore needed for future SVAT prediction. The numbers of approaches of LUCC modelling are numerous. Difficulties are due to complex interactions of social and biophysical drivers of change. In this study a model of LUCC was built using information of past changes derived by remotely sensed data. Special focus was on forest development patterns. A training period of 14 years between 1987 and 2001 was chosen. Two LUC classifications were accomplished to Landsat data of the start and end date of this period. A change detection of the training period provided the basis for predictive LUCC modelling. Potential drivers for LUCC were applied to the model as GIS layers. The modelling procedure consisted of a combination of Markov chain analysis (MCA) for quantitative modelling and multi-layer perceptron (MLP) for revealing potential locations of change. A multi-objective land allocation (MOLA) served as final integration step. 14 LUC transitions were considered in the modelling procedure. Unconsidered LUC classes were assumed to stay constant in the future. The model results were maps of LUC for 2006, 2010 and afterwards for every 10 years up to 2100. An internal validation was performed with the training data. The results of the prediction were validated by comparing the model output of 2006 to an ASTER LUC classification of the same time. The validation methodology comprised crisp and fuzzy map comparison using Kappa statistics. The study area featured a deforestation of 13.61% in the training period. The model was able to explain deforestation in the training period 51% better than just by chance. The location of predicted deforestation reached a better than chance agreement of 30%. Predicted quantities of deforestation were 59% conforming with the reference. The validation of the prediction indicated the difficulty of modelling human impact on the ecosystem. Prospects and limitations of the model were identified with suggestions for future research tasks. The results of this study are assumed to present a good groundwork for future SVAT models.