this is for holding javascript data
Gustaf Granath added The_wildfire_started_on_the__.tex
over 8 years ago
Commit id: 0cbbdc18e780f34eff9c88229dc18097a5767278
deletions | additions
diff --git a/The_wildfire_started_on_the__.tex b/The_wildfire_started_on_the__.tex
new file mode 100644
index 0000000..151612a
--- /dev/null
+++ b/The_wildfire_started_on_the__.tex
...
The wildfire started on the 31 of July 2014 in the county of Västmanland, located in the central parts of Sweden. The forest fire lasted for 12 days and a total of 14 000 ha were consumed by the fire. During the initial days (31/7 – 3/8) the spreading of the fire were of moderate intensity but during the fourth of august the wind and fire intensity increased which drastically enlarged the fire affected area. Seven streams and ten lakes, larger than 1 ha, located inside the affected area and three lakes adjacent to the area were investigated. Two selected reference streams (Soltorpsbäcken and Säckenbäcken) were located in the same region as the fire affected area. One measured stream (Gärsjöbäcken) is part of the Swedish regional monitoring program (RMÖ) since 1995.
Sampling and chemical analysis
The first post-fire measurements of the streams were made on the 21 of August (2014). The fire affected area was during this time closed of and the measurements were therefore made outside the bounderies. The measurement location for Gärsjöbäcken was moved to the original place used in the regional monitoring program when the area was accessible. Three of the seven streams (Gärsjöbäcken, Vallsjöbäcken and Myckelmossbäcken) were measured every second to third week, except during the coldest period between 16th of December to the 27th of January. Synoptic measurements were made at 6 occasions during the eleven months after the fire. The lakes were sampled at one occasion on the 28th of October (2014). The water sampling procedure and water chemistry analysis were made according to the Swedish monitoring program (Fölster et al., 2014) using SWEDAC accredited methods at the geochemical laboratory at the Department of Aquatic Sciences and Assessment at the Swedish University of Agricultural Sciences (Sonesten, 2015). The chemical analysis of the regional monitoring program is presented in table X.
Catchment delineation
The catchments of the sampled streams were produced in ArcGIS 10.3, software from ESRI, using a national elevation model from Lantmäteriet (2015a) that had a resolution of 2x2 m and accuracy of 0.5 m. When rain hits the surface it will run in the steepest slope direction which is determined in the elevation model. By grouping the surfaces of the steepest slopes with the same direction watersheds were delineated.
Image analysis
Image analysis was made from three band infrared colored orthophotos that were photographed on the 16th of august, 2014, using an UltraCamEagle097 at 3700 meters (Lantmäteriet, 2015b). The orthophotos had a resolution of 0.25 m in the national projection SWEREF 99 TM. The image analysis was preformed in ArcGIS 10.3, software from ESRI. The ortophotos were analysed through supervised classification. Training samples were collected from uniform surfaces in four orthophotos. The training samples were sorted into classes that described different types of surfaces. To insure that different classes included different information the histogram, scatterplot and statistics were evaluated. Classes and selection of training samples were remade until the distributions in the histogram and scatterplots represented individual and non overlapping information. To ensure that the classes contributed with independent information low values of covariance were ensured. A signature file was created from the different classes of training sample classes. Nr ortofoton was classified using maximum likelihood classification.
The image classification was validated in the field at locations separate from the location of the four orthophotos used for the signature file. Location points with known image classification were observed in the field and coordinates of field observations were compared with the classification results. For each class in the image classification two or more location points were chosen which resulted in a total of 50 points.
The aim of the image analysis was to classify forest- and peatlands. In the fire affected area there were minor areas of roads, lakes, houses and farmland which were excluded from the image classification. Scorched peat and canopies had similar coloured surfaces and could not be separated in the classification. To separate peatland, drained forested petland and woodland a canopy height model (CHM) were produced. This was performed in LAStool, software from rapidlasso using laser scanned data from Lantmäteriet (2015c). The laser scanning is made at 1700 – 2300 m and the instrument can obtain up to four returns per pulse. The point density is approximately 0.5 -1 point per square meter and the accuracy is approximately 0.1 m for flat hard surfaces but can deteriorate in regions with steep terrain. The CHM were calculated from normalized laser data and noise reduced. A pixel size of 1.5 m was chosen due to the point density. Open peatland were extracted from the image classification by combining areas beneath one meter in the CHM with types of peat soil in soil maps from the Geological Survey of Sweden (SGU). The soil map was in scale 1:50 000 and 1:100 000 with a pixel size of 25 m. Drained forested peatlands were assumed to be equal to forested peat soils. To extract drained forested peatlands from the image classification areas of peat soil, according the map from SGU, were combined with areas with a canopy height of two meters or higher in the CHM.
In areas with high burn severity the canopy were consumed and more of the ground layer were present in the orthophoto. To distinguish between areas of severely burned canopies and pre-fire clear-cuts, shapefiles of announced clear-cuts from 1998 were observed. The shapefiles were evaluated with orthophotos, photographed before the fire in April and June, and the CHM to extract areas of clear-cuts made before the fire but with canopies lower than 2 m. The image classification of the four different types of land use was extracted to the watersheds of the measured streams.
