0073-B1

Forest Monitoring in Russian National Parks Based on Integration of Remote Sensing and GIS

N. Malysheva^[1], S. Knjazeva, T. Zolina, O. Orlova, I. Voukolova and N. Rajchenko

Abstract

Diverse ecosystems are unique, valuable for scientific research, picturesque and attractive in a recreational respect. They are also more attractive for nature conservation and ecological tourism than the typical ones. The existing net of Russian National Parks is an example of these ecosystems. The areas of the parks vary from ten thousand to million hectares. According to Russian Nature Protection Law, one of the primary tasks of national parks is to conduct ecological monitoring, to preserve unique nature areas or restore disturbed nature complexes. As the main ecosystem component in existing Russian national parks, forests are considered to be the main indicators of unfavorable impact as well.

The forest monitoring strategy and a set of monitoring methods corresponding to parks' peculiarities are introduced. The results of testing the forest ecosystem monitoring for Russian national parks are demonstrated.

Introduction

National parks promote nature preservation and cultural heritage of a country. Russia, with its 35 national parks, is very rich in outstanding nature landscapes undisturbed by human activities. Park's areas vary in considerable range from ten thousand to million hectares. The largest, "Yugyd-Va," with its virgin boreal forests covers about 2 mln. ha. The smallest one, the "Curonian Spit," with unique landscape mosaic of forests and sand dunes, covers only 7.9 thousand ha.

Unlike many parks in other countries, "housekeeping" activities in Russian national parks are permitted only to the extent of maintenance of ecosystems in conditions attractive for visitors. National parks' services take care of preservation or restoration of unique and picturesque ecosystems intended for tourism, recreation, and ecological education. Ecological monitoring of a territory provides valuable information for the management of the parks. Forests, as the main element of natural complexes of many Russian parks, are the main objects in the monitoring of ecosystems.

The value of information provided by monitoring in Russian parks outgrows the national boundaries. This information is a very important part of long-term observations for international research programs on global climate changes, biodiversity, and global observation of forest coverage. It is also a basis for estimation of sustainability of human activities indices. In this context, when performing research in Russian national parks, one must consider internationally adopted criteria and indicators.

Presented is a set of aggregate indices that meet the requirements of international criteria and indicators of sustainable impact on the environment in protected areas. Remote sensing information was a primary source for data collection and many indices estimation.

Monitoring strategy

The monitoring strategy is developed for every park separately. In each case we analyzed the park's area, exotics of environmental conditions, ecosystems preservation degree, the urgency of nature protection actions, and so on. Such approaches to monitoring may be as diverse as the parks themselves. This principle of differentiation, together with monitoring objectives, should also apply to observation design. For example, three observation levels are involved in a full scale monitoring system. They are: space images, airborne survey and ground plots. In case of remote regions, with virgin forests undisturbed by recreation and former human activities, data collection, overall estimation of the degree of preservation and ecosystems dynamics could be limited to space image analysis. In the case of disturbed ecosystems of a small park, forest health assessment and nature preservation actions' control will be of great importance, and a monitoring system could be designed with two observation levels: airborne survey and selective ground plots. In addition, certain survey techniques, scales and spectral ranges should be chosen in order to be able to make damage assessments and identify not only forest stands as a whole but tree pathology as well.

The principle of selective surveys with respect to monitoring objectives helps to choose the most effective territory monitoring plan, to estimate ecosystem preservation degree, and then to conduct a detailed study of areas with anomalous aggregate indices or unfavorable trends.

Territory monitoring planning should also use the principles of stratification, specification of the sampling units and methods of selecting those units, again with respect to monitoring objectives.

A monitoring system is designed to be implemented in phases: observation planning first, then selecting of data collection method, obtaining data, data processing and analysis, and interpretation and summarizing of the results.

For phase of planning the stratified sampling principle should be the leading one. Stratification implies a division of an area into long-term sampling units in order to obtain aggregate indices and then detailed data. Area stratification could be based on either ecological or administrative and human activities characteristics. The ecological approach to stratification assumes that a territory is divided into relatively homogeneous formations, such as river basins and landscape complexes. The administrative approach assumes that the territory is divided into so-called functional zones (zones of certain conservation, protection and land use regimes) and is primarily desirable for administrative and economic purposes. The use of GIS applications makes it possible to present monitoring results in both of these two spatial frames.

Monitoring methods

Wide geographical variety of research areas made it possible to test different monitoring methods. Experiments have been conducted in different geographical subregions: from the smallest among Russian National parks, the Curonian Spit, to the largest in European part of Russia, Vodlozero Park. The following will give a more detailed description of these methods.

Data collection based on processing of space images

Evaluation of ecosystems conservation and its dynamics for Vodlozero Park is made by analyzing space images. Vodlozero is the last undisturbed wetland taiga complex of such a considerable size (about 500 thousand ha) in Europe.

Multi-temporal RS information consists of the following:

• CIR space photos obtained in 1978 from the Kosmos satellite in scale of about 1:250000 with a resolution of 5-10m;

• CIR space photos obtained in 1992 from the Resours-F satellite in scale of about 1:800000 with a resolution of 20m;

• Multispectral scanner images (MSU-SK) obtained in 1998 from Resours-O with a resolution of about 150m;

The space images are processed by means of two-stage classification using IDRISI software. Uncontrolled classification is carried out at first. Then the images were classified based on maximum likelihood to reference areas. The identification error did not exceed 10% for all classes.

The areas (pixels) of the classes of identified objects are calculated based on the results of multitemporal image classification. Then indices reflecting park' s ecosystems peculiarities and specificity characteristics (e.g. percentage of forest cover, share of bogs, share of water surfaces) were estimated. The indices are calculated for the park as a whole, for strata corresponding to river basins of a certain order and for test areas with unfavorable ecosystem dynamics. In order to conduct the calculations the images are superimposed on the boundaries of parks and river basins- the layer produced by means of ARC/INFO.

The comparison of indices obtained from space images in 1978, 1992 and 1997 is of great interest for evaluation of ecosystem conservation and its dynamics caused by natural and human factors. The interval between observations was 14 and 19 years.

The monitoring objective in the central part of the park is the conservation of virgin boreal ecosystems. The analysis of indices from space images of different years has shown that the area of bogs complexes in this part has increased by 15-30%. On the contrary, the percentage of forest coverage has decreased. The unfavorable trend in boreal ecosystems dynamics is probably a result of a worsening hydrological regime, growing bog areas, and the changing of forest resourcefulness and ecological potential.

A disturbed balance in natural complexes' is also registered in another test area, the park's buffer zone. The strictly protected zone of the park is adjoining the buffer zone that was formerly disturbed by intensive logging. The comparison of aggregate indices (see table) leads to a conclusion about changes in natural complexes' balance at the test area. Considering the length of park boundaries, the effect of industrial activity in the buffer zone on conservation of the park's ecosystems and its dynamics will be best evaluated by space image analysis.

Table 1. Aggregate indices for Vodlozero National Park buffer zone

The above results have proved the disturbance of the hydrological regime at the test area under intensive harvesting. The stretching of the clear cuttings up to strictly protected zones inside the park becomes evident by overlaying the classification results with the layer of vector boundaries. The lack of space data for the territory between mentioned observations makes it impossible to reach an accurate conclusion of what year the clear cutting was performed. Therefore, it is not clear when harvesting took place-before or after the park was founded in 1991. Probably, the blocks with clear cuttings from a neighboring forest were assigned to the park's territory the year it was formed. Clear cuttings in the buffer zone don' t appear on space images starting from 1997. Compared with the situation in 1992, forest areas increased by 15% due to regeneration of clear cuttings during 5 years. At the same time, bog areas have increased by 7%, and by 12% during 19 years. Changes in the balance of indices prove the disturbance of the hydrological regime caused by harvesting in the past.

Space image-aided ecosystem monitoring shows the necessity of detailed research with more precise methods in the frontier park's territories and delineative areas of the highest priority for observations. Figuratively speaking, the space images sound the alarm, but more detailed methods, i.e. airborne survey and field observations, will help to improve information reliability. Experimental study shows that space images are appropriate for evaluation of gradual, long-term ecosystem dynamics caused by both natural conditions and human interference in vast territories. Could these methods be used for small parks, and if yes, in what aspect of research? Let's illustrate the reasonable prospects for the space data applications in the context of the smallest Russian national park, the Curonian Spit.

The Curonian Spit belongs to the largest accumulative phenomena of the Baltic Sea. Continuous 300m-long sand dunes are up to 1 km wide and are among the highest in the world. They stretch 70 km along the peninsula. The divers-dune relief is combined with green forests, bright whiteness of sandy beaches and blue of the Baltic Sea.

Multispectral SPOT imagery gathered in 1996 is analyzed in the study. Two available controlled classification algorithms were used for SPOT analysis. They are maximum likelihood and Euclidean distance classifiers. The classification errors were estimated by control sampling with class models not involved in the study.

As a result of SPOT image classification, specific land categories are distinguished with the following accuracy. The sand dunes without vegetation and water objects are identified with probability 1, agricultural land and dunes covered by vegetation-with probability 0.96-0.98. The two classes of forest are identified with probability 0.8 for coniferous stands and 0.7 for deciduous stands. Though the coniferous stands are, in turn, could be of either of two subdivisions, the probability of their identification do not exceed 0.55. This example demonstrates the usefulness of SPOT imaging only for preliminary stratification of park areas and planning the sampling aircraft observations.

Thus, multilevel data collection for monitoring determines the level of space images resolution. The scanner high resolution and medium resolution data (SPOT and Resours-O) should be applied in research of large park areas as a whole and their zones. The high-resolution satellite images with 5-20 m (SPOT, Resours-F and etc.) are preferable for planning and later conducting a detailed survey of test areas.

Data collection based on digital airborne survey

For small parks with disturbed ecosystems, such as the Curonian Spit, the monitoring objective is nature protection management and forest health assessment. In this case, two observation levels including the airborne survey and ground sampling are necessary. Airborne surveying conducted by means of digital video or photo cameras is an irreplaceable source of expeditious information for park's service routines.

Modern digital survey equipment provides images with appropriate resolution ready for computer processing. Applicability of airborne video imaging depends on producing images in different scales, i.e. an overall view and detailed view, at the same time from the same aircraft heights due to the change of the lens focusing during the flight. The overall view observation level is similar to moderate scale aerial photography (1:10000-1:15000), and detailed level is comparable with large-scale aerial photography (»1:1000 and more). The combination of data on two levels is helpful for evaluating forest sites and its characteristics. For instance, it is possible to evaluate a disturbance degree for forest as a whole, then to estimate the forest stands and, at last, to characterize the separate tree crown. Video survey is suitable for monitoring of thinning cuts, regeneration and state of forest plantings, recreation impact, performance of nature protection measures, forestry activity and etc.

The overall view video images serve as backgrounds for placement of ground sampling plots and for georeferencing the detailed video frames in GIS system. It is also suitable for correction of forest site boundaries and territory division into zones with different industrial activities and nature protection regimes.

Image mosaics, interpretation results, and obtained indices are loaded into GIS. The topographic maps, forest management planning maps, thematic maps, forest inventory data, ground sampling data in combination with video image data are also loaded. The data could be used either separately or in combined analysis of video and photo information. As a rule, the image interpretation is not regarded to be a self-reliant process but a part of complex data analysis via GIS capabilities. All information sources including the video images should be overlaid upon basic topographic maps for accurate analysis. The rectified video mosaics would be more preferable for interpretation in GIS but separate video frames would not.

Video images interpretation for forest monitoring is an interactive process that combines computer analysis and a specific expert's knowledge with summary to be performed by an expert. Airborne survey over the territory for forest monitoring could take place once every 3-5 years depending on speed of changes. In case of natural disaster, the observation system should be supplemented by efficient observations conducted in short time span. For instance, the forest stands of the Curonian Spit were damaged by a hurricane in winter 1999-2000. The efficient airborne video survey to evaluate the severity of wind damage was done. Three forest stand classes such as sustainable, with disturbed sustainability and weakened sustainability were identified as a result of video image comparative interpretation before and after the hurricane.

Field observations must be included in monitoring system because remote sensing without common forest stand characteristics is never sufficient for maintenance of sustainable management of parks. For example, interpretation of detailed images can't provide data on changes in height and crown diameter, regeneration ability of tree species or information about forest health damage, pests and diseases. The down to earth data gathering is also needed for experts' education and training and thus for perfection of video image interpretation.

GIS application

GIS has been just recently introduced to Russian forest management practices and many National parks don't possess forest map databases produced by common forest inventory and planning operations. Therefore, a digital map database should be prepared before conducting monitoring and visualizing outputs as a map. It is not a simple task because forest maps are not practically referenced to real coordinate system of topographic maps. Organizational and technical difficulties, information source errors and limitations peculiar to software packages were the reasons of testing a variety of packages. In order to create National Parks' GIS, Mapinfo, TopoL and ArcInfo for WS were tested and specific advantages of these packages were combined.

Digital map databases for several national parks were formed or adapted for using in monitoring practice. The adaptation means the rectification of information sources and georeferencing to a coordinate system of a topographic basis map. The constituent parts of GIS are presented in figure 1.

Figure. Constituent parts of GIS for National Parks

The methodology we introduced will help to carry out monitoring tasks that deal with forest health, fire protection, forest management operations on a base of unified software and unified base map. This principal will ensure the comparability and compatibility of the monitoring results.

The GIS application will provide cartographic support for park management. Advantages of GIS in spatial analysis and different map compilations as a result of monitoring will be illustrated in the examples of Vodlozero Park and Valday National Park.

Visual representation of indices obtained from space images in combination with map data enabled us to undertake analysis of spatial ecosystem variety in river basins and over functional zones of Vodlozero Park. As a result, it was possible to conclude what kind of areas is preferable for development of recreation and ecological tourism. Measures to be undertaken to restore forest ecosystem potential were recommended in cases where an unfavorable trend of increasing bog area was revealed.

Of paramount importance for national parks is the biodiversity analysis and mapping by means of GIS applications. Flora and fauna diversity is of specific concern for Valday National Park, where GIS and an attribute database were used to compile thematic maps.

The territory is situated in the northern part of the Valdai Upland and has a very complex and picturesque terrain that was formed during last glacial epoch. The belt of large lakes stretching from north to south was one of the ancient stems of settlement of Northern Slavic tribes and historical commercial routes in the European part of Russia.

The maps provide characteristics of natural complexes and biodiversity, namely, mammal habitats and size of populations, forest and water bird species, sites of endemic herbal plants, endemic fish species. The basic information on biodiversity was collected by the Park staff during common field observations.

National parks monitoring system is currently in transition from a research phase to practical implementation. The nearest future tasks include GIS education of the staff, updating databases and loading the ancillary data, and further enhancement of monitoring systems with assistance of their originators.