Remote sensing for sustainable forest management

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/235334384 Remote Sensing for Sustainable Forest Management - ‘A case study of Monitoring Pai Forest Irrigated... Conference Paper · November 2009 CITATIONS READS 0 55 3 authors, including: Naeem Shahzad Institute of Space Technology 14 PUBLICATIONS 5 CITATIONS SEE PROFILE All content following this page was uploaded by Naeem Shahzad on 19 December 2016. The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document and are linked to publications on ResearchGate, letting you access and read them immediately. REMOTE SENSING FOR SUSTAINABLE FOREST MANAGEMENT – ‘A CASE STUDY OF MONITORING PAI FOREST IRRIGATED PLANTATION BY USING LCCS AND OBJECT BASED CLASSIFICATION TECHNIQUE’ Urooj Saeed1, Naeem Shahzad1, Rab Nawaz1 and Hammad Gilani2 1: World Wide Fund for Nature (WWF) Pakistan Ferozepur Road Lahore, Pakistan 2: International Centre for Integrated Mountain Development (ICIMOD), Nepal Abstract: Current study aims to analyze the temporal variations in the Pai Forest irrigated plantation by using Object based Image Classification Technique through generating Segments on the Pan sharpened (0.6m) Quick Bird data. A standardized legend was developed by using Land Cover Classification System (LCCS) designed by Food and Agriculture Organization (FAO) and United Nations Environment Programme (UNEP). The study reveals a significant decrease of 40.57% in the ‘mixed forest class’ i.e. Acacia spp., Prosopis cinerea, Salvadora spp. and Prosspis juliflora. from the years 2003 to 2008. This decline could be attributed to the extraction of forest for fuel wood. On the other hand an increase of 55 hectares in Eucalyptus spp. class has been observed which seems to be due to the plantation activities in the area. The spread of exotic vegetation i.e. Prosopis juliflora is 128% and it is critical to eradicate some of the Forest compartments. Prosopis spp., an exotic species of vegetation has increased by 128%. It is recommended to study and halt any impact of Prosopis spp. on indigenous flora and fauna. It will help the foresters to develop forest products in a sustainable manner that can diversify and support livelihoods of local communities. The study can be effectively used as a baseline to strengthen the National level vegetation mapping projects. The developed thematic maps can be effectively merged with other land cover maps at regional and global scale. Key Words: LCCS, Object based Image Classification, Segmentation, Quick Bird. Introduction Pai Forest is situated in Nawabshah District of the Sindh Province. Geographically it ranges from 68.2134ºE, 26.1012ºN to 68.2842ºE, 26.1263ºN with an area of 2,348 hectares (5,802 acre). Map of study area is shown in Figure 1. The area has been divided into different forest compartments for the purpose of plantation. Each forest compartment comprises of 16 hectares (40 acre) which is being irrigated by fourteen tube wells [1]. During early twentieth century, the forest lost its connection with the riverine track due to the construction of protection embankments along the Indus River. It is now dependent on the sanctioned irrigation water supply which is inadequate and infrequent to sustain the entire area. This situation is leading to a continuous degradation of forest and wildlife habitat [1]. Figure 1: Location map of the study area Presented at a workshop on SAR and optical applications for natural resources and environment management, I SNET , November 2-5, 2009, GORS headquarters, Damascus, Syria Purpose of the Study • • To develop thematic maps of Pai Forest using temporal satellite images To monitor forest cover through change analysis using high-resolution satellite images Methodology Satellite data and software used High resolution QuickBird satellite images of two different dates (2003 and 2008) were procured from Digital Globe. The acquired images were in the Universal Transverse Mercator (UTM) coordinate system with Spheroid and Datum as WGS 84. The characteristic details of the satellite images are given in table 1. Table 1: Characteristics of satellite data Satellite Acquisition Date 04-02-2003 08-07-2008 QuickBird QuickBird Spatial Resolution (m) XS (2.4), Pan (0.6) XS (2.4), Pan (0.6) Spectral Bands 4, 1 4, 1 For satellite images interpretation and processing Digital Image Processing (DIP) software ERDAS Imagine 8.7®, Definien Developer 7.0® ,Land Cover Classification System (LCCS) 2.4.5® were used. All the maps were developed in ArcGIS 9.0®. Microsoft Word and Microsoft Excel were used for documentation and graphical analysis. A field visit of Pai Forest was arranged to collect ground truth data. One hundred and twelve Ground Control Points (GCPs) were collected as sample areas for the land cover maps development. LCCS Legend Definition Classification in a simplest way can be defined as an abstract representation of the ground situation. On the other hand, legend can be defined as the translator of that abstract. It should be therefore scale and source independent which can lead integration of Land Cover of diverse areas in a same database [Antonio and Louisa 1998]. For land cover legend standardization, Food and Agriculture Organization (FAO) and United Nations Environment Programme (UNEP) introduced a Land Cover Classification System (LCCS) for legend definition. LCCS provides harmonized and standardized legend for the Land Cover. The classification legend follows the dichotomous structure which can be identified and recognized anywhere in the world. The classification system leads to mutually exclusive Land Cover classes, which comprise of a unique Boolean formula (coded string of classifiers used), standard name and unique numerical code [Antonio and Louisa 1998]. LC LCCS Code LCCLevel LCCS own Legend LCCs Label 1 5003-15 A4-A13A16 Settlement Low Density Urban Area(s) 2 6005-6 A5-A12 Landsoil 3 20036 A6A10B4C2-B13C5 Grasses 4 10891-12006 A4B1B6C2D3B4C3C7C19D4 Agriculture Land 5 20594-13233 A3A10B2XXD1E1F1-B6 Eucalyptus spp 6 20602-13233 // 20606-13305 // 20608-53024 A3A10B2XXD1E2F1-B6 // A3A10B2XXD1E2F2F6F7 G3-B6F9G9 // A3A10B2XXD1E2F2F6F7 G3F2F5F10G2-B6F9G9G5 Acacia spp./Prosopis cinerea/Salvadora spp./Prosopis juliflora 7 20602-13314 A3A10B2XXD1E2F1-B7 Prosopis juliflora Stony Bare Soil And/Or Other Unconsolidated Material(s) Interrupted (Cellular) Closed Short Grassland Scattered Clustered Field(s) Of Surface Irrigated Graminoid Crop(s) (One Additional Crop) (Herbaceous Terrestrial Crop Sequentially). Broadleaved Evergreen Medium High Trees, Single Layer Broadleaved Deciduous Medium High Trees with Open Medium High Shrubs // Broadleaved Deciduous Medium High Trees with Open Medium High Shrubs and High Emergents Broadleaved Deciduous Low Trees, Single Layer Table 2: LCCS legend for Pai Forest Presented at a workshop on SAR and optical applications for natural resources and environment management, I SNET , November 2-5, 2009, GORS headquarters, Damascus, Syria Seven thematic classes were defined on the basis of structure and density of the features (Table 2). For example, pure patches (single layer) of medium high trees of Eucalyptus spp. (broad leaved) were present in the study area. This class was characterized as ‘Broadleaved Evergreen Medium High Trees, Single Layer’ with LCCS level ‘A3A10B2XXD1E1F1-B6’. Similarly, Prosopis spp. was defined as ‘Broadleaved Deciduous Low Trees, Single Layer’ with LCCS level ‘A3A10B2XXD1E2F1-B7’. The LCCS level contains the specifications of defined landcover class. The details of the class can be traced by using the above mentioned LCCS level. As the legend comprises of boolean formula, standard name and unique numerical code, it can be easily merged with other land cover maps at regional and global scale. Object Based Image Analysis Classification is the process of assigning pixels of a continuous raster image to predefined classes [4]. For the land cover mapping, different conventional classification techniques (unsupervised classification, supervised classification, hybrid classification etc.) are being used by the GIS Professionals. In this study advanced and the most recent classification technique i.e. Object Based Classification was applied on the satellite images. In conventional classification techniques, a specific class is assigned to a particular group of pixels on the basis of spectral reflectance values. These techniques do not cater to shape, pattern, texture and other visual image interpretation elements. Whereas in Object Based Image Analysis (OBIA), classes are defined by using spectral values as well as image interpretation elements [5]. Following figure describes the fundamental steps of OBIA. Figure 2: Fundamental steps of OBIA Definien provides various algorithms for segmentations i.e. Chessboard Quadtree, Contrast Split, Multiresolution and Spectral Difference Contrast Filter. Multi-resolution segmentation algorithm with parameter values i.e. scale = 10, shape = 0.1 and Compactness = 0.025 was used to get a segmented layer. Contradictory to the traditional classification techniques, it was quite efficient to extract segment layer from pan sharpened QuickBird image. On the segmented layer, different training samples were collected by using ground truth data, delineated objects, reflectance values and visual image interpretation elements. On the basis of these samples, a preclassification was performed which was refined to get the final classified layers. Thematic layers of 2003 and 2008 were generated by using QuickBird images. Output classified layers were exported in the ERDAS Imagine format (.img). Area of each class was also calculated for the change analysis. Results and Discussions As the Pai Forest area is heterogeneous and very few pure patches of certain vegetation types were picked as training samples, it was not possible to classify each class separately. Hence a mixed forest class of, Acacia spp., Prosopis cinerea, Salvadora spp. and Prosopis juliflora was formulated. An enormous decrease of 483 ha (40.57%) in mixed forest class has been observed within a period of five years. Analysis clearly reveals an increase of 75 % in Eucalyptus cover from 2003 to 2008. Prosopis spp. an exotic and perennial weed has highly invaded in the study area over the past few years with an increase of about 151 ha (128%) in the area. According to locals, Prosopis spp. was first time observed in Presented at a workshop on SAR and optical applications for natural resources and environment management, I SNET , November 2-5, 2009, GORS headquarters, Damascus, Syria 1990s (pers. comm. February, 2008) in the study area. It is now considered as one of the dominating vegetation of the area. 1400 1200 1000 800 600 400 200 0 2003 So il La nd G ra ss es /G ra zi ng Ar ea s La nd ar ve st ed Fo re st Ag ri c ul tu re /H M ix ed Pr os op is Eu ca ly pt us ju l if lo ra 2004 sp p. Area(ha) Figure 3: Temporal land cover/Land use map of Pai Forest Figure 4: Graphical comparison showing five year change in land cover classes Agriculture land and “grasses and/or grazing areas” are seasonal classes, the variation in area and intensity of these classes can not be assessed due to the use of satellite imagery of two different seasons. The classes ‘land soil and/or trees shadow” highlight the open land area. Tree shadows (mostly of Eucalyptus spp.) were also merged in this class to get more accurate figures. “Settlements/Buildup Area” class is a man made feature class. No significant change has been observed in the area/extent of this class. Presented at a workshop on SAR and optical applications for natural resources and environment management, I SNET , November 2-5, 2009, GORS headquarters, Damascus, Syria Conclusion and Recommendations The results demonstrate that the decrease of 483 ha in the Mixed Forest class defines high degradation rate for fuel wood. Acacia spp. dominates the Mixed Forest class and it is used as a fuel and fodder by the local communities. While considering Prosopis juliflora (an exotic and invasive species), an increase in the area from 117 to 268 ha is due to its wide spread and invasive nature. It is recommended to study and halt any impact of prosopis spp. on indigenous flora and fauna. It will help the foresters to develop forest products in a sustainable manner that can diversify and support livelihoods of local communities. Eucalyptus spp. is being planted in the study area in certain compartments by replacing other vegetation types such as Prosopis cineraria (kandi). It is widely accepted that Eucalyptus has high water consumption rate. As water is already scarce in the area to irrigate the Pai Forest, it is highly recommended to review the existing plantation regime Eucalyptus spp, if any. It would be pertinent here to highlight that plantation of Eucalyptus spp. has been banned in the Punjab Province. Since agriculture/harvested land are season dependent, the results cannot make any definite conclusion whether there is any significant change in this class. Acknowledgements Authors express their deepest gratitude to ESRI and Lieca for providing the software to WWF - Pakistan. Authors would like to thank WWF – Pakistan’s ‘Indus for All Programme’ for providing financial support to carry out field visits and data purchase. Special thanks to Dr. Ghulam Akbar, Director Indus for All Programme and Ms. Uzma Khan, Manager Conservation Programme, WWF – Pakistan for their cooperation, constructive comments and moral support throughout the study. References [1]: www.foreverindus.org [2]: Sadia Qamar, “Pai Forest faces conservation problems” Nawa-i-waqt Group of News papers October 17, 2007. [3]: Lillesand. T. M. and Kiefer, R. W. 2003. Remote Sensing and Image Interpretation. 4th ed. John Willey & Sons, Singapore. [4]: Sabins, F. F., Jr. 1997. Remote Sensing Principles and Interpretation. 3rd ed. New York: W. H. Freeman & Co [5]: Flack, J., 1995. “Interpretation of remotely sensed data using guided techniques”, Ph.D. Thesis, School of Computer Science, Curtin University of Technology, Western Australia. [7]: Antonio, et al. (1998) “Land Cover Classification System Classification (LCCS), Concepts and User Manual” Food and Agriculture Organization of the United Nations, Rome [8]: Perera ANF, et al. 2005. “Turning invasive Prosopis to improve livelihoods in Sri Lanka” Commonwealth Forestry Association World Congress. 28 February- 5 March 2005. Colombo, Sri Lanka. [10]: Swain, P. H. 1973. Pattern Recognition: A Basis for Remote Sensing Data Analysis (LARS Information Note 111572). West Lafayette, Indiana: The Laboratory for Applications of Remote Sensing, Purdue University [12]: Final report on “Sindh vision 2030” prepared by Planning and Development Department, Government of Sindh on July 2007 [13]: www.sindhwildlife.com.pk Presented at a workshop on SAR and optical applications for natural resources and environment management, I SNET , November 2-5, 2009, GORS headquarters, Damascus, Syria View publication stats