Development of a Vegetation Density Map of Brunei Darussalam Kazimierz Becek Geography Department, Faculty of Arts and Social Sciences Universiti Brunei Darussalam Jalan Tungku Link, BE1410 Gadong, Brunei Darussala kbecek@gmail.com Abstract In this paper, a feasibility study of the development of a forest density map for Brunei Darussalam using readily available data is presented. Vegetation density is calculated from the elevation bias of the space shuttle topography mission data product (SRTM). This bias is roughly equal to about 50% of the vegetation height for a fully stocked forest, according to various studies. Naturally, any variations in the density of the vegetation will lower the level of the bias. With a map of the forest of Brunei Darussalam as a guide to the seven major forest types, the average SRTM elevation bias was calculated for every forest plot. This was done by subtracting the ‘bald earth’ digital terrain model (DTM) from SRTM. The DTM was developed from 1:10,000 contours. Subsequent analysis of the preliminary results indicates that the density of forest plots varies significantly. Visual inspection of some of the forest stands confirmed that forces of different types (forest fire, logging) have caused a drop in the plot-based forest densities. By relating the SRTM elevation bias of a forest plot to the elevation bias of a healthy forest plot, a map of relative vegetation density was developed. The relative vegetation density map can be used to estimate the amount of biomass in the country, even more accurately than using a map of vegetation height, although the primary purpose is monitoring in relative terms of forest degradation or depletion. 1. Introduction In recent years, an increasing number of countries (the USA and UK among others) have created, or are in the process of developing, a map of vegetation height in their territories. A diverse range of methods has been utilized, including LiDAR and airborne InSAR surveys. The need for such a map can be justified by the fact that the vegetation height is a parameter strongly correlated with the amount of aboveground biomass. While the area of vegetation was traditionally used to estimate the biomass, vegetation height would increase the accuracy of the estimate. Hence, this should lead to increased reliability of the assessment of carbon flux between the atmosphere and vegetation. It has to be noted, however, that the ‘vegetation height’ approach ignores the fact that the vegetation density can still be different for the same vegetation height. This can be due to both natural and anthropogenic factors (BECEK and ODIHI, 2008). In the remote sensing literature, vegetation density is referred to by a few indices, including the Leaf Area Index (LAI) and the Normalised Difference Vegetation Index (NDVI). Both indices suffer from several biases caused by saturation at a certain density level, seasonal variations, and problems with their measurements (PEPER and McPHERSON, 1998). All of this constitutes a body of evidence supporting a statement that those indices measure the translucency properties of vegetation cover rather than the biomass density as it should be understood, i.e., as the amount of organic material per volume unit. Apart from destructive measurements, there is no way other than using an allometric equation, which relates biomass to certain physical parameters of vegetation available through using remote sensing methods. As an example, the diameter at chest height, steam density, and tree height can be provided (FRANKLIN, 2001). A relatively new method of acquiring certain physical parameters of vegetation is provided by synthetic aperture radar interferometry (InSAR). This method is based on the fact that electromagnetic waves do not fully penetrate the vegetation cover, which is manifested in the vegetation bias of digital elevation products developed using InSAR (BECEK, 2008). The elevation bias can be calculated as the difference between the InSAR elevation and the ‘bald earth’ elevation of corresponding locations. This method was successfully used to estimate the amount of biomass in a landscape to a regional scale that was reported, for example, in (KELLNDORFER et al., 2004, SIMARD, et al., 2006, 2008, and WALKER, et al., 2005). The approach adopted in these investigations relied on a calibration procedure that allowed the conversion of the vegetation bias to the average canopy height. The calibration was performed using a LiDAR survey, which allowed use of the first/last return signal calculation of the true tree height. Once the canopy height was known, an estimate of the biomass was calculated. In this paper, vegetation density is measured in relative terms, i.e., as a ratio between the density of a plot of forest of a certain type and the typical density of a fully stocked forest plot of the corresponding forest type. Required densities are calculated from elevation bias of the space shuttle topography mission elevation data product (SRTM) (SLATER, et al., 2006). This bias is equal to about 50% of the vegetation height for a fully stocked forest (CARABAJAL and HARDING, 2006). Obviously, any change to vegetation density will alter the SRTM elevation bias (BECEK, 2008). This relationship can be used to study vegetation density. The main goal of the investigation is to demonstrate the feasibility of developing a map of vegetation density using commonly available datasets. These may include a digital terrain model (DTM), SRTM data, a forest map, and a multispectral satellite image. As a test area, the territory of Brunei Darussalam was used. 2 Area of Interest The following description of the area of interest was extracted from BECEK and ODIHI, 2008. The test site is composed of the territory of Brunei, a country located on the island of Borneo (4.5º N 114.5º E) with an area of about 5,765 km2. It is bordered on the north by the South China Sea and elsewhere by the Malaysian state of Sarawak. The climate is described as wet tropical; the monthly mean temperature is around 27.3 ºC. The annual mean rainfall varies between 2,300 mm and more than 4,000 mm. The climate of Brunei is influenced by the southwest monsoon (April-August) and the northeast monsoon (October-January). Brunei has mainly low relief, especially in the coastal areas and much of the western part of the country comprising the three districts of Brunei-Muara, Tutong, and Kuala Belait. In this part, flat alluvial deposits are common. The eastern part of the country is the Temburong District. The southeastern part of this district comprises the Temburong Mountain range with the highest peak of 1850 m in Bukit Pagon. A substantial area of the lower region of the Belait River system is covered by a peat land swamp, of which the Badas forest is a part.  Figure 1. Geographical context of the investigated site Source: BECEK and ODIHI, 2008. The present relief of the country was formed during the last significant sea level subsidence some 5,000-6,000 years ago. Brunei is drained by four major rivers. They are Sungai Belait (209 km), Sungai Tutong (137 km), Sungai Temburong (98 km), and Sungai Brunei (41 km). The total area of the natural inland waters is less than 0.25% or 15 km2. The eastern part of Brunei—the Temburong District—is detached from the rest of the territory by the Malaysian valley of Sungai Limbang. The estimated forest cover in Brunei in 2005 was 48% and 28%, respectively, for primary forest (48%) and secondary forest (FAO, 2005). The forests in Brunei are dominated by the mixed dipterocarp forest (46%) and peat swamp forest (18%) (See Table 1). The average forest stand area of a given forest type is 492 ha, varying in range from 5.9 ha up to 40,000 ha. The country is located in the tropical belt, which is characterized by persistent cloud cover (in average 7 octas between 9 and 11 a.m. local time), which precludes the acquisition of passive remote sensing data. 2. Materials and Method 2.1 Data The following datasets were used in the study:
The forestry map was manually digitized. About 1,100 forest plots have been identified. The major forest types of Brunei Darussalam and corresponding basics statistics are shown in Table 1. Multispectral data were used to calculate the NDVI. The NDVI map was subsequently downsampled to the 90-m pixel size. The pixels were also made spatially coincidental with the SRTM and DTM pixels.  Table 1. Major forest types in Brunei Darussalam (ANDERSON and MARSDEN, 1984). 2.2 Method The following data processing operations were performed:
 In order to identify a typical SRTM elevation bias for a forest type, a typical forest plot must be identified first. This search was necessary because of the time lapse since the forest map was drawn (1984). Since that time period, forests have been altered by natural or anthropogenic degradation or depletion processes, and the forests may not represent a typical forest plot of a certain forest type. To facilitate the selection, it was assumed that forest plots that had preserved their pristine status and can be considered as typical are those that have the lowest variation of their spectral characteristics. In this case, the NDVI was selected. Consequently, the average NDVI and its variation were calculated for every forest plot. Next, forest plots were sorted in ascending order of the NDVI variation. The top 1/3 of the plots are the most homogeneous, and were considered typical for a given forest type. Figure 2 shows the concept of selecting typical forest plots for a given forest type. Using those plots, a typical SRTM elevation bias as an average of all typical plots was calculated. Table 2 shows the typical SRTM elevation bias for the major forest types.  Figure 2. NDVI and its standard deviation for plots mangrove forest plots. The plots were ordered by ascending standard deviation of the NDVI. The top 1/3 were taken as typical mangrove forest plots. In the third step, the average vegetation density for every forest plot is calculated. It expresses in relative terms the density of a given forest plot in relation to the typical density of a plot of forests of a given type.  Table 2. Typical SRTM elevation bias for the major forest types in Brunei Darussalam 3. Results Figure 3 shows the final product of the investifations, i.e. plot based forest density map of Brunei Darussalam as per February 2000 SRTM data acquisition.  Figure 3. Map of the relative forest density of Brunei Darussalam as at February 2000. 4. Discussion and Conclusions The described procedure leads to the production of a map showing a relative vegetation density map of an area of interest. It is possible to convert this map to a map showing the density of biomass expressed in kg m-3 or Tha-1. In order to do that, detailed quantifications of the biomass of typical forest plots of various forest types are necessary. It could be done using a ground survey and/or LiDAR-based tree height extraction. However, this is not necessary as long as the absolute vegetation density values are not required. But this is the most frequent situation because the major purpose of creating such a map is the monitoring of change to the vegetation cover. A limitation of the final product—the vegetation density map—is the fact that SRTM represents the situation on the ground in February 2000 when the SRTM mission was flown. However, certain updating of the bias should be possible by using currently operational radar platforms such as the TerraSAR X-band satellite. To exploit this avenue, in-depth investigations of the issue are needed. References ANDERSON and MARSDEN (Forestry Consultants) Ltd, 1984. Brunei Forest Resources and Strategic Planning Study. The Forest Resources of Negara Brunei Darussalam. The Government of His Majesty the Sultan and Yang Di-Pertuan of Negara Brunei Darussalam. Final Report. Volume 1. BECEK, K., 2007. Comparison of Decimation and Averaging Methods of DEM’s Resampling. Proceedings of the Map Asia 2007 conference, Kuala Lumpur, Malaysia, available online, accessed July, 2008, at http://www.gisdevelopment.net/technology/ip/ma07267.htm BECEK, K., 2008. Investigation of Elevation Bias of the SRTM C- and X-Band Digital Elevation Models. Proceedings of the XXI Congress of ISPRS, Beijing, China, Vol. XXXVII (B1-1), pp. 105-110. BECEK, K., and ODIHI, O.J., 2008. Identification and Assessement of Factors Affecting Forest Depletion in Brunei Darussalam. Proceedings of the XXI Congress of ISPRS, Beijing, China, Vol. XXXVII (B2-1), pp. 209-213. CARABAJAL C.C. & HARDING, D.J., 2006. SRTM C-band and ICESat Laser Altimetry Elevation Comparisons as a Function of Tree Cover and Relief. Photogrammetric Engineering & Remote Sensing (PE&RS), Vol. 72, No. 3, pp. 287-298. FAO, 2005. Global Forest Resources Assessment 2005. FAO Forestry Paper 147, United Nations, Rome. Available online (July, 2008): http://www.fao.org/forestry/fra2005. FRANKLIN, S.E. 2001. Remote Sensing for Sustainable Forest Management. Lewis Publishers. KELLNDORFER, J.M., WALKER, W.S., PIERCE, L.E., DOBSON, M.C., FITES, J. HUNSAKER, C., VONA, J., CLUTTER, M, 2004. Vegetation Vegetation height derivation from height derivation from Shuttle Radar Topography Mission and National Elevation data sets. Remote Sensing of Environment, Vol. 93, No. 3, pp. 339-358. PEPER, P.J. and McPHERSON, E.G., 1998. Comparison of five methods for estimating leaf area index of open-grown deciduous trees. Journal of Arboriculture. 24(2), pp. 98-111, Avaiable online http://www.fs.fed.us/psw/programs/cufr/products/6/cufr_92.pdf, (accessed, July, 2008). SIMARD, M., RIVERA-MONROY, V.H., MANCERA-PINEDA, J.E., CASTAÑEDA-MOYA, E., TWILLEY, R.R, 2008. A systematic method for 3D mapping of mangrove forests based on Shuttle Radar Topography Mission elevation data, ICEsat/GLAS waveforms and field data: Application to Ciénaga Grande de Santa Marta, Colombia. Remote Sensing of Environment 112 (2008) pp 2131–2144 SIMARD, M., ZHANG, K., RIVERA-MONROY, V.H., ROSS, M.S., RUIZ, P.L., CASTAÑEDA-MOYA, E., TWILLEY, R.R. & RODRIGUEZ, E., 2006. Mapping Height and Biomass of Mangrove Forests in the Everglades National Park with SRTM Elevation Data. Photogrammetric Engineering & Remote Sensing (PE&RS), Vol. 72, No. 3, pp 299-311. SLATER, J. A., GARVEY, G., JOHNSTON, C., HAASEN, J., HEADY, B., KROENUNG, G. & LITTLE, J., 2006. The SRTM data “finishing” process and products. Photogrammetric Engineering and Remote Sensing, Vol. 72, No. 3, pp. 237-247. WALKER, W., KELLNDORFER, J. & PIERCE, L. 2005. Quality Assessment of SRTM C-and X-band Interferometric Data: Implications for the Retrieval of Vegetation Canopy Height. Remote Sensing of Environment, Vol. 106, No. 4, pp 428-448. |