Satellite-based prediction of rainfall interception by tropical forest stands of a human-dominated landscape in Central Sulawesi, Indonesia

• Published in: Journal of hydrology (Amsterdam) Vol. 364; no. 3; pp. 227 - 235
• Main Authors: Nieschulze, Jens, Erasmi, Stefan, Dietz, Johannes, Hölscher, Dirk
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier B.V 30-01-2009; [Amsterdam; New York]: Elsevier; Elsevier
• Subjects: agroforestry; Animal and plant ecology; Animal, plant and microbial ecology; Applied geophysics; Biological and medical sciences; Canopy; Earth sciences; Earth, ocean, space; Exact sciences and technology; Fundamental and applied biological sciences. Psychology; Hydrology; Hydrology. Hydrogeology; Internal geophysics; land use change; Local maxima; Logging; overstory; prediction; Quickbird; rain; rain interception; remote sensing; satellites; selective felling; Synecology; Terrestrial ecosystems; Texture; Theobroma cacao; tropical forests; Indonesia; Celebes; Far East; Asia; Indonesia, Sulawesi; Logging; Texture; Quickbird; Canopy; Local maxima; textures; regression analysis; variograms; vegetation; rain water; image analysis; land use; imagery; high resolution; landscapes; satellites; models; rainfall; reflectance; Space remote sensing; forests; hydrologic cycle; prediction; vegetation effects; standard deviation
• ISSN: 0022-1694; 1879-2707
• DOI: 10.1016/j.jhydrol.2008.10.024

Rainforest conversion to other land use types drastically alters the hydrological cycle in which changes in rainfall interception contribute significantly to the observed differences. However, little is known about the effects of more gradual changes in forest structure and at regional scales. We studied land use types ranging from natural forest over selectively-logged forest to cacao agroforest in a lower montane region in Central Sulawesi, Indonesia, and tested the suitability of high-resolution optical satellite imagery for modeling observed interception patterns.
Investigated characteristics indicating canopy structure were mean and standard deviation of reflectance values, local maxima, and self-similarity measures based on the grey level co-occurrence matrix and geostatistical variogram analysis. Previously studied and published rainfall interception data comprised twelve plots and median values per land use type ranged from 30% in natural forest to 18% in cacao agroforests. A linear regression model with local maxima, mean contrast and normalized digital vegetation index (NDVI) as regressors was able to explain more than 84% (
R
adj
2
) of the variation encountered in the data. Other investigated characteristics did not prove significant in the regression analysis. The model yielded stable results with respect to cross-validation and also produced realistic values and spatial patterns when applied at the landscape level (783.6
ha). High values of interception were rare and localized in natural forest stands distant to villages, whereas low interception characterized the intensively used sites close to settlements. We conclude that forest use intensity significantly reduced rainfall interception and satellite image analysis can successfully be applied for its regional prediction, and most forest in the study region has already been subject to human-induced structural changes.