4 FIRE MONITORING EXPERIMENTS WITH THE
SYSTEM
So far the prototype of the fire detection system has been
tested and demonstrated in three phases. Only two NOAA
satellites were used in each experiment.
4.1 Test Phase in Summer 1993
In summer 1993, the system operated from 1 June to 20
August. More than 330 images were analyzed by the system
(Häme and Rauste 1995). Of these 330 images, the system
reported 66 images as containing a possible fire. Because
the automatic screening of false alarms was not in use in
1993, these 66 images were inspected visually. In four cases,
no other reason but a high-temperature target on the ground
could be identified for the hot spot. Two of the four detected
fires were outside Finland and therefore practically impossible
to verify. One of the Finnish fires was verified as a real forest
fire, where the total area burned was 30 ha. The NOAA scene
where the fire was detected was received 1-2 hours after the
start of the fire.
Another test was carried out using a limited set of NOAA
images acquired over Greece. The analysis of these scenes
was hampered by the afternoon heating of mountainous areas.
4.2 Demonstration Phase in Summer 1994
In the demonstration phase of the forest fire workstation in
summer 1994 (from 5 July 1994 to 15 August 1994), 105
NOAA AVHRR scenes were processed. Thirteen fires were
detected in the fully automatic processing of these scenes
(table 1, figure 3, left).
Table 1: Fires detected by the prototype system in 1994.
Acquired at Latitude Longitude Country
1 94-07-05 16:54 62°19’ 24°54’ Finland
2 94-07-06 15:00 62°31 25°32" Finland
3 94-07-14 17:00 56°54’ 24°14’ Latvia
4 94-07-14 13:21 60°06’ 30°56’. .. Russia
5 94-07-14 15:02 59°17 25°25... Estonia
6 94-07-15 13:09 60°44’ 29°53" Russia
1 94-07-15 13:09 58?46' 34?03' Russia
8 94-07-17 12:44 59°08’ 37?49' Russia
9 94-07-17 14:25 55221! 23?58' Lithuania
10 94-08-01 14:42 54?31' 20°56’ Russia
11 94-08-02 16:49 65?28' 29?11' Finland
12 94-08-03 14:17 62?04' 26°55’ Finland
13 94-08-06 17:02 55°22 24°15’ Lithuania
In table 1, the column "Acquired at” gives the date and time
(in UTC) of the NOAA scene where the fire was detected.
A verification was done for all fires that were located in Fin-
land, Estonia or Latvia. All 4 fires located in Finland (verified
by the Finnish Ministry of the Interior, Rescue Department)
were prescribed (man-made) burnings for forest improvement
purposes. One fire was in Estonia. This was verified by the
Estonian fire authorities. The fire was a real forest fire. One
of the detected fires was in Latvia. Also this was a real fire
(verified by the Latvian fire authorities).
4.3 Pilot Phase in Summer 1995
In 1995, the system was demonstrated in a pilot phase. Be-
tween 4 July and 8 September, the alert messages generated
by the fully automatic fire detection system were sent by tele-
fax to the Helsinki Regional Dispatching Centre. There the
location of the detected fire was used to redirect the fax to
the regional dispatching centre where the fire was located or
to the country where the fire was located. This redirected
fax included a request for verification of the fire.
In total 205 NOAA AVHRR scenes were processed between 4
July and 8 September. The system detected over 14 000 hot
spots. Only 85 of the hot spots were classified as real fires
(figure 3, right). Table 2 shows the fires detected in the area
where verification was possible. Data for the remaining fires
can be found in World Wide Web, URL:
http://www.vtt.fi/aut/ava/rs/proj/fire/fires95.html
Table 2: Fires detected by the prototype system in 1995 in
areas where verification is possible.
Acquired at Latitude Long. Country V
95-07-09 08:16 56.275 22°11" Latvia P
95-07-11 06:21 59°56’ 23°09' AHK/Tammisaari |
95-07-12 15:51 62°25" 27°40’ AHK/Varkaus P
95-07-13 17:10 60°36" 25°34’ AHK/Porvoo P
95-07-20 17:29 61°08" 32°25' Carelia/Russia -
95-07-23 15:12 60°37" 24°43" AHK/Riihimaki F
95-07-27 15:25 62°58" 26°34’ AHK/Kuopio P
95-07-28 16:43 63?04' 25?48' AHK/Aàünekoski P
95-07-29 17:13 62°19" 25°12" AHK/Jyvaskyla P
95-07-30 08:45 58°41’ 25°00’ Estonia 2
95-07-31 15:38 63°00" 28°24’ AHK/Kuopio P
95-08-02 16:35 61°24" 23?39' AHK/Tampere P
95-08-02 18:01 65°07" 29°44’ AHK/Kajaani P
95-08-02 18:01 7^061?22"*23929' "AHK /Tampere ^ "P
95-08-03 16:13 62°24" 24°54" AHK/Jamsa P
95-08-06 16:48 50?27':? 23?57':"Latyla p
95-08-16 16:30 61?24' 23?39' AHK/Tampere p
95-08-19 15:25 64?42' 24?31' AHK/Raahe |
95-08-23 15:38 59°17°21°50 "Estonia T
95-09-03 16:39 64?43' 24?31' AHK]/Raahe |
95-09-04 16:17 60°38" 26°49" AHK/Kotka ?
Key for the verification results (column V above):
P = prescribed burning
| = industrial site
- = verification pending
F = forest fire
2 = false alarm in an area where 2 days
earlier had been a forest fire
? = reason for hot spot not known
In table 2, the regional dispatching centre (AHK) is indicated
for fires in Finland. The verification results in column V of
table 2 are mainly based on the immediate feed-back from the
regional dispatching centres. These verification results were
augmented after the fire season in cases where no answer was
received and in cases where other problems were encountered.
In the screening of false alarms, the elimination of cases where
the imaging geometry was close to specular reflection was
the most efficient single criterion. A more detailed analysis
586
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
