This file was created by scanning the printed publication. Errors identified by the software have been corrected; however, some errors may remain. Acquisition of Spatial Data by Forest Management Agencies: a Review Michael J o C o Weir1 Abstract. - Forest management agencies have a long tradition of surveying and mapping for the acquisition of a wide range of spatial data. These include: i) external (cadastral) boundaries; ii) permanent, internal, forest boundaries; iii) forest stand and other impermanent boundaries; and iv) roads and other topographic features. In recent years, the well-established processes of graphical map compilation and revision have been increasingly replaced by digital mapping techniques. Based on a limited survey, together with published information, this paper reviews the surveying and mapping techniques currently employed by forest mapping agencies for the task of digital database establishment and maintenance. Accuracy tolerances specified for different categories of information vary considerably. The survey showed that not all agencies apply checks to ensure that specified tolerances are achieved. It appears also that some agencies apply techniques whose achievable accuracy is inconsistent with specified tolerances. INTRODUCTION The acquisition of spatial data to assess the state of forest resources has its origins in central Europe during the 15th century (Griess, 1989). Since the early 19th century, when forest management based on reliable area statistics gradually became common practice in Europe, forestry agencies have employed personnel and established facilities for the specific tasks of surveying and map production. In Canada alone, about 30 million hectares are inventoried annually by provincial forestry agencies. These agencies maintain more than 37,000 map sheets at scales varying from 1:10,000 to 1:20,000 (Leckie and Gillis 1995). Spatial data acquisition by forest management agencies worldwide therefore accounts for a significant portion of mankind's total effort and investment in resource mapping. The aim of this paper is to present a review of the surveying and mapping techniques currently employed by forest mapping agencies for digital database establishment and maintenance. The presented information is mainly based on data Lecturer, lnternalional Institute for Aerospace Survey and Earth Sciences (ITC), Enschede. The Netherlands. 309 from a small survey of forestry agencies known to use GIs in production (research organizations, were not included). The aim of this survey was to determine: i) how forest management agencies acquire spatial data for input to GIs; ii) the accuracy tolerances (if any) applied; and iii) policies concerning the updating of the spatial database. A questionnaire covering these topics was sent to 33 agencies in various parts of the world and 25 replies were received. The remainder of the paper is divided into three sections. The first section briefly describes the main categories of spatial data required by forest management agencies. The technology used by the 25 surveyed agencies to acquire these data is summarized. In the second section, specified accuracy tolerances are examined in the light of the surveying and mapping technology employed. The paper concludes with some remarks on spatial data accuracy and error modelling in forest management. SPATIAL DATA ACQUISITION Although maps are probably the documents most frequently consulted during forest operations (Moser 1970), technical details their production are not widely published, neither in the forestry nor in the surveying and mapping literature. Leckie and Gillis, (op. cit.) provide extensive details of current forest mapping practice in Canada. A recent report (Kennedy et al. 1994) summarizes forest mapping activities in Europe. Probsting (1994) provides a useful review of applications of aerial photography in European forest mapping. Information content The management of forests and forest land requires a wide range of spatial information, usually obtained from a variety of sources. For example, a GIs to support the formulation of "Indicative Forestry Strategies" in Scotland uses 34 sources of biophysical and administrative data (Aspinall et al. (1993). Although requirements vary from agency to agency depending on forest conditions and management goals, maps and GIs databases generally comprise four main categories of spatial data, namely: cadastral boundaries, "internal" administrative (for example, compartment) boundaries, topographic details and information on the forest resources. Information on spatial data acquisition by the 25 management agencies was compiled for each of these categories of information (table 1). It should be noted that the data in the table refer only to the number of instances where a particular technique is mentioned as being used. The data do not consider the area concerned or the amount of equipment employed. Furthermore, many agencies apply more than one technique to acquire certain categories of spatial data. Table 1. - Sources of spatial data required by 25 forest management agencies (number of occasions mentioned). Source Cadastral boundaries Topographic features Internal boundaries Forest boundaries Supplied by other agencies as: maps digital data Acquired by the forest agency using: Land survey GPS Graphical transfer from aerial photographs Graphical stereoplotting Digital stereoplotting Orthophotography Satellite imagery Not required Cadastral information An exact knowledge of the area and limits of the forest land is essential for proper management. In many parts of Europe, large scale cadastral maps are the source documents for the compilation of the "basic forest map" (German Forstgrundkarte). Forest management agencies generally obtain inforrnation on land ownership boundaries from cadastral survey agencies. In some countries, however, the forest management agency itself may have official cadastral survey tasks such as the determination and demarcation of parcel boundaries on forest land (Wander 1981). Topographic information General topographic inforrnation is required for planning forest operations such as road construction and timber harvesting and also to provide a metric framework within which sample plots for forest inventory can be located. Most importantly, details such as roads or drainage features may define (parts of) forest stand boundaries. Topographic information therefore forms an integral part of any (graphical or digital) forestry database rather than simply a background against which details, such as forest types, are mapped. The survey of 25 forest agencies showed that they obtain topographic information primarily from other agencies. However, surveys by forestry agencies themselves are generally required to map additional details not shown on general purpose topographic maps and "new" features such as recently constructed forest roads. Land surveying, GPS and a range of photogrammetric methods are all employed for this purpose. When relatively approximate methods of graphical transfer from aerial photographs are employed, these additional topographic features are likely to be acquired with a spatial accuracy lower than that achieved by the national mapping agency, leading to inconsistencies in spatial data quality among different sources. Permanent boundaries established by forest agencies Many forestry organizations establish and maintain a "quasi cadastral" survey of the land under their jurisdiction. This may take the form of compartments or timber concession areas, the boundaries of which must be demarcated, surveyed and mapped. The survey showed that a wide range of surveying and mapping techniques are employed to acquire data on permanent "internal" forest boundaries. Unlike "external" property boundaries which, for reasons of accuracy, are primarily surveyed in the field, permanent forest boundaries need not be surveyed to cadastral mapping standards. Photogrammetric techniques are therefore feasible and, in fact, apparently predominate over field surveys. Forest information Information on stand boundaries and other "specialized" forest information is not generally available from existing sources. Acquisition of this category is therefore one of the main tasks of any forest surveying and mapping agency. Just over half of the agencies employ simple methods of graphical transfer from aerial photographs for this purpose. The operating costs of these "simple" graphical methods are not significantly lower than those associated with the use of rigorous photogrammetric methods (Weir 1981). It is therefore surprising to find this labour intensive approach still so widely used by forestry agencies in developed countries. This technique is employed by many provincial forestry agencies in Canada (Leckie and Gillis, op. cit.), although some are now replacing it by more efficient photogrammetric methods such as digital mono plotting. As table 1 indicates, forest management agencies make only limited use of satellite imagery for forest type mapping. Clearly, the spatial resolution of the present generation of earth observation satellites is inadequate for mapping at the scale, level of detail and positional accuracy generally required by agencies practicing intensive forest management. In many developing countries, however, satellite imagery may be the only source of timely spatial data on forest resources, although informal discussion with foresters from these countries suggest that there too, aerial photography remains the principal source of spatial data. Frequency of revision Unlike a soil or geological map which, if carefully produced, may serve as a useful document for many decades, a forest management map must be updated at regular intervals. Consequently, map revision is an important aspect of forest management mapping. In the intensively managed forest of western Europe, where management plans are traditionally prepared every ten years, management maps are revised on a corresponding 10-year cycle. Nowadays, map revision is increasingly carried out as a process of (digital) spatial database maintenance. Table 2 gives details of the frequency of updating employed by 16 of the surveyed agencies and indicates a tendency towards continuous or annual updating. Corrie et al. (1994) describe the database maintenance procedures employed by a timber company for 2.4 million hectares of productive forest in British Columbia. The database, which includes 250 base and thematic layers, is continuously updated, with all changes distributed to local offices on a nightly basis. Table 2. - Frequency of database updating by 16 forest management agencies. Continuous Several times per year Annually Every 2 - 5 years Every 10 years Cadastral boundaries Topographic features 7 9 1 3 1 1 3 2 2 Internal boundaries Forest boundaries In fact, experience with database updating is still limited. At the time of the survey (mid-1994), only two systems had been in operation for more than ten years. Eight agencies had being using their system for less than five years and some agencies had not yet done any updating. Interestingly, a few agencies indicated that forest stand information is updated by means of a complete remapping rather than by considering only the actual changes. Although this would appear to negate the advantages of having a digital database, there may be a limit to the amount of revision which can take place. One agency, which has been employing GIs since 1977, updates the database on a 2-year cycle and makes completely new overhaul every ten years. ACCURACY STANDARDS In forest management, spatial data are required to provide not only locational information but also, most importantly, area data for the calculation of timber volume and other quantities. Ideally, therefore, the errors associated with these area data should be known. Leckie and Gillis (up. cit. p81) report that most Canadian forest inventory maps are considered to have accuracies "in the 10 to 25m range", although as Gernrnell et al. (1991, p22 1) note, "most existing forest inventories do not have estimates of the accuracies of their polygon labels and boundaries". The survey of 25 forest management agencies suggested that, among the four main categories of information, cadastral boundaries tend to have relatively stringent tolerances. As indicated in table 1, however, this category of data is generally provided by other agencies according to predetermined standards. The other three categories of information are mostly all acquired to the same standards (although some agencies relax the tolerances for - relatively uncertain - forest stand boundaries). Actual values vary markedly from one agency to another or are expressed as a range (for example, 10 to 30m) within which map accuracy should fall. Tolerances for stand boundaries set by individual forest management agencies in Europe are generally in the range 2 - 10m rmse relative to nearby fixed detail. Although virtually all of the 25 surveyed agencies indicated that they set accuracy standards for the acquisition of spatial data, almost half (12) indicated that they do not apply any quality control measures to check if accuracy tolerances are actually being met. Interestingly, eight of these 12 agencies employ simple graphical methods for mapping forest boundaries from aerial photographs. Tests using a variety of instruments, photo-scales and operators (Weir, op. cit.) have shown that the accuracy of these methods is highly sensitive to terrain conditions, base map quality and operator skill. Their performance (under optimal conditions, 0.8 - 1.3mm rms plotting error) is generally inconsistent with required standards. CONCLUSIONS This paper has shown that many forestry agencies already work to, or at least aim at, "near professional" standards of surveying and mapping. In some cases, however, there is a need to define these standards more clearly and, in particular, to apply appropriate quality control measures. In forestry, spatial data acquisition is not simply a matter of map production. Data on the areas of individual forest stands form an essential input for forest management planning. Spatial data quality therefore has a direct impact on forest management operations and business decisions (Prisley 1994). The establishment and maintenance of a spatial database can represent a considerable investment for a forestry agency, and it may be difficult to justify measures to improve spatial data quality (Keefer 1994). In recent years, GIs researchers have laid much of the ground work needed to handle spatial error in natural resource databases. A lot of the research on error propagation (see, for example, Heuvelink 1993) is, however, concerned with situations in which the accuracy of the basic data is predetermined. Because they acquire a large part of their primary data themselves, forest management agencies are in the fortunate position of being able to chain backwards through the various GIs operations in order to define the accuracy of primary data acquisition needed to achieve the desired quality of the information product. In this way, the methods and costs of spatial data acquisition can be better matched to the value of the resource being mapped. ACKNOWLEDGEMENT This paper is based on information about spatial data acquisition kindly supplied by forestry agencies in many parts of the world. REFERENCES Aspinall, R.J ., Miller, D.R. and Birnie, R.V. 1993. Geographical information systems for rural land use planning. Applied Geography, 13:54-66. Corrie, G., Reedijk, W. and Lohia, K. 1994. A GIs to improve the competitive advantage of TimberWest Forest Limited. Proc. GIS'94, Vancouver:83-88. Gernrnell, F., Goodenough, D.G., Fung, K. and Kushigbor, C. 1991. Resource spatial and attribute information extraction from remotely sensed data. Proc. Spatial Data 2000, Oxford:221-231. Griess, 0 . 1989. Die Forsteinrichtung, ein klassisches geographisches Informationssystem. Proc. Symposium on New Technologies in Forest Management Planning, IUFRO S4.04, Luxemburg:8pp. Heuvelink, G.B.M. 1993 Error Propagation in Quantitative Spatial Modelling: Applications in Geographical Information Systems. Nederlandse Geografische Studies 163, KNAG, Utrecht: 15lpp. Keefer, B.J. 1994. Impact of spatial accuracy on business decisions: a forest management perspective. Proc. International Symposium on the Spatial Accuracy of Natural Resource Data Bases, Williamsburg :18-23. Kennedy, P.J., Paivinen, and Roihuvuo, L. (eds.) 1994. Proc. International Workshop on Designing a System of Nomenclature for European Forest Mapping. European Commission, Luxemburg:455pp. Leckie, D.G. and Gillis, M.D. 1995. Forest inventory in Canada with emphasis on map production. Forestry Chronicle, 7 1(1):74-88. Moser, W. 1970. Wie genau muB eine Forstkarte sein? Allgemeine Forstzeitung, 8 1(8):205-207. Prisley, S. P. 1994. Why natural resource information managers must be concerned about spatial accuracy. Proc. International Symposium on the Spatial Accuracy of Natural Resource Data Bases, Williamsburg:24-34. Probsting, T. 1994. Einsatz von Luftbildern in der Forstwirtschaft in Europa - Ein ~berblick.In T. Probsting (ed.) Photograrnrnetrie & Forst. Albert-LudwigsUniversitat, Freiburg:277-296. Wander, R. 1981. Die Forstvermessung in Landesteil Baden. Allgemeine Vermessungsnachrichten 819:329-338. Weir, M.J.C. 1981. Simple plotting instruments for resource mapping. Proc. International Conference on Matching Remote Sensing Technologies and their Applications, London:223-232. BIOGRAPHICAL SKETCH Michael Weir is a lecturer at the International Institute for Aerospace Survey and Earth Sciences (ITC), Enschede, The Netherlands. He is mainly responsible for teaching surveying, photogrammetry and GIs to foresters and other resource specialists from developing countries.