Arief Wijaya
Arief Wijaya received his doctoral degree in the field of Remote Sensing from TU-Bergakademie Freiberg, Germany in 2010. He was a recipient of Georg Forster Research Fellowship from Alexander von Humboldt Foundation in 2011 undertaking Post-Doctoral research at TU-Dresden, Germany to model carbon biomass over tropical peat swamp forest of Indonesia.
Prior to his tenure at Thünen Institute, he worked at Center for International Forestry Research (CIFOR) based in Indonesia, and is responsible to integrate socio-economic and spatial data for analyzing drivers of deforestation and forest degradation in Indonesia, Vietnam, Peru, Brazil, Cameroon and Zambia. He was also an observer of UN Climate negotiations, especially for MRV REDD+ and climate change mitigation issues. In 2015, he is part of technical team of Ministry of Environment and Forestry in Indonesia responsible for preparing National Forest Reference Emissions Level (FREL) document for submission to UNFCCC.
He is currently working within Landscape Forestry in the Tropics (LaForeT) project established at Thünen Institute of International Forestry and Forest Economics in Hamburg, Germany and focuses on the analysis of drivers of deforestation/reforestation and spatial modeling of tropical landscape dynamics in the Philippines and Ecuador.
Supervisors: Richard Gloaguen, Hermann Heilmeier, Uta Berger, and Lou Verchot
Address: Thünen-Institut für Internationale Waldwirtschaft und Forstökonomie
(Thünen Institute of International Forestry and Forest Economics)
Leuschnerstraße 91
21031 Hamburg (Germany)
Prior to his tenure at Thünen Institute, he worked at Center for International Forestry Research (CIFOR) based in Indonesia, and is responsible to integrate socio-economic and spatial data for analyzing drivers of deforestation and forest degradation in Indonesia, Vietnam, Peru, Brazil, Cameroon and Zambia. He was also an observer of UN Climate negotiations, especially for MRV REDD+ and climate change mitigation issues. In 2015, he is part of technical team of Ministry of Environment and Forestry in Indonesia responsible for preparing National Forest Reference Emissions Level (FREL) document for submission to UNFCCC.
He is currently working within Landscape Forestry in the Tropics (LaForeT) project established at Thünen Institute of International Forestry and Forest Economics in Hamburg, Germany and focuses on the analysis of drivers of deforestation/reforestation and spatial modeling of tropical landscape dynamics in the Philippines and Ecuador.
Supervisors: Richard Gloaguen, Hermann Heilmeier, Uta Berger, and Lou Verchot
Address: Thünen-Institut für Internationale Waldwirtschaft und Forstökonomie
(Thünen Institute of International Forestry and Forest Economics)
Leuschnerstraße 91
21031 Hamburg (Germany)
less
InterestsView All (26)
Uploads
Papers by Arief Wijaya
and Grey Level Co-occurrence Matrix (GLCM) texture measures. Different texture layers derived from Geostatistics method, namely fractal dimension, semivariogram, madogram, rodogram, pseudo-cross variogram and pseudo-cross madogram, were incorporated for the land cover classification of tropical rainforests in East Kalimantan, Indonesia. Texture layers of grey level 2 co-occurrence matrix (GLCM) channels, i.e. variance, contrast, dissimilarity, and homogeneity, were also used for the classification. Two classification methods, using Support Vector Machine and Minimum distance were applied for image classification. Landsat 7 ETM images combined with textural information is used for land cover classification of tropical rainforest area. Band 5 of Landsat data was used to compute texture layers using the GLCM and Geostatistics methods. This band was chosen because it has the highest variance of training data compared to other spectral bands. The results were compared to find out how the extra information given by the texture enhances the classification. According to the accuracy assessment using error matrix, combinations of image and texture data performed better with 81% of accuracy compared to those of image data only with 76% of accuracy.
Through the experiences and analyses of five REDD+ countries, two donor organisations and several researchers and negotiators, the papers gathered examine:
- success factors for building capacity and implementing national forest monitoring
- stepwise approaches for bridging capacity gaps through continuous improvement
- key components and attributes of an effective national forest monitoring system
- data and technology needed for forest monitoring
- the conservativeness principle, benefit distribution, and a framework for REDD+ reference levels (stepwise approach)
- assessment of current and required methodological guidance.
By publishing these papers for a wider audience, this collection aims to help all those invested in the success of REDD+ to learn from others’ experiences.
This joint publication of the CIFOR Global Comparative Study on REDD and the GOFC-GOLD Land Cover Office synthesises the main outcomes of that meeting.
This study explores the possibility of Landsat 7 ETM data acquired in 2003 to detect single tree felling in three stages. In the first stage, seven bands of Landsat image were classified using fuzzy c-means, neural network and maximum likelihood methods. Pixels labelled as single tree felling class by these classifiers were masked out and assigned as ‘first order single tree felling class’. The remaining pixels were defined as ‘unclear single tree felling class’ and classified in the second stage applying neural network method and some ancillary data. The single tree felling class pixels from the second classification was defined as ‘second order single tree felling class’ and the accuracy was assessed using confusion matrices. The first- and second order single tree felling pixels were merged and filtered using expert knowledge-based GIS layers resulting in more reliable illegal logging map.
The results of this study found the fuzzy c-means method has less accuracy than neural network and maximum likelihood techniques for classifying the single tree felling class. In the first classification, the maximum likelihod, fuzzy c-means and neural network methods show 53.3% agreement in discriminating pixels as single tree felling class. Multi-source classification of neural network in the second stage resulted in 80% of accuracy. The best result was found combining multi-spectral bands of Landsat ETM data, aspect, elevation, skewness and variance in the input layer of neural network classifcation. The rule-based classification in the last stage found 8.6 – 15.2 % of total RKL 1 area identified as illegal logging sites.
Factor Analysis combined with Principle Component Analysis (FA-PCA), Optimum Index Factor (OIF), and 6 Bands ALI image are used as the input of this research. Combination of fuzzy classification and cluster shape is applied to training area that already collected. We use Fuzzy C-Means, Fuzzy Shape, and Fuzzy Adjusted for image classification considering sphere and ellipsoid as cluster shape forms. Conventional maximum likelihood classification and Neural Networks method are used as comparison of the fuzzy classification result.
The classification results show that fuzzy approach can be applied as alternative to solve mixed pixel problems in medium spatial resolution images. By adjusting the cluster size in classification could increase accuracy of the result. The fuzzy adjusted with ellipsoid cluster shape and PCA input was the best classification for mixed training areas with the accuracy of up to 82.11%.
This study investigates the habitat suitability for Sumatran tiger (Panthera tigris sumatrae) in Tesso Nillo National Park and surrounding landscape in Sumatra, Indonesia. We derived land use and distance from road networks based on supervised maximum likelihood classification and the interpretation of multispectral Landsat ETM bands with 30m resolution. The stream networks are generated from smoothed SRTM DEM calculating flow direction and flow accumulation of the catchment over the study area. Slope, elevation and aspect are also calculated from the SRTM data. These features are overlaid each other and the value of each grid cell is constructed in a database.
Field work is conducted to collect ground truth data and to study the behavior of the Sumatran tiger and its prey with respect to certain environmental variable. Ultimately, a habitat suitability map for the Sumatran tiger is generated using binomial multiple logistic regression technique integrated with GIS environment during post field work. Based on the suitability map, the biophysical properties and anthropogenic factors affecting the distribution of Sumatran tiger and its prey and the possibility for defining corridor development for the conservation of this species are discussed.
Keywords: Sumatran tiger, land use change, habitat suitability map
This study concerned on a tropical rainforest in East Kalimantan, Indonesia. The satellite remote sensing data was atmospherically corrected using Dark Object Subtraction (DOS) technique, and topographic correction was conducted using C-correction method. Stand volume was estimated using field data and remote sensing data, thus using Levenberg-Marquardt algorithm neural network method was employed. Due to unavailability of actual biomass data, the stand volume estimate was converted into the above ground biomass using equations developed for the tropical Amazonian forest assuming similar forest vegetation over both areas. Spatial distribution of the AGB and the error estimate were then interpolated using kriging.
The results showed integration of field measurement and remote sensing data has better prediction of stand volume validated with the actual stand volume data. The AGB estimate showed a great correlation with predicted stand volume data, number of stems, and basal area. Spatial distribution of the AGB described a correlation between forest biomass and land use/land cover in the study area.
Keywords: above ground biomass, stand volume, remote sensing, neural network, kriging
More information about the conference:
http://www.wzw.tum.de/lifo2008//index.php
Keywords : stand volume estimate, neural network, Levenberg-Marquardt, Kriging, multi-linear regression