Browsing by Author "Luck, Wolfgang"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Generating automated forestry geoinformation products from remotely sensed imagery
    (Stellenbosch : Stellenbosch University, 2018-12) Luck, Wolfgang; Van Niekerk, Adriaan; Stellenbosch University. Faculty of Arts and Social Sciences. Dept. of Geography and Environmental Studies.
    ENGLISH ABSTRACT: Private industry, national government departments, and the international community rely on geoinformation to optimise activities, enforce legislation, and assess global markets concerned with the use of natural resources. One of the main challenges faced by the remote sensing community is to derive thematic information from available imagery at the required speed, consistency, quality, and cost. The international space science community has organised itself within the Committee on Earth Observation Satellites to combine information systems for providing global information services as part of the resulting Global Earth Observation System of Systems (GEOSS). The European Union also initiated a programme called Copernicus, formally known as Global Monitoring for Environment and Security (GMES), which aims to provide information services as a contribution to GEOSS. Copernicus/GMES provides three core services consisting of land, marine, and emergency response, and two pilot services relating to the atmosphere and security aspects; each core service consists of several service elements. The forest monitoring programme is part of the land core services. At present only Landsat archives provide a consistent and affordable data source to serve the land core services on local and regional levels. This is particularly true for all services addressing the thematic field of forestry, as the relatively low vegetation dynamics of forestry (compared to other cultivated crops) can be served by Landsat data as it optimally provides one cloud-free (cloudless) observation (not necessarily cloud-free scene) per season. Landsat data is also freely available from the United States Geological Survey (USGS) and several other ground receiving stations around the world. This makes it financially viable to use Landsat data for providing operational services that would otherwise be too expensive to deliver using commercially purchased satellite imagery. Another advantage of Landsat data is that it is well calibrated, and includes multispectral bands covering the full passive remote sensing spectrum from blue light to thermal radiation. While the blue band is useful for the characterisation of atmospheric effects, the red, near-infrared, and shortwave infrared bands are suitable for the characterisation of different vegetation types, in particular forests. The spatial resolution of Landsat imagery (15–30 m) is also suitable for forestry applications, in particular when the monitoring of tree clusters instead of individual trees is required. Given these unique attributes of Landsat imagery, this study focused on the development of a processing chain for the automatic extraction of forestry geoinformation products from Landsat thematic mapper (TM)/enhanced thematic mapper (ETM)+ imagery. The products generated consist of a plantation and indigenous forest mask, and a broad genus classification for plantation forests. The products were generated and validated in three regions in South Africa, namely Cape Town (Western Cape Province); the Natal Midlands (KwaZulu-Natal Province); and the eastern escarpment and Lowveld region (Limpopo and Mpumalanga provinces). The results show that the products have an overall accuracy exceeding 93% for all areas, which will make them useful for forestry operations and planning. Although the resulting forestry products are evaluated in a South African context, the methodology can also be applied in other regions. The methods can also be adapted for application on other data sources such as those offered by the recently-launched Copernicus Sentinel 2 satellite and other commercially-operated satellites such as RapidEye, Resourcesat, and Spot.