Christoph Kleinn is Professor
of Remote Sensing and Forest
Inventory at the
University of
Göttingen, Göttingen, Germany.
Advancement of technology may improve efficiency of forest assessments, leading to cost savings and/or improving accuracy and precision - but revolutionary changes are not expected.
National forest inventories (NFIs) provide information relevant for national-level decision making, policy formulation and monitoring for forestry and related sectors, as well as for forestry planning in smaller geographical or political units at the subnational level. Since the results are an input to global forest assessments, there is also regional and global interest in high-quality national forest information. This article examines technological and methodological developments that can facilitate implementation of NFIs and provide improved information, thus helping to improve management and policy formulation through better-informed decisions at all levels.
New technologies and methodologies may affect an NFI in all phases (inventory planning, implementation, analysis and reporting). The extent to which they help improve NFIs can be evaluated according to the following criteria.
Do they help meet information needs better, by providing more information, new information or better information (higher precision of estimates, more accurate measurements)?
Do they reduce cost, by streamlining the NFI process at different stages? Fieldwork with its many costs (transport, DSA, helpers, tools, measurement devices) is a heavy cost factor in most NFIs [for example, in the Forest Inventory and Analysis Programme of the United States Forest Service about two-thirds of the programme cost arise from field-plot work (Guldin, 2000)]. However, it is a central and indispensable source of information for many forestry variables; it is, therefore, natural to seek options for cost reduction particularly in the context of fieldwork.
Do they improve the visibility or political standing of NFIs, for example by providing better and more widely available information, by increasing the overall credibility of results or by integrating new sets of attributes that increase the interest in the information to a broader group of users?
Structure, design and implementation of individual NFIs are as variable as the political and biophysical environment in which they take place. The potential role and impact of new technologies and methodologies will vary with the general biophysical, organizational and political conditions in each country, and also with the national history and background of the NFI.
Forest inventory fieldwork is an indispensable source of information for many forestry variables - technological developments that help facilitate fieldwork and reduce its costs are always welcome
- M. GARCÍA
Early NFIs relied almost completely on field observations. In the 1960s and 1970s, when FAO supported and carried out many Technical Cooperation Projects (TCPs) implementing NFIs in developing countries, field observations were still the single most important information source. Aerial photographs were used in some cases, and at the end of the 1970s, first attempts with the newly available Landsat multispectral scanner (MSS) were made. Satellite data and imagery then entered rapidly into NFIs, particularly in tropical countries, and the emphasis shifted from field observations to image interpretation. In some cases, mere satellite-imagery-based forest cover mapping was then denominated as "forest inventory". Map products derived from satellite image interpretation have since been the dominant output of forest inventories in many tropical countries. Mapping studies cost less than fieldwork (if the necessary hardware and software are available); need less planning, smaller teams and less broad expertise; are independent of weather; and provide maps as the major product, which are usually more easily accepted and "marketed" than statistics and tables with error specifications. However, ground truthing is essential to the interpretation of remote-sensing imagery.
Today, an efficient integration of different information sources is sought, including remote sensing. Fieldwork is particularly indispensable for a large range of variables that cannot be observed through remote-sensing technology within an acceptable range of accuracy, or at all. Remote sensing, combined with ground control, is the choice for mapping and landscape-level analysis.
Many non-traditional carriers for aerial cameras and video cameras, such as model airplanes, balloons, small dirigibles and ultra-light airplanes, offer good opportunities for local imaging, yet their implementation requires equipment, infrastructure and skills that are usually not available in an NFI. Digital photography (a form of terrestrial remote sensing) is not expected to open new avenues, since photographic documentation has not been of great importance in NFIs.
In the NFI context, remote-sensing data and imagery may be used for the following functions.
The estimation of forest area (or forest-type area) is a natural task for remote-sensing imagery. Given the right imagery and a workable definition of cover categories, this classification can be carried out with high accuracy. However, field sampling also usually provides sufficiently precise results, with superior classification accuracy. Rare classes will be observed better through remote-sensing imagery, when detectable with sufficient accuracy. Imagery-based area estimation not only provides summary statistics, but also allows description and analysis of the spatial arrangement and fragmentation of the forest area. However, sample-based approaches can be developed to estimate fragmentation status (e.g. Kleinn, 2000) where full-cover mapping is not required.
Only a few attributes can be directly observed from remote-sensing data and imagery. Crown sizes of individual trees and crown cover can be taken from suitable imagery. These attributes may support classification and can be taken as co-variables in analysis and modelling. Height of individual trees can be taken from large-scale air photos, and laser scanning offers possibilities for automated generation of height profiles in larger areas, but the utility of this potential for NFIs is expected to be minor.
Remote sensing allows spatially explicit observation of changes of forest types over large areas. For this particular task, remote sensing is indispensable.
Remote-sensing data and imagery co-registered and matched with field observations can be used in modelling. Whole area coverage is then not necessary; limited local imaging will suffice. Air photos or very-high-resolution satellite imagery plots around the location of the field samples may be taken to raise the general information content of field samples and make it possible to extend the spatially very limited field plot information to a sort of landscape-level representation of attributes such as fragmentation and spatial arrangement of land use classes and landscape elements.
The major analytical utility of remote-sensing-derived imagery in small-area NFIs lies in complementing the information from field plots or, in more general terms, in linking remotely sensed data to other geo-referenced data. Models relating remote-sensing features (in particular those determined through the ever-increasing spatial and spectral resolution of the imagery) to attributes relevant to the NFI may improve estimations. If applied and extrapolated to the whole area (if entirely under cover), such models may generate thematic maps that contain forestry relevant information. The Finnish NFI has long implemented a corresponding modelling approach (Tomppo, 1990), and ongoing research in geostatistics is expected to lead to spatially more accurate results and higher precision of estimations.
For mapping, spatial analysis and geo-referenced information, remote-sensing techniques offer the only operational options. Traditional NFI reports have contained more tables than maps, but maps and geo-referenced information are expected to be the major standard product of future NFIs. The step from large-area forest-cover mapping to landscape inventory covering other land use types (or tree inventory covering all tree resources) is not great; thus the scope of NFIs is easily extended, increasing their utility for other sectors. The mapping options offered by remote-sensing technology contribute significantly to better visibility and to a more client-oriented presentation of NFI results.
The development of satellite navigation systems was a major breakthrough in many fields. The Navigation System with Timing and Ranging - Global Positioning System (NAVSTAR-GPS), operated by the United States Ministry of Defense, was the first system in place; it was developed in the late 1970s and 24 satellites were launched from 1989 to 1994. Up to 1 May 2000, the signals were intentionally degraded to decrease precision for common users. Also, in times of political crisis the system may only be available to a selected core group of users (to which NFIs usually do not belong). The acronym GPS is also used for satellite navigation systems in general.
The Russian Government operates a satellite navigation system called GLONASS, and the European Commission recently launched a programme for its own system, named "Galileo", which is expected to be operational in 2008. The existence of three independently operated satellite navigation systems is expected to improve coverage significantly in the future, but only if receivers are developed that can process signals from all three systems. Coverage will particularly be improved for applications in which the range of the visible sky is limited, as in dense forests. Precision of position is likely to increase because of a higher likelihood of encountering better satellite constellations in all locations.
Use of GPS technology requires investment in hardware and training, but its cost is relatively low and its use is not complicated. Field crews usually prefer GPS receivers over the conventional position determination procedure. Satellite-based navigation systems have five basic applications: location (determining a basic position), navigation (getting from one location to another), tracking (monitoring the movement of people and things), global mapping and precise timing (Trimble Navigation Ltd, 2002). Only the first three have potential utility in NFIs.
Location and navigation, especially the latter, are the most important functions of satellite navigation systems in NFIs. To establish new field plots or to find plots that were established earlier, target points are geographically defined by coordinates and need to be located in the field. The traditional approach to navigation is to use maps to identify optimal starting points and to measure distances and azimuths, and then to follow these indications in the field with angle and distance measuring devices. Reference objects and observations are recorded on a sketch map to find the plots again during the next measurement cycle. Maps and local expertise will continue to be needed to identify the optimal access route to the field plots, but GPS navigation may replace the subsequent distance and angle measurements. This technology may make it possible to navigate with higher accuracy, particularly in regions where road infrastructure is not dense and where there are not many map-identifiable references.
GPS works perfectly in most regions when navigation is outside closed canopy layers. However, measurements inside dense forests are still not consistently effective everywhere. Searching for a canopy gap that allows suitable measurements can be time consuming, and the route from the gap to the field plot must still be navigated traditionally.
What are the benefits of GPS in the field? In many cases, particularly when open areas have to be crossed, navigation is much faster (and less costly) with GPS. The higher accuracy does not offer obvious benefits for the statistics generated by field-plot-based inventories. However, it is crucial for co-registration of field plot observations with maps or satellite imagery for modelling purposes. Location accuracy reduces "white noise" in the model. Thus the gain in accuracy offered by GPS is likely to improve the model and the corresponding extrapolations and predictions (e.g. Halme and Tomppo, 2001).
Satellite navigation systems can facilitate orientation under difficult topographic conditions like these
- M. GARCÍA
When field crews use GPS to navigate to field plots, their route can be automatically recorded so that a sketch map of the access path, which is traditionally drawn by hand and at rough scale, is directly registered. The resulting sketch map can be enhanced and complemented by verbal descriptions of reference points and waypoints. Tracking may also be useful for supervision of fieldwork, to cross-check whether the field teams came sufficiently close to the projected field-plot locations.
The main tree attributes measured in the field are diameters and heights. In addition, measurements of distances and angles are required for navigation, plot establishment and recording of tree positions when GPS is not used.
Ultrasonic and laser devices facilitate distance and angle measurements. Electronic distance measurement greatly increases the speed of navigation to plot locations, although GPS may replace it. Electronic distance meters (ultrasonic or laser based) facilitate the measurement of tree positions by distance and azimuth from the plot centre. If the measurement device is connected to a mobile data logger, data can be stored directly without the need for paper forms.
Some new ideas are currently being tested such as horizontal laser scanning to record tree diameters; from a plot centre, the surroundings are laser-scanned and the backscatter of the trees makes it possible to record their diameters at various heights. This technique does not currently have practical relevance, however.
Automated measurement of some variables may be possible in the future. This may speed up field measurements but is not likely to increase accuracy significantly.
Distance measurement by tape is common practice; electronic distance meters could expedite this task
- J. Fallas
Mobile data loggers have been used in NFIs since the late 1980s. Data input may be done with a keyboard or by cable or wireless from electronic measurement devices. The technology is not new, but its potential is perhaps not utilized to the extent possible. It is advantageous to perform data entry as close as possible (in terms of time and space) to where the data are generated. Procedures for checking data directly in the field allow field crews to correct errors and inconsistencies immediately, which improves data quality.
A next step will be direct connection of mobile data loggers to a central database via mobile communication and Internet access. The central database would then be permanently updated and checking procedures could be adjusted immediately and uniformly for all field crews. However, such online data entry is expected to provide less of a gain in quality than immediate digital storage of data in the field; a periodic data transfer (once a week or so) appears to be sufficient.
Mobile telephone communication increases safety for field crews because they can communicate more easily in cases of emergency, and it also allows online data entry to a central database. However, coverage for mobile phones is poor in many regions, particularly in the sparsely populated rural areas where many NFI field plots are located. Therefore, mobile telephone communication directly from the field is not technically feasible everywhere.
Depending on the forest type to be inventoried, even a diameter measurement can be a challenge
- RETANA
Geographic information systems (GIS) and Web-GIS offer many possibilities for presentation of results and for improved data and user access to information. Online presentation of NFI results increases their visibility and utility, particularly when the information system offers interactive information retrieval. Such approaches are under development, e.g. the European Forest Information System (EFIS) (Kennedy et al., 2001).
Much development is to be expected in model development, not only in remote-sensing data analysis as discussed above, but also in more traditional aspects of modelling related to volume functions, which are available worldwide. The accuracy and local validity of such models is frequently in doubt. Models for estimating biomass and biomass components are much less available than those for carbon estimation. Moreover, uncertainty in biomass estimations has been identified as one of the major factors that determines the uncertainty in carbon estimations. The quality of data, particularly volume-related attributes, could be considerably increased by improvement of the traditional biomass models.
In addition, improvements are expected in models relating attributes observable in NFIs to aspects of biological diversity.
Growth models and large-area prediction of the development of forests and their functions (scenario modelling) depend partly on NFI data (e.g. Päivinen, Roihuvuo and Siitonen, 1996). Fostering the use of NFI data in other areas, such as politics, economics and research, would increase the visibility of NFIs. However, rights to the use of NFI data are not clearly regulated everywhere. More open policies on the use of data would probably attract the interest of more institutions.
Sampling is employed for the assessment of most attributes gathered in NFIs. Full-cover assessment by satellite imagery is technologically possible for only a few attributes (such as forest area and related variables). Forest inventory designers have done pioneering statistical work in sampling theory. In early NFIs, sampling was driven by a blend of statistical and practical considerations. The large set of attributes assessed in NFIs does not allow simple optimization of sampling and plot design.
Systematic sampling is the most frequently used sampling design in NFIs, sometimes in combination with stratification and/or other estimators. The most frequently applied plot design employs clusters of subplots; in temperate and boreal forests these tend to be fixed-area circular plots or point samples (a specific plot design used in forest inventory where a fixed plot area is not defined), while in tropical regions they tend to be elongated rectangular plots (strip plots). In some NFIs, however, for example in Switzerland and formerly in China, individual plots have been used rather than clusters of plots.
New sampling techniques are constantly being developed, discussed and presented in forestry research. Some of the more recently described techniques, such as adaptive cluster sampling and importance sampling, allow impressive gains in precision for specific inventory questions. However, the established combination of systematic sampling with cluster plots does not yet seem to have a serious competitor. Adjustments of the cluster plot design for the assessment of particular attributes (for example, in the context of observing indicator variables for biological diversity, or for more efficient recording of rare objects) may have promise. Efficient integration of data from different sources, particularly integrating data from field samples with remote-sensing imagery, may be of benefit to NFIs (e.g. Schreuder, 2001).
Areas where new methodologies and technologies are expected to benefit NFIs
Methodology/ |
Main phases of a national forest inventory |
||||
|
Planning |
Implementation |
Data quality |
Analysis |
Reporting | |
Remote sensing |
x |
|
x |
x | |
|
Satellite navigation |
x |
|
x |
||
Measurement devices |
x |
||||
Mobile information and |
x |
x |
|||
Software and algorithms |
x |
x |
x | ||
Sampling options |
x |
|
x |
||
NFIs are dynamic undertakings, prone to changes in scope and objectives. Although revolutionary changes in technologies and methodologies are not expected, technological developments will continue to contribute to change in NFIs at all stages, from planning to presentation of results
(see Table).
The technological and methodological developments described in this article for the estimation of some core attributes - such as forest area, volume/biomass/carbon, ownership, industrial and non-industrial wood extraction and biological diversity - are not expected to contribute significantly to improving the precision and accuracy of the estimation of state and change. Improvement is mainly expected through methodologies such as improved algorithms and models, particularly for the estimation of the derived attributes of volume, biomass and carbon and for estimates related to wood extraction and biological diversity.
Any new procedure needs to be integrated into the NFI process carefully; as expressed by Iles (1995), "the big gains are to be made by systems which show balance, backup and flexibility". In many regions the major impediment to NFI implementation is its cost, and the challenge is to provide economic justification for the expenses of a national-level assessment of the forest resource.
Bibliography
Guldin, R.W. 2000. Remote sensing at the dawn of the new millennium: a Washington, DC perspective. Keynote remarks presented at the RS2000 Conference, Albuquerque, New Mexico, USA, 13 April 2000.
Halme, M. & Tomppo, E. 2001. Improving the accuracy of multisource forest inventory estimates by reducing plot location error - a multicriteria approach. Remote sensing of Environment, 78: 321-327.
Iles, K. 1995. Constructing a safety net in the circus of high technology. Paper presented at IUFRO World Congress, Tampere, Finland. Available on the Internet: www.island.net/~kiles/p-iufro.htm
Kennedy, P., Folving, S., Munro, A., Päivinen, R., Richards, T., Köhl, M., Voss, M. & Andrienko, G. 2001. European Forest Information System - EFIS. A step towards better access to forest information. Paper presented at the IUFRO conference Collecting and Analyzing Information for Sustainable Forest Management and Biodiversity Monitoring with Special Reference to Mediterranean Ecosystems, Palermo, Italy, 4-7 December 2001. Available on the Internet: www.geolab.unifi.it/iufro_conference/present/kennedy/1.htm
Kleinn, C. 2000. Estimating metrics of forest spatial pattern from large area forest inventory cluster samples. Forest Science, 46(4): 548-557.
Päivinen, R., Roihuvuo, L. & Siitonen, M., eds. 1996. Large-scale forestry scenario models: experiences and requirements. Proceedings of an International Seminar and Summer School, Joensuu, Finland, 15-22 June 1995. EFI Proceedings No. 5. Joensuu, Finland, European Forest Institute (EFI).
Schreuder, H. 2001. Sampling using thematic mapping, very large scale photography, and ground sampling for measurement and strategic surveys. In Forest biometry, modelling and information science. Proceedings of a IUFRO 4.11 conference, London, UK, June 2001. Internet document: cms1.gre.ac.uk/conferences/iufro/proceedings/Schreuder4.pdf
Tomppo, E. 1990. Designing a satellite image-aided national forest survey in Finland. In The usability of remote sensing for forest inventory and planning, p. 43-47. Proceedings of an SNS/IUFRO workshop, Umeå, Sweden, 26-28 February 1990. Report No. 4. Umeå, Sweden, Swedish University of Agricultural Sciences, Remote Sensing Laboratory.
Trimble Navigation Ltd. 2002. How GPS works. Internet document: www.trimble.com/gps/triangulating1.html
