Browsing by Subject "accuracy"
Now showing items 1-2 of 2
-
(2015)Field work is needed to obtain reliable estimates when forest inventories are carried out. Field measurements traditionally have been the main source of information for inventories. But nowadays, also remotely sensed data collected using active or passive sensors mounted on satellite and aerial platforms are used to help in the estimation of forest parameters. Although the use of remotely sensed data is of great help in forest inventories, field data still plays a very important role as reference data, for results calibration and accuracy assessments. Considering that time and budget required for field work are generally some of the main concerns in forest inventory planning, the development of faster, cheaper, simpler, more accurate or more reliable field inventory methods and tools is a topic of great interest. TrestimaTM is a forest inventory system based in the interpretation of images taken with a mobilephone. Its accuracy and efficiency in estimating forest parameters was studied using sample plots in Russian. A total of 156 field plots were measured. The forest parameters measured were: the plot basal area and sample trees’ diameters and heights. The data collected with Trestima was meant to replicate a typical relascope sample plot inventory (variable radius plot inventory). Measurements obtained using traditional tools were used as reference data. The data collected for the inventory included plots at forest stands with different structures: from young to mature stands; and mixed stands to stands dominated by different species (most often Norway Spruce, Picea abies, (L.) H. Karst). The plots’ basal areas ranged from 7 to 62 m2/ha, the tree diameters from 3 to 60 cm and the tree heights from 3 to 35 m. The time used to measure the plots with the Trestima and the reference methods were collected. The data for each forest variable and the time invested in taking the measurements were organized as paired samples and compared using the statistic estimators of bias and RMSE, as the paired Student's t-test. Compared to the reference measurements, Trestima underestimated the basal area with a bias of 1.2 m2/ha (3.7%), but the differences were not statistically significant. In mixed stands, Trestima overestimated spruce basal area (bias of 13.9%), but for spruce dominated stands underestimated it (bias of 4.9%). Trestima overestimated tree diameters with a root mean squared error (RMSE) from 5.5 to 7.9%, depending on the tree species. but underestimated tree heights with an average RMSE of 3.7m (17.5%). The Trestima sample plot measurements were done faster than with traditional tools. Trestima measurements were in average 1.6 minutes (14.8%) faster. The Trestima system provided results comparable to the reference method for all the measured forest parameters. The worse results were obtained for the measurement of the tree heights. The interpretation of the results for the basal area, indicated that the system could benefit from taking into consideration stand structure, especially for species specific estimations. Trestima provided faster measurements of the forest parameters. One important advantage, is that Trestima produces automatically geographically referenced data, which can be used during later analysis, for example, interpretation of remotely sensed data or forest planning.
-
(2019)In forest inventories, the field data is needed as a reference data, calibration and for assessing the accuracy. Gathering the field data needs resources, such as work forces, equipment and data management operations. This means that time and budget as well as quality must be carefully considered when National Forest Inventory activities are to be planned. Therefore, the development of cost efficient, simpler, safer and more accurate and reliable field data measurement methods and tools are topics of great interest. One of the field measurement variables is the upper stem diameter, which has been part of the NFI field data set until the 11th NFI. The measured upper stem diameter helps to produce more reliable tree stem volumes with 3-parameter Laasasenaho volume model. However, only if the quality of the measurements is on an adequate level. The upper stem diameter has been traditionally measured with parabolic caliper assembled in the top of the 5-meter aluminium rod. This equipment combination takes time to assemble and disassemble, it is not very compact to be carried around as well as it can be a health and safety issue during the thunderstorm. Therefore Criterion, laser-based dendrometer performance was further examined in this study as an optional measurement equipment for future upper stem diameter field measurements. The Criterion upper stem as well as dbh measurement precision was analysed based on the 326 sample tree measurements with Sonar, Caliper and Criterion. The standard error for Criterion was 17,26 mm in dbh measurements for all species and 10,36 mm in d6 measurements for all species. The reference standard errors from earlier studies were 2,70 mm for dbh with Steel Caliper and 7,00 mm for d6 with Caliper. When analysing the Criterion performance with reference to mean of the measurements, the standard error in dbh measuements for all species was 9,72 mm and for d6 measurements 7,07 mm. Furthermore, the accuracy and precision were analysed with Bland-Altman technique, which is a suitable method for comparing two measurements of the same variable when both have some errors. The Bland-Altman results supported the earlier findings. Within and between observer analysis showed that the impact in measurement accuracy or precision is not caused by the observers. The comparison of the sample tree volumes produced the results that the relative standard error was increased by 2,13 % for all species when 2-parameter volume model was compared with 3-parameter model. Likewise, the relative bias was increased by 1,53 %. In efficiency experiment the scenario where only one sample tree with measured d6 was chosen from each of the sample plot and dbh percentile of p70 was providing most accurate and precise sample plot volumes with RMSE of 3,92 m3/ha and bias -0.78 m3/ha. The results show that there is a real challenge to achieve reliable and accurate upper stem diameter measurements and therefore new measurement methods need to be further studied and analysed.
Now showing items 1-2 of 2