Available online at www.sciencedirect.com ScienceDirect Agriculture and Agricultural Science Procedia 11 (2016) 90 – 94 International Conference on Inventions & Innovations for Sustainable Agriculture 2016, ICIISA 2016 Identification and Seasonal Analysis of Degraded Tropical Peatland by Using ALOS AVNIR-2 Data Dandy Aditya Novresiandi a, b,*, Ryota Nagasawa c a United Graduate School of Agricultural Sciences, Tottori University, 4-101 Koyama-cho Minami, Tottori 680-8550, Japan b Center for Remote Sensing, Bandung Institute of Technology, Jalan Ganesha 10, Bandung 40132, Indonesia c Faculty of Agriculture, Tottori University, 4-101 Koyama-cho Minami, Tottori 680-8550, Japan Abstract Tropical peatlands are being subjected to the consequences of rapid economic development without any consideration of the importance of sustainable management practices Sustainable management of tropical peatlands is an important element in controlling carbon emission However, the available information of tropical peatlands lacks of accuracy and is outdated, especially in terms of medium to high resolution Thus, development of reliable monitoring techniques is a significant step towards the sustainable management of tropical peatlands The remote sensing (RS) application is suitable as a tool to monitor tropical peatlands, whereas direct measurements are generally labor-intensive, time-consuming and limited to accessibility In this study, methodology to identify degraded tropical peatland was developed by using the McFeeters Normalized Difference Water Index (McFeeters-NDWI), which was derived by Advanced Land Observing Satellite (ALOS) Advanced Visible and Near Infrared Radiometer type (AVNIR-2) data Additionally, a seasonal analysis was carried out to examine the characteristics of degraded tropical peatland during the rainy and dry seasons from the viewpoint of the medium to high resolution of optical RS Overall, a relationship was discovered such that the wet shrub class was considered as the degraded tropical peatland area, and was identified as being in between -0.43 to -0.11 of the McFeeters-NDWI value The wet-shrub class yielded a producer’s accuracy of 80.6% and a user’s accuracy of 91.8% Afterwards, the seasonal change was discovered to slightly shift the threshold values (TrVs) in the identification of degraded tropical peatland by as much as -0.05 However, the interval of the TrVs for the wet shrub class was stable and remained unchanged © 2016 The Authors Published by Elsevier B.V This is an open access article under the CC BY-NC-ND license © 2016 The Authors Published by Elsevier B.V (http://creativecommons.org/licenses/by-nc-nd/4.0/) Peer-reviewunder under responsibility responsibility of of the the Faculty Faculty of Animal Sciences Sciences and Peer-review of Animal and Agricultural Agricultural Technology, Technology, Silpakorn Silpakorn University University Keywords: tropical peatlands; McFeeters-NDWI; ALOS AVNIR-2; remote sensing * Corresponding author Tel.: +81-9057871656 E-mail address: dandyaditya@gmail.com 2210-7843 © 2016 The Authors Published by Elsevier B.V This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Peer-review under responsibility of the Faculty of Animal Sciences and Agricultural Technology, Silpakorn University doi:10.1016/j.aaspro.2016.12.015 Dandy Aditya Novresiandi and Ryota Nagasawa / Agriculture and Agricultural Science Procedia 11 (2016) 90 – 94 Introduction Tropical peatlands play an important role in the global carbon balance, while being recognized as one of the largest terrestrial carbon stores (Jauhiainen et al., 2005) Therefore, tropical peatlands have a direct relationship with the process of global climate change (Jaenicke et al., 2008) Unfortunately, tropical peatlands are being subjected to the consequences of rapid economic development without any consideration of the importance of sustainable management practices (Rieley et al., 2008) Excessive land conversions to commercial plantations, drainage and illegal logging have led to fires, as well as large increases in carbon emissions to the atmosphere (Rydin and Jeglum, 2006) Sustainable management of tropical peatlands is an important element in controlling carbon emission However, the available information on tropical peatlands lacks accuracy and is outdated (Page et al., 2007), especially in terms of medium to high resolution Thus, the development of reliable monitoring techniques is a significant step towards the sustainable management of tropical peatlands The remote sensing (RS) application is suitable as a tool to monitor tropical peatlands, whereas direct measurements are generally labor-intensive, timeconsuming and limited to accessibility The use of the RS application to the monitoring of tropical peatlands has been increasing expeditiously in recent years, along with the availability of RS data sets (e.g., Page et al., 2002; Wijaya et al., 2010; Wahyunto et al., 2012) The data from the Advanced Land Observing Satellite (ALOS) Advanced Visible and Near Infrared Radiometer type (AVNIR-2) are a particular concern, as they provide a medium to high spatial resolution of 10 meters for optical RS data (JAXA, 2008) The present study was carried out to develop methodology for identifying degraded tropical peatland using ALOS AVNIR-2 data The McFeeters Normalized Difference Water Index (McFeeters-NDWI) was evaluated to measure the amount of wetness in shrub areas, as well as examine the characteristics of degraded tropical peatland from the perspective of the medium to high resolution of optical RS In addition, a seasonal analysis was carried out to examine the characteristics of degraded tropical peatland during the rainy and dry seasons from the perspective of the medium to high resolution of optical RS Materials and Methods 2.1 Description of the study area and dataset The study area was taken from the catchment area of the Kahayan River in Central Kalimantan, Indonesia (Fig 1) In general, the condition of tropical peatland in this area is mostly in a severely degraded condition (Jaenicke, 2010) A sparse to medium vegetation layer in the form of shrubs covers the degraded tropical peatland in this area This condition was considered as a key parameter in order to identify the degraded tropical peatland In this study, ALOS AVNIR-2 data, acquired on 11 January 2009 (rainy season) and 17 October 2010 (dry season), were used as the primary data An existing land use/cover map, published by the Indonesian Geospatial Information Agency, which was updated by using visual interpretation, was used as a reference map In addition, data collected from a ground truth survey conducted between 23 and 28 August 2013 was used to provide basic information about the study area Fig Map of Indonesia showing the location of the study area (hatched rectangle) 91 92 Dandy Aditya Novresiandi and Ryota Nagasawa / Agriculture and Agricultural Science Procedia 11 (2016) 90 – 94 2.2 Remote sensing techniques and methodology Geometric and radiometric corrections were made to the data processing for both sets of data (Oo and Takagi, 2010) Thus, eight classes of land use/cover―waterbody light, waterbody mix, waterbody dark, artificial, forest, bare land, shrub and mixed plantation―were extracted by performing a supervised classification using the Maximum Likelihood method From these classes, the shrub class was used as an approach to identify degraded tropical peatland Rainy season data were used to develop the methodology for the identification of degraded tropical peatland due to less cloud contamination compared to dry season data Therefore, McFeeters-NDWI was derived using the combination of green and near-infrared bands of the ALOS AVNIR-2 data The McFeeters-NDWI has been proposed as a parameter for discriminating and improving the existence of open water features among soil and other terrestrial vegetation features from the perspective of optical RS data The McFeeters-NDWI was derived by the following equation (McFeeters, 1996): McFeeters NDWI GREEN NIR GREEN NIR (1) Where, GREEN and NIR are the green and the near-infrared bands of optical RS data in the reflectance unit, respectively Generally, the open water features yield positive values, while soil and other terrestrial vegetation features generate zero or negative values Subsequently, threshold values (TrVs) were derived by means of spatial analysis for discriminating shrub class into wet shrub and dry shrub classes The McFeeters-NDWI image was then overlaid with the reference map for spatial analysis Thus, regions that exhibited similar patterns of degraded tropical peatland areas were combined as a “wet shrub”, while its McFeeters-NDWI values were considered as the TrVs for the wet shrub class Accuracy assessment, using a confusion matrix, was carried out in order to assess the accuracy of the TrVs Hence, a total of 388 points was generated on the shrub areas of the reference map, with each point situated within a x km mesh In additional, validation using the spectral signature of ALOS AVNIR-2 was carried out in order to examine associations between the derived wet shrub class and the degraded tropical peatland area A particular value of spectral signature for bands 2, 3, and was found to be represented by the thickness of peatland (Wahyunto et al., 2012) Furthermore, in order to conduct seasonal analysis, dry season data were processed by the same procedures as the rainy season data Thus, McFeeters-NDWI images derived by both sets of data were then overlaid for seasonal analysis purposes The difference between the TrVs, in relation to rainy season and dry season data, was examined, as well as their intervals, in order to understand the effect of seasonal change upon the identification Results and Discussions As shown in Fig (a), the McFeeters-NDWI image of the shrub class area was generated by the rainy season data This image was then overlaid with the reference map for spatial analysis Thus, a relationship was discovered between the McFeeters-NDWI and the shrub area, such that the McFeeters-NDWI was found to produce a higher value for the wet shrub area than for the dry shrub area The higher wetness level of the wet shrub compared with the dry shrub was clearly distinguished by the McFeeters-NDWI Hence, TrVs were derived in order to separate the shrub class into wet and dry classes The wet shrub was identified as being in between -0.43 and -0.11 of the McFeeters-NDWI value Subsequently, accuracy assessment using a confusion matrix was carried out to assess the accuracy of the TrVs, which yielded a Producer’s Accuracy (PA) of 80.6% and a User’s Accuracy (UA) of 91.8% for the wet shrub class, whereas the dry shrub class achieved a PA of 87.4% and a UA of 86.6% Overall, the TrVs yielded an Overall Accuracy (OA) of 84.2% and a Kappa Coefficient of 0.69 Concurrently, additional validation using a spectral signature of ALOS AVNIR-2 was performed to examine the association between the derived wet shrub class and the tropical peatland area The spectral signatures for bands 2, and of ALOS AVNIR-2 were found to represent the thickness of peatland, whereby a lower spectral signature meant a thicker tropical peatland (Wahyunto et al., 2012) A shallow (less than 100 cm of thickness) tropical peatland ought to have spectral signatures of less than, or equal to, 51.2% of band 2, 43.6% of band 3, and 77.6% of Dandy Aditya Novresiandi and Ryota Nagasawa / Agriculture and Agricultural Science Procedia 11 (2016) 90 – 94 Fig McFeeters-NDWI images of (a) rainy season and (b) dry season data band These values were used as a boundary to determine the association between the wet shrub class and the presence of tropical peatland Thus, a spectral signature less than, or equal to, the boundary value ought to be recognized as tropical peatland and vice versa As a result, a relationship was discovered between the wet shrub class and the tropical peatland area The spectral signature for the derived wet shrub class was situated below the boundary of the spectral signature for peatland (Table 1) Therefore, the wet shrub class was considered as the degraded tropical peatland area Table The spectral signature comparison between shallow tropical peatland and the wet shrub class Class Name Shallow peat (less than 100 cm) wet shrub Spectral signature Band % Digital Number 51.2 130.6 42.35 108 Band % 43.6 31.76 Digital Number 111.2 81 Band % 77.6 42.35 Digital Number 197.9 108 Furthermore, to conduct a seasonal analysis, dry season data were processed by the same procedures as the rainy season data As shown in Fig (b), the McFeeters-NDWI image of the shrub class area was generated by the dry season data This image was then overlaid with the McFeeters-NDWI image of the rainy season data for seasonal analysis Thus, the seasonal analysis found that there was a shift of -0.05 in the TrVs between both sets of data 93 94 Dandy Aditya Novresiandi and Ryota Nagasawa / Agriculture and Agricultural Science Procedia 11 (2016) 90 – 94 (Table 2) Therefore, the seasonal change was discovered to moderately shift the TrVs in the identification of degraded tropical peatland Nevertheless, the interval of the TrVs for the wet shrub class was stable and remained unchanged by as much as 0.32 The rainy season data TrVs were found to be closer to positive values, which indicate that the degraded tropical peatland was slightly wetter during the rainy season and vice versa However, this condition did not extremely affect the TrVs, since they only moderately shifted, while the interval of the TrVs for the wet shrub class was stable and remained unchanged Consequently, in the present study, the McFeeters-NDWI provided better analysis for identifying degraded tropical peatland without being severely influenced by the seasonal effects Table TrVs comparison between rainy season and wet season data Data rainy season dry season Threshold Values (TrVs) dry shrub class -0.44 䍸 dry shrub -0.49 䍸 dry shrub wet shrub class -0.43 䍸 wet shrub 䍸 -0.11 -0.48 䍸 wet shrub 䍸 -0.16 open water class open water 䍹 -0.10 open water 䍹 -0.15 Conclusions The study shows that optical RS application could be advantageous as an approach to identify degraded tropical peatland in relation to the sustainable management of tropical peatland In this study, the methodology to identify degraded tropical peatland was developed by using the McFeeters-NDWI, which was derived by ALOS AVNIR-2 data The degraded tropical peatland was identified to be associated with the wet shrub class generated by the developed methodology Additionally, the developed methodology was useful in understanding the seasonal effects of degraded tropical peatland, whereby the McFeeters-NDWI provided better analysis for identifying degraded tropical peatland without being severely influenced by the seasonal effects Finally, this study should aid to improve the state of knowledge about tropical peatland monitoring, especially involving the use of medium to high resolution optical RS In terms of future research, findings obtained from this study ought to be used as considerations for further analysis and developing methodology, along with the use of Synthetic Aperture Radar (SAR) data to improve the reliability of tropical peatland identification by using an RS application References Jaenicke, J., Rieley, J O., Mott, C., Kimman, P., and Siegert, F., 2008, Determination of the amount of carbon stored in Indonesian peatlands Geoderma 147, 151-158 Jaenicke, J., 2010 3D modelling and monitoring of Indonesian peatlands aiming at global climate change mitigation Ludwig Maximilians University of Munich, Germany Jauhiainen, J., Hidenori, T., Juha, E P H., Pertti, J M., and Harri, V., 2005 Carbon fluxes from a tropical peat swamp forest floor Global Change Biology 11, 1788-1797 JAXA, 2008 ALOS data users handbook: revision C Japan Aerospace Exploration Agency McFeeters, S K., 1996 The use of the Normalized Difference Water Index (NDWI) in the delineation of open water features International Journal of Remote Sensing 17 (7), 1425-1432 Oo, K S., and Takagi, M., 2010 Land cover classification for forest management using ALOS AVNIR-2 images International Symposium on Social Management Systems 2010 Kochi, Japan, paper #SMS10-170 Page, S E., Siegert, F., Rieley, J O., Boehm, H V., Jaya, A., and Limin, S., 2002 The amount of carbon released from peat and forest fires in Indonesia during 1997 Nature 420, 61-65 Page, S E., Banks, C J., and Rieley, J O., 2007 Tropical peatlands: distribution, extent and carbon storage – uncertainties and knowledge gaps Peatlands International 2007 (2), 26-27 Rieley, J O., Wüst, R A J., Jauhiainen J., Page, S E., Wösten, H., Hooijer, A., Siegert, F., Limin, S H., Vasander, H., and Stahlhut, M., 2008 Tropical peatlands: carbon stores, carbon gas emission and contribution to climate change processes, in “Peatlands and Climate Changes” In: Strack, M (Ed.) International Peat Society, pp 148-181 Rydin, H., and Jeglum, J., 2006 The biology of peatlands Oxford University Press, United Kingdom Wahyunto, Supriatna, W., and Agus., 2012 ALOS satellite data to explore areal extent of peatland case study: Kubu Raya District, West Kalimantan Province Report and proceedings of ALOS application and verification project in Indonesia, 26-31 Wijaya, A., Marpu, P R., and Gloaguen, R., 2010 Discrimination of peatlands in tropical swamp forests using dual-polarimetric SAR and Landsat ETM data International Journal of Image and Data Fusion 1(3), 257-270 ... signatures for bands 2, and of ALOS AVNIR- 2 were found to represent the thickness of peatland, whereby a lower spectral signature meant a thicker tropical peatland (Wahyunto et al., 20 12) A shallow... (less than 100 cm of thickness) tropical peatland ought to have spectral signatures of less than, or equal to, 51 .2% of band 2, 43.6% of band 3, and 77.6% of Dandy Aditya Novresiandi and Ryota Nagasawa... identify degraded tropical peatland in relation to the sustainable management of tropical peatland In this study, the methodology to identify degraded tropical peatland was developed by using the