Identifiering av fel i skogliga beståndsregister med hjälp av satellitdata
This thesis has focused on the possibility to use a satellite image method developed by Metria Mijöanalys to identify stands that are wrongly registered in Holmen Skogs database. The method is only intended for identifying stand with poor descriptions in Holmen?s database that needs to be updated and not for updating the stand information.
The study area is located several km west of Hudiksvall on the borderline between Gävleborgs- och Västernorrlands län. The area is covered by a SPOT image from 2005. By which Metria Miljöanalys has made estimates of forest parameters for all stands in the area, and compared them with corresponding estimates according to the stand database. The stands in which the estimated and registered values differed more than a certain ?threshold? value was marked as deviant. The total number of stands in the studies area is 8941, of which 1473 were marked as deviant.
To evaluate Metrias method, a number of random chosen stands marked as deviant where measured in field. Prior to the field inventory, the populations were divided into two strata, in one the populations marked as deviant and in the other strata, the rest. In both strata, a number of random stands were chosen. Most of the chosen stands were marked as deviant.
In this thesis, the method of using satellite images to estimate variables, such as volume and distribution of tree species, has been evaluated. The results show that at the threshold value used by Metria Miljöanalys (2*Ssat), about 25 % off all included stands were marked as deviant, of which about 50 % were truly deviant which corresponding to about 25 % of the truly deviant stands. The results also show that the method estimates the volume better than other variables. About 65 % of the stands registered as deviant in volume, truly was deviant. The time needed to measure all the 1090 stands marked as deviant by Metria Miljöanalys in field is about 3300 h.
Depending on the intentions with the method, two ways of working with the method can be chosen. A low threshold value gives a high number of deviant stands. This also means that a high number of stands that isn?t truly deviant will be marked as deviant. On the opposite, a high threshold value will generate a lower number of stands marked as deviant. Of the stands marked as deviant, a higher number will be truly deviant. But this also means that a higher number of truly deviant stands won?t be found. This thesis also shows that the method should be used to find deviant population, regardless of which variable that differ.