The Contribution of Remote Sensing for Afforestation and Forest Biodiversity Assessment

Barbara Koch

Forestry Faculty, University of Freiburg

Remote Sensing and Landscape Information Systems

Tennenbacherstr. 4, 79085 Freiburg, Germany

Tel: ++49 761 203 3694

Fax: ++49 761 203 3701

e-mail: ferninfo@ruf.uni-freiburg.de

Remote sensing has been a tool for environment assessment and monitoring for many years. At the beginning for many projects the successful demonstration of the tool remote sensing was more important than the provision of repetitive information for different user communities. During the last years the aforementioned attitude changed towards more service oriented applications. Within the field of environmental studies the integration of remote sensing for afforestation and biodiversity assessment is of increasing interest. While afforestation is especially in connection with climate change and carbon emissions of interest within the frame of European environmental policy, biodiversity has only become world wide of relevance after the United Nation Conference in Rio de Janeiro 1992, when political support was stated for biodiversity development and preservation.

The presentation will try to give an overview on the status quo integrating remote sensing for assessment and analysis of afforestation and biodiversity.

While talking about assessment or mapping of afforestation by remote sensing it is necessary to have a clear and uniform definition on afforestation. Only a clear, uniform and widely accepted definition allows the processing of useful products. But what is afforestation and what is the difference to the term reforestation, and do we mean afforestation only or also reforestation when we use this term? For further discussions how to integrate remote sensing for afforestation assessment it seems of advantage, to have a clear definition of the term afforestation. Nevertheless is there a clear definition?

When discussing afforestation it is a prerequisite to know the definition of forest. We all have experienced that the definition of forest is difficult and there are many definitions differing from country to country. It is not possible to go into all different kind of forest definitions, nevertheless it becomes clear while talking of afforestation it always needs the reference to what is a forest or woodland. Otherwise it is not possible to define afforestation. There is a need to know how the class physically looks like and how it is legally defined, before it can be mapped.

When it is clear how the class forest is defined, it still needs the differentiation between afforested land and other land regrown by forests. The IUFRO working group 6.03.02 found 34 definitions for afforestation. For example:

• Establishment of forest crops by artificial methods, such as planting or sowing on land where trees have never grown. [Source: Timber Harvesting and Engineering Glossary, USFS],
• The process or act of changing land into forest by planting trees, seeding, etc. on land formerly used for something other than forestry. This can obviously be contrasted with deforestation. [American Forestry; v100; 23-25; 1994.] [New Scientist; v143; 30-35; 1994.],
• establishment of forest on unforested lands or areas that have been deforested for at least 30 years (WWF/IUCN 1996),
• a land use change from different uses to forests. Both natural regeneration and artificial planting count as afforestation. (Austrian Forestry Act (Federal Legal Gazette no. 440/1975, as amended Federal Legal Gazette 231/1977, 142/1978 and 576/1987).

The definitions show that afforestation is in general used for a change in land cover or land use in favour of forests or wood land. In contradiction to regeneration and reforestation, afforestation is used for lands which have never or for a long time of period not been forested. The time of period which is considered as long time differs quite strongly between 30 and 100 years. In most cases afforestation is meant as an active transaction and not a natural development. Therefore afforestation is an action term, that means action takes place to change land cover or land use. While inventory or monitoring this action it has to be asked how long this term can be used, after the action is completed? There is no answer yet, but a definition is needed in order to map this class.

In respect to the aforementioned it becomes clear, that only a clear definition of the term afforestation provides a basis to decide if and how an assessment of afforestation by remote sensing is possible.

Due to the very unclear definition in many papers which have been screened, this paper reports now the status quo for application of remote sensing to map afforestation and reforestation.

The studies on mapping of afforestation and reforestation by use of remote sensing data report in general a good accuracy for the assessment of regeneration classes based on the spectral response of optical multispectral satellite data (Weißflog 1993), but most papers also point out problems to separate regeneration from pasture or idle land (Keil et al 1989, Schardt 1990, Bodmer 1993, Nagata et al. 1997). In general the assessment of structural information helps to map afforestation and reforestation. Therefore visual interpretation or textural measures are included in some studies. While textural measures only seem successful by use of digital aerial photography (Hudak 1996), the interpretation of textural information is helpful also for high resolution satellite data.

The application of radar data seems to be successful to get clear boundaries between wooded and non-wooded land while using L- or P-band data in cross or horizontal polarisation (Koch et al., Yanasse et al., Ranson & Sun, Saatchi et al.). A separation of regeneration from other forest classes is also reported (Kasischke et al. 1997). Nevertheless most articles dealing with radar and forest assessment are focused on deforestation and not afforestation.

The remote sensing literature indicates that there is potential for the assessment of aforestation or reforested area. Nevertheless the results presented do not give clear statements, because in most cases they are not explicitly dealing with aforestation and refer in general not to a detailed definition on aforestation.

In order to make application of remote sensing methods successful at least on European level common definition of aforestation areas is needed. On sublevels for countries and states own definitions, in respect to their needs, have to be defined. Only if the remote sensing community puts up a suitable and clear classification system, remote sensing can act as an successful mapping tool.

The remote sensing classification system might differ from the terrestrial classification systems, but has to be consistent in itself.

In respect to the work done in former years a broad knowledge base can be used to set up a classification system. A good reference is the FIRS report (1996) on "Definition of a System of Nomenclature", which gives an excellent overview on the feasibility of forest attributes for different remote sensing facilities.

While afforestation is often discussed together with climate change, carbon emission, water management and soil erosion, the impact of afforestation on biodiversity was not a major theme until now. Nevertheless the habitat changes connected with afforestation have an impact upon biodiversity. With afforestation non-forest cover types, most likely pasture or idle land, will be changed into wooded land. An increase in forest area will effect all different kind of taxa and change species composition in all levels of biodiversity. While the shift towards forests reduces some cover types more than others, like pasture land is more likely changed into forests than arable land, some species groups will suffer more than others. There is the opinion that afforestation might lead to a reduction in biodiversity, but there is no good evidence for this and there is no idea to what level biodiversity is influenced under different scenario as well as to what extent this has to be rated negatively.

The aforementioned proves that the impact of afforestation on biodiversity is one area which needs to be studied intensively among others. Independent from afforestation the assessment and monitoring of biodiversity in general is an important topic. Since the United Nation conference on Environment and Development in Rio, in 1992, the importance to preserve and improve biodiversity was political accepted and initiated world wide activities to characterise and measure biodiversity. Nevertheless since then there is confusion about what is meant by biodiversity and how biodiversity should be measured (Jeffer 1996). Using the term biodiversity, even missing a detailed definition it should be clear that this term includes all organisation levels, from the biom to the gene and is not restricted to species. Even so this is obvious, in the past it was given emphasis to species diversity on the habitat level. So comments Plachter (1997), one of the leading scientists in the field of biodiversity studies in 1997 that most biodiversity investigation are restricted to one indicator species or in listing of endangered species. He requests new computer based methods to realise an area wide and quantitative assessment. In respect to the perception of the insufficiency of species counting and the consciousness about not being able to provide full assessment of species (Bastian und Schreiber 1994) a new idea has developed to characterise more the biodiversity of living spaces. Hunter (1990) for example describes for the classification of forest diversity 7 criteria:

• species composition
• age structure
• horizontal spatial heterogeneity
• edges
• islands
• vertical structures
• dead trees

In this respect he also mentions how valuable it is not only to have a qualitative description but a quantitative approach. He reflects the opinion of many ecologist within the North-American region, like Turner & Gardner (1991), who believe that the description of the landscape structure is superior for the ecological valuation of living spaces than the assessment of species.

In this context it becomes obvious that structural diversity, neighbourhood relationships and dynamic processes in space and time is important. For the analysis and visualisation of spatial relationships Geographic Information Systems have achieved high relevance while for the assessment terrestrial approaches are still most important.

Nevertheless terrestrial approaches often have a deficit in assessment of structural criteria and neighbourhood relationships especially on landscape scale level, because terrestrial methods have to restrict to samples or small areas. A study by Kareiva and Anderson (1988) showed that in 95% within a sample of published paper the plot size was less than 1 ha.

However, to improve the understanding of ecosystem functions and processes and to develop a holistic description of the landscape, both intensive studies of small areas and assessment of much larger areas are required. Assessments of biodiversity need to include investigations at different scales. Crossing these scales can be very difficult and pattern as well as scale represent a central problem in ecology (Levin 1992). Remote sensing can provide information from species to global scale. The grain, the smallest unit in landscape ecology is defined by the pixel size or the scale dependent resolution of aerial photographs. Remote sensing can provide information on species and structural diversity and their changes over time either as primary or secondary information. Primary information refers to the use of information directly from the digital image values like statistics from spectral or textural features, while secondary information refers to information from classified or interpreted data. Based on classified data information on area metrics neighbourhood relationships, structural elements and species types is derived. This allows measures and indices indicating biodiversity values, visualised in maps.

Within the past remote sensing has demonstrated to be a useful tool for biodiversity studies, particular in relation to the assessment of ecosystem diversity at landscape level. A number of different remote sensing technologies are available and the required resolution of the end product determines which will be the most appropriate. Until now aerial photography (Münch 1993, Ihse 1994, Köhl 1996, Miller et al. 1996, Holopainen &Wang 1998) and digital optical satellite images (Skole and Tucker 1993, Aspinell 1994, Aspinell and Birnie 1994,Gulinck 1994, McCormick&Folving 1998, Marchetti et al. 1998, Jusoff 1998, Legg &Laumonier 1998, Wellens et al. 1998, Albert 1998, Dewi et al. 1998) are the two main technologies used to derive information. We all know the potential of aerial photography getting detailed information on tree and stand composition as well as horizontal and vertical structural information. The limitation associated with the use of satellite data in terms of pixel size, stereoscopic information and weather dependency are likely to decrease within the next years. A new generation of high resolution satellites is planned, with improved temporal frequency of data acquisition and often stereo capability.

In addition there are other promising remote sensing technologies which will in future possibly provide very interesting information for the assessment of biodiversity. One of such seems to be radar image data. Much research has been carried out over vegetated areas. This has shown that the interactions between microwaves and vegetation is very complex. However there is evidence that surface roughness and stand geometry are the main driving parameters influencing the backscatter response. Investigations carried out do not give a consistent picture but the results seem to prove, that deforestation, forest boundaries and the detection of flooding under vegetation canopy are successful applications (Keil et al. 1994, Haas et al. 1996, Koch et al 1996, Evans & SChmullius 1997, Kasischke et al. 1997, Kremmers 1997, Saatchi et al. 1997 and Horlacher et al. 1998). The integration of radar data might therefore improve the remote sensing data for biodiversity studies, especially in area with high percentage of cloudy weather situations. But since there is no significant evidence that space-born radar data itself or the integration of space-born radar data will distinctly improve remote sensing data sets for biodiversity studies, until today the use of radar is more a matter of research than an operational tool for use in biodiversity studies. This seems at least true for areas where the chance for the assessment of optical satellite data sets is quite good.

Another quite promising remote sensing tool for biodiversity studies is laser scanner data. Until now laser scanning is expensive and restricted to airborne applications, therefore its use is more for local investigations. Besides the mentioned restrictions laser scanner data seem to be very valuable for biodiversity studies, because they provide detailed information on height and vertical structures (Nilsson 1996, Hugh & Wehr 1997, Friedländer 1998). The exact digital measurement of terrain height, vegetation height and the visualisation of vertical structures provides information which is requested from many biodiversity scientists. The possibility to have an automated quantitative measurement of the above mentioned 3 D structural criteria will improve the quality biodiversity studies.

Thermal remote sensing is another method which was applied for ecological studies. Luvalle & Holbo (1991), Pierce & Congalton (1988) and Mayer (1988) have provided information on temperature distribution pattern. The relatively rare application of this technique to biodiversity studies results from the difficulties in calibrating radiative thermal energy to surface temperature as well as the small number of satellites with thermal channel, their fixed day-time data take and their relatively coarse spatial resolution.

In a wider sense remote sensing cannot only provide information on the surface from a bird’s view perspective but also terrestrial photogrammetry has high potential for large scale assessment of biodiversity indicator parameters in cases where particular plant types and their structure is important. With this method single trees and bushes or groups or section of landscapes are measurable. Terrestrial photogrammetry can contribute to microhabitat analysis. The structure and dimension of trees and bushes can be assessed quantitatively. Measurements of crown cover, crown structure, stem form and other vertical and horizontal structures are possible (Koch & Reidelstürz 1998).

While satellite remote sensing data are most appropriate for small scale to medium scale studies, airborne remote sensing is used for medium to large scale studies. The manifold applications of airborne or spaceborne remote sensing have been demonstrated by European scientists, like in the workshop on "Remote Sensing in Landscape Ecology Mapping" (EUR Report 1995) or during the Conference on "Assessment of Biodiversity for Improved Forest Management" in 1996 (EFI Proceedings 1996). McCormick & Folving (1998) structured the components of biodiversity in three main categories, composition, structure and development. For all three categories remote sensing can provide substantial information based on different indicators measurable out of remote sensing data. For satellite data McCormick & Folving (1998) describe some of the measurable indicator variables and procedures for information extraction.

According to literature there is no doubt that remote sensing , in spite of limitations, like weather dependency or restriction to surface information and pixel size, is avery valuable tool for biodiversity studies. The major benefit of remote sensing is the total coverage of extended areas, which assures easy assessment of structural and compositional parameters in a synoptic way. The range of measurable indicator parameter with total area coverage allows the assessment of habitat quality within certain landscapes. This approach supports the tendency of biodiversity specialists to focus on environmental conditions, because in general a species is best protected by protection of the habitat quality. Remote sensing tools are particular useful when assessing habitat quality, at landscape scale Depending on the appropriate scale of assessment different remote sensing tools are applicable, ranging from terrestrial photogrammetry to low resolution satellites. Operational tools at present are most likely aerial photography and optical satellite data.

Despite of this success story remote sensing is still under-utilised in studies of forest biodiversity (Stom &Estes 1993). What are the reasons for this? One might be that the remote sensing community was only concentrating on biodiversity aspects within the last years, but the major problem is that remote sensing and biodiversity research communities need more interactions. Results from remote sensing and ground studies are not enough related to each other. In order to improve the acceptance of remotes sensing in biodiversity studies a close co-operation between the two research groups is needed to develop procedures and strategies to integrate remote sensing tools successfully into biodiversity studies.

I thank Mr. Thomas Pappvary for assistance with the literature for this paper