Browsing by Subject "Weibull-distribution"

Sort by: Order: Results:

Now showing items 1-2 of 2
  • Nykänen, Mikko (Helsingfors universitet, 2011)
    In Finland collation of data needed in forest planning is changing from traditional standwise field inventory to area-based airborne laser scanning and aerial photography inventorying. The aim of this study was to compare prediction accuracy of stand total volume and diameter distribution using following methods: MSN, PRM, ML, FMM and Weibull-distribution. Results were calculated separately for pine, spruce, birch and other tree species. In addition need of calculation time and storage space was established. Sampling plot and stand information used in this study were collected in the vicinity of Evo, Finland. Total of 249 sampling plots were measured. 12 clear-cutting areas measured by logging machine were used as reference data. In addition area-based laser scanning and aerial photographbased features were used in estimation of variables of interest. Results were calculated in all 12 stands and stands which area was over 0,5 hectares (8 pcs). Number of grid neighbours differed between 1 and 10. Depending on method and number of neighbours the relative RMSE and bias of stand total volume varied between 20,76 – 52,86 % and -12,04 – 46,54 % respectively in all stands and between 6,74 – 59,41 % and -8,04 – 49,59 % respectively in stands which area was over 0,5 hectares. Calculation time varied strongly depending on method and number of neighbours. With more developed programming and programs calculation times could decrease substantially. Storage space needed in saving information is not an issue in tested methods even in large-scale applications. According to diameter distribution PRM-method predicts narrow distribution if sampling plot consists of only few trees nearly same size. This affected results especially in PRM2.
  • Lyytikäinen, Satulotta (Helsingin yliopisto, 2019)
    A diameter distribution describes the size distribution of a forest and is used, for example, in forest planning. Stand characteristics can also be produced without a diameter distribution, but the diameter distribution allows utilization of tree-level models in the calculation of stand level variables and improves the estimation accuracy. In this thesis estimation methods for diameter distribution prediction were compared and the suitability of the methods and materials evaluated by using error indices. The two comparative diameter distribution estimation methods used in this study were the k most similar neighbor method (k-MSN) and the prediction of diameter distribution using the theoretical distribution model: Weibull-distribution. Both study methods utilised laser scanning-, field plot- and harvester measurement data collected from mature Norway spruce dominant forests in Southern Finland. When the k-MSN method is applied, the measured diameter distribution and stand characteristics of the nearest neighbor were imputed to the target grid of the stand. Stand characteristics were used for the prediction of Weibull-distribution parameters. Weibull parameters were estimated by modelling and by parameter recovery. Field and harvester plots were used as a reference data for the k-MSN. The suitability of the Weibull distribution for the description of the diameter distribution was evaluated by estimating the diameter distribution directly on the harvester data of the corresponding region. Maximum likelihood estimation and prediction of the Weibull-distribution were applied. As a result of the suitability testing, the Weibull distribution is seen as suitable for describing the diameter distribution of a mature, even aged, stand. With the maximum likelihood method and parameter recovery, the produced diameter distribution corresponded appropriately with the measured diameter distribution. The results of the k-MSN show that the imputation of diameter distribution gave the most accurate prediction of diameter distribution. Harvester or field plots gave nearly as accurate results and are both suitable for diameter distribution prediction