Land Cover Classification of Asia Region Using NOAAAVHRR 1-KM Dataset
Cheng-Gang Weng & Ryutaro Tateishi
Center for Environment Remote Sensing, Chiba Univ. Japan
Abstract:
With the development of global economy, environmental research has become more important than ever, especially in Asian region. The objectives of this study are to produce land cover classification dataset for different purposes in Asian region using NOAA AVHRR 1-km dataset. In this study, we used not only the NDVI data, but also the surface temperature data which has the great potential to yield a reliable land cover classification.
1. Introduction
Until now, NOAA's Advanced Very High Resolution Radiometer (AVHRR) serves am important role in global/continental land use management and planning, and analysis of land cover change. To data, most land cover mapping applications at broad spatial scales have been based on multi-temporal Normal Difference Vegetation Index (NDVI) data (Tucker et al. 1991). In order to extract more fully information contained in AVHRR data, the thermal data can also be used for land cover mapping. Kerber and Schutt (1986) have been AVHRR channel3 data, which is sensitive to reflected and emitted radiation, to locate the boundary between forest and non-forest. More recently, Lambin and Ehrlich have used multi-temporal NDVI and land surface temperature (Ts) and the ratio between these two variables to perform the land cover mapping based on AVHRR Global Area Coverage (GAC) data set over African continent.
2. Source Data
Global NOAA AVHRR 1-km 10-days composites dataset from April, 1992 to March 1993 are the main data used in this study. In addition, digital elevation data and country or regional level vegetation and land use maps are also used for the analysis of land cover classification.
2.1 AVHRR Data
Global NOAA AVHRR 1-km 10-days composites NDVI, channel4 and channel5 data from April, 1992 to March, 1993 were used in this study for development of land cover mapping. The NOAA AVHRR 1-km dataset is based on the Interrupted Goode Homolosine map projection. In order to corresponding different purposes, the data were transformed to latitude/longitude map projection (Plate Carree Projection) with 30-seconds resolution. Because the main purpose of this study is to develop land cover classification mapping of Asia continent and demonstrate the role of NOAA, AVHRR multi-temporal NDVI and brightness temperature data for vegetation monitoring and land cover classification at broad range area like Asia continent, the NDVI, channel4 and channel5 data were cut from East 25 degree to West 165 degree in longitude, North 90 degree to South 15 degree. When it is compared with seashore lines of Digital Chart of the World (DCW), there is less 0.5 pixel bias at all parts after re-sample for geometric registration. The size is 20400-pixel row by 12600-pixel line.
2.2 Digital Elevation Model (DEM) data
Elevation data are used to model ecological governing natural vegetation distribution, and are important for identifying land cover types and stratifying seasonal regions representing two or more disparate vegetation types. In this study, Digital Elevation Model, Version 1 (January 1998), developed by Global One-kilometer Base Elevation (GLOBE) Task Team was used.
2.3 Map Data
In this study, maps of ecoregions, vegetation, land use, and land cover are used in interpretation period and served as references date to collect ground truth information and guide class labeling.
3. Land Cover Classification System
It is known that there are two needs for land cover data: scientific needs and social needs. Some land cover surface properties and some land use information can be derived directly from remote sensing data. As required information is diverse due to various purposes, land cover data can act as a common environmental variable from which further land information can be derived. Based on this a suitable land cover classification system is used in this study. It is hierarchical structure and is easy for land cover type interpretation (Table 1. Tateishi and Wen,
4. Ground Truth Collection
Ground truth data in this study means geographically specified regions which are identified one of classes in the land cover classification system by class code. Collection of good ground truth data is a key issue for reliable land cover mapping.
In this study, ground truth data were collected mainly existing land cover maps and land use maps which were got from the members of LCWG/AARS of Asia and Russia. And one part ground truth data were collected by field surveys in central-Asia area. The detailed period and purposed countries are as followed:
-
From August 23, 1996 to September 2, 1996 from Almaty to Akmula of Kazakhstan
-
From July 5, 1997 to July 23, 1997 from Akmula to Kustanaj of Kazakhstan
-
From April 26, 1998 to May 8, 1998 from Almaty of Kazakhstan, through Uzbekistan, to Ashkhabad of Tukmenistan.
5. Methodology
channel 15 using split window algorithm (Price,1984).The formula of TS
Ts = T4+ 3.33(T4-T5)-273.15
Where T4 and T5 are brightness temperature in AVHRR channel4 and channel5 (in degree kelvin)
Authors applied maximum value compositing (Holben, 1986) to this AVHRR 1 km data, selecting independently the maximum value of 10-days composites NDVI and Ts over every monthly period . Ts responds both to short-term variations in energy balance, related to rainfall events and changes in soil moisture, and to seasonal changes. The monthly compositing of Ts data artificially removes the short time scale variations in 10-days compositing Ts, leaving only the seasonal trend. It mainly includes lower frequency information, which is related to land cover types (Lambin and Ehrlich, 1995). The monthly compositing of NDVI can remove cloud effects. And both of the monthly compositing of Ts and NDVI can also solve the data volume problems. Since Ts displays the opposite trend to NDVI when moving from dense to sparse vegetation landscape, the ratio between Ts and NDVI variables increases the potential for discriminating between broad vegetation classes. The slope of Ts/NDVI has been interpreted biophysically as regional surface resistance to evapotranspiration (Nemani and Running , 1989). This provides theoretical support for using this ratio in land cover analysis. The maximum Ts and maximum NDVI ratio (Ts/NDVI) were then computed for every monthly period in this study.
