The purpose of this research is to understand the windfall damage caused to artificial conifer forests in northern Kyushu by typhoons No. 17 and 19 in 1991. In September 1991, Typhoons No. 17 and 19 struck the northern part of Kyushu with record-breaking violent winds. Typhoon No. 19 in particular set new records for maximum momentary wind speed, which was measured 60.9m/s at the summit of Mt. Aso, 52.9m/s in the city of Kumamoto and 44.4m/s in the city of Hita.

According to Forestry Agency, forest damage extended to 30 of Japan's prefectures and the total damage sustained by the country as a whole reached approximately 216.0 billion Yen.

Much of the damage was concentrated in Fukuoka, Saga, Kumamoto and Oita Prefectures. On private forest land in Oita Prefecture, which suffered the heaviest damage, total damaged land area was approximately 220km², or more than 10% of the prefecture's total forest area.

This numerical value, however, was compiled and published from reports from various municipalities. Better assessments require a more objective and uniform method for grasping the extent of such damage. For our research, we used the forests in northern Kyushu that are included in the images used.

To perform our analysis, we conducted a sample of damaged forests by sampling the changes in overlapping periods before and after the damage occurred.

The images used were MOS/MESSR (path 25, rows 73-74, dated April 15, 1991 and November 5, 1991). The number of clear weather days during 1991 was low, and the only images of the entire target region taken before and after the damage that have almost no clouds are MESSR images. Because we clarified from preliminary analysis that the forest damage appeared only in MESSR band 2 (red), we utilized band 2 level slice.

Because the target region is surrounded on three sides by the sea and the atmospheric conditions differ, we subdivided the image into small divisions and processed each division.

In addition, we used national land information elevation data (250m mesh) to understand the damage conditions suffered by the sampled damaged forest areas by slope azimuth (refer to the flow chart).

As a result of accuracy evaluations using a training area, the sampling accuracy for damaged forests was 85%.

Viewed by municipality, the percentage of damaged area was high in the towns of Honyabakei, Kusu and Amagase in Oita Prefecture and in the village of Hoshino and Katsuyama town in Fukuoka Prefecture.

Moreover, with regard to the relationship between slope azimuth and damage, we also clarified that regardless of the municipality the damage was concentrated on slopes facing south-southwest.

Although the forest crown was stripped by the high winds and fell to the ground, there was no noticeable decrease seen in the near-infrared radiation range on the satellite images taken immediately after the damage and the destruction of green leaves was not observed, because the crowns and trees lived for a considerable period of time.

We believe, however, that reflection of visible rays will increase from the change in the shape change of the forest canopy. Given this consideration, after sampling the change immediately after the damage by using visible light wavelengths it is necessary to allow a sufficient time to pass and then shift to a wavelength region selected according to the conditions of the damaged forest, such as sampling the change by using near-infrared radiation or vegetation indices.

For our actual analysis, we subdivided and processed the image in order to remove regional atmospheric effects.

Takao G.: Detection of the Windfall Damage to Forests Caused by the Typhoons 9117 and 9119, Proceedings of the 13th Asian Conference on Remote Sensing, pp. A-2-8-1 - A-2-8-6, 1992 Forestry and Forest Products Research Institute: Monthly News, No. 20, 1992