A two-level nested model for extracting positive and negative terrains combining morphology and visualization indicators

Topography affects the resource distribution and limits physical living and production conditions in mountainous regions. Depicting the specific positive and negative terrains in karst region would support for understanding ecosystem process and land resources managing in ecologically fragile areas. However, positive and negative terrains, specific topographic types in karst region, have been difficult to extract by the automatic machine quantitative method. The novelty of this study is developing a two-level nested (TLN) model to quantify and extract positive and negative terrains features of karst through utilizing visualization indicators, openness and the sky-view factor (SVF), combing an insight of morphological analysis. The algorithms of the two visualization indicators detected the surface morphological changes in 3D sphere and realized self-adaptive of complicated topographic changes. There were three main modules included in the two-level nested model: optimum radius analysis of selected indicators, optimum threshold judgement of positive and negative terrains edges and a morphological postprocessing module. The two-level chained process designed in the second module reached the quantitative potential critical thresholds detection and optimizing expression of positive and negative terrains boundaries, which integrated analyzing the variation trend of openness difference, SVF value and SVF change. Through the TLN framework, the accuracy and stability of positive and negative terrains boundaries extraction by openness and SVF were further strengthened. The TLN model was tested in seven counties where the karst terrain morphologies were quite different. The boundaries extracted by our TLN model were compared with the visual interpretation results, and the accuracy assessment validated that the overlapping areas of test and reference results were more than 90% and the boundaries' distance deviations were less than one grid. Positive terrain was predominant in all study areas and the proportions varied from 75.33 to 85.23%. Different positive and negative terrains percentages and formations indicated different phases in landform and ecosystem evolution process. The largest proportion of cultivated land appeared in the 0-150m buffer ring where mutations were detected, which implied an increased ecological risk of karst region. Therefore, it is urgent to resolve the increasingly serious ecological and environmental issues while ensure agricultural production and economic development. The positive and negative terrains distribution patterns could provide theoretical reference for a rational land use planning in the future.