Meteorological disasters/Agriculture and forestry damage/Insect damage. Forest disasters/Insect damage.
Disaster name
Japanese Red Pine Wilt Disease Damage in Shibata and Toyoura-cho, Niigata Prefecture

Japanese Black Pine Wilt Disease Damage in Niigata and Maki-cho

Author of WEB
Abe Nobuyuki

Case Study

No. 1

1. Analysis objective

Pine forests throughout Japan have been damaged by insects such as pine bark beetles, greatly reducing pine forest resources. The damage initially occurred in western Japan, and has been gradually spreading into eastern Japan. In Niigata Prefecture, the damage is currently spreading actively from regions where the insects have been eradicated, resulting in the presence of many areas where the blight is expanding.

"Understanding the actual conditions in this type of damaged forest area is not an easy task. Although damaged areas are reflected in statistics as a numerical value, the entire area of a forest grove damaged by pine bark beetles, for example, will be added as damaged forest when a forest owner finds damaged trees. In many instances, such figures are only remotely related to actual conditions.

If forest changes can be understood by applying the results of satellite data analysis, authorities can be expected to correct the information in their woodland holdings registers and prefectures, will be able to maintain forest information that more accurately reflects actual facts.

"Using Niigata Prefecture as a target, we attempted to sample portions of pine forests that have been damaged.

The subject areas were Japanese red pine forests in Shibata and Toyoura-cho in Niigata Prefecture, and Japanese black pine forests near the coast in the city of Niigata and in Maki-cho (Figure 1).

These cities and towns first began to suffer damage to their pine forests around 1985, with the devastation reaching a peak around 1990. Such damage is still continuing to spread.

2. Analysis procedure Analysis flow chart

To analyze the occurrence of forest damage from changes in brightness values, we used satellite data from two points in time. The satellite data utilized was Path 108, Row 34 LANDSAT/TM data, specifically 5 photographs taken on June 4, 1984, June 2, 1989, June 26, 1992, June 24, 1997 and August 1, 1999. In addition, we used 4 photographs taken on October 10, 1984, December 8, 1988, November 4, 1993 and December 4, 1998 to obtain samples of the base land cover.

We made geometric corrections of all of the data by affine transformation, which enabled us to control the overlapping error at the same position to less than +- 1 pixel.

For the analysis, we mainly used the pattern development method(1). Unlike the traditional supervised classification method, the pattern development method creates basic spectral patterns (S, V, W) from the portions of TM data for soil, vegetation and water, and recreates each pixel spectral pattern by merging the three basic pattern alignment.

A = CsS + CvV + CwW

The coefficient (Cs, Cv, Cw) of each basic pattern is called the pattern value. These values may be thought to be proportional to the respective cover ratios in each pixel. The pattern development method has the advantage that multiple data from different observation days and places can be analyzed using the same information space.

To sample the spectral features created by the albedo value of the nth wavelength of water, vegetation and soil, the three base ground covers make an image that standardizes the albedo values to provide the representative samples for water, vegetation and soil.

We determined this by averaging multiple samples taken from images in winter when living organism activity has declined the most, images of deciduous trees in June when they are the most active vegetation, and images in summer and winter, for water, vegetation and soil respectively.

Develop the albedo spectrum for arbitrary pixels by using the three base patterns. The development coefficients correspond to the quantities related to the quantities of light reflected from each respective land cover within one pixel. The development coefficient of vegetation corresponds to the albedo volume limited to just the vegetation cover in one pixel.

Information on the percentage within one pixel that is accounted for by the vegetation cover, or information on the type of vegetation and activity of the vegetation, is included in the vegetation development coefficient (2).

3. Analysis results

We conducted a field survey in the city of Niigata and in Maki-cho during October and November 1999. The damage rate was shown by the ratio of the total sectional area of damaged trees to the total basal area of a plot. We designated the damage situation as lightly damaged forest for a damage ratio from 1-33%, moderately damaged forest for a ratio from 34-66%, and severely damaged forest for a ratio from 67-100%. The damage ratios for each plot are shown in Figure 2.

An example of a typical pine wilt disease damage forest area is shown in Photograph 1 (City of Shibata jurisdiction).

The pattern development method development coefficients are the most important data for pursuing this type of analysis.

The development coefficients for June 4, 1984 are shown in Figure 3, and the development coefficients for August 1, 1999 are shown in Figure 4. The development coefficients of vegetation in June 1984 were between 0.2-0.3, indicating mostly pine wilt disease damaged forests.

On the 1984 land cover classification map, the locations in pixels classified as conifers where the development coefficient Cv of vegetation increased by 0.05 or more (in the Case Study of the city of Niigata and Maki-cho) or 0.1 or more (in the Case Study of the city of Shibata and Toyoura-cho) in the 1999 images were assumed to be pine wilt disease damaged forests.

The results for the city of Niigata and Maki-cho are shown in Figure 5, and the results for Shibata and Toyoura-cho are shown in Figure 7. In Figure 6 we showed the locations where Cv increased by summing as pine wilt disease damaged forests. From these figures one can see that the damage in both Shibata and Toyoura-cho has been increasing annually.

This also shows the tendency for Cs to increase when damage first begins to appear, followed by a spreading number of locations centered on those areas where Cv has increased.

This is thought to result from a loss of top layer leaves when the pine wilt disease damage first begins to appear, followed by an increase in Cs from the effects of lower layer vegetation or the soil, and finally by an increase in Cv as broad-leaf trees invade the area. If the change in Cv is large, the damage level is high; that is, this indicates severely damaged forest area. Analysis of many damaged forest areas will be necessary in order to determine how to decide the threshold value of Cv.

Today it would be best to regard areas within the jurisdictions of Shibata and Toyoura-cho where Cv is 0.2 or higher as locations that are severely damaged.

From the standpoint of monitoring, on the other hand, it is necessary to know the damage tendency followed until damage becomes extreme damage.

In this Case Study, the change in Cv is thought to be 0.1-0.2. Such locations are considered to be lightly damaged.

By comparatively examining the results of samples taken from satellite data and the actual conditions in the field, it was possible to sample the pine wilt disease damaged locations (3).

On the damage map shown in Figure 7, the areas in light pink indicate locations where the Cv value changed to 0.1-0.2, and those areas in dark pink indicate locations where the Cv value changed to 0.2 or higher.

4. Results from using the analysis results

This analysis was conducted as part of joint research carried out with the Conservation Division in the Agriculture, Forestry and Fisheries Department of Niigata Prefecture. As shown in Figure 7, by utilizing past satellite data it is possible to understand not only present conditions but also the dynamic change in the amount of damaged area.

The most critical point is to understand whether the pine wilt disease is still spreading, or whether it has reached a halt. Being able to understand damaged areas from the increase in Cv value from images, as performed in the analysis reviewed here, makes its simple to understand changes in damage. This approach was also viewed positively by government authorities.

Issues to be addressed are the small number of target municipalities and the need to expand the scope of this research as widely as possible in the future, and the fact that during the initial stages when pine wilt disease is first occurring the damaged area is very small and the LANDSAT/TM resolution of 28.5 x 28.5m is not adequate. We therefore want to use higher resolution satellite data for future studies.

5. Sources

(1) Fujiwara N., Muramatsu K., Awa S., Hazumi T. and Ochiai Fumio: Pattern Expand Method for Satellite Data Analysis
Journal of the Remote Sensing Society of Japan, Vol. 16, No. 3, pp. 17-34, 1996

(2) Furuumi Shinobu., Hayashi Ayami, Shiono Y., Muramatsu K. and Fujiwara N.: Vegetation Change in Kansai District by Pattern Decomposition Method using Landsat/MSS and TM data
Journal of the Remote Sensing Society of Japan, Vol.17, No.4, pp.34-49, 1997

(3) Yui Ryuuta, Nakamura Maki, Tatsuhara Satoshi and Abe N.: Monitoring Pine Tree Health Using LANDSAT TM Data, Conference Papers from the 111th Japan Forest Society Academic Conference, pp.488-489, 2000