Limiting Factors. A. License shall be in the subject area currently being taught by the instructor.
Limiting Factors. Illumination angles, sensor viewing angles, species, stand density, and image quality can all affect the feasibility of identifying individual tree canopies for both automated and manual methods. Imagery must be of a high enough resolution for individual crowns to be visible; this usually means multiple pixels need to be able to capture a tree crown. Ideally, if a crown is 3 feet in diameter, a pixel of 1- foot (but preferably even higher to accommodate the mixed pixel issue) is needed, providing at least 9 pixels covering the crown. This is an ideal example where pixel mixing is not an issue. Differing image quality over an area can cause results to vary across a study site. There are specific factors that can influence the TDA, by the nature in which it functions. One of the most significant is how the tree basins [i.e. gaps in the canopy] are calculated. Much of this hinges on a brightness value, which means that shade can heavily influence the accuracy in choosing between Figure 24. Plot showing no trees detected due to poor illumination and crown shadowing. trees and basins. The most extreme occurrence of this was plot 2321, in which the algorithm detected no trees. Field data indicates that plot #2321 has 11 living tally trees and a fairly open canopy. But because there is so much shade present in the image, trees are not represented by pixel brightness, and therefore no trees were recognized by the algorithm. This could be a specific issue when trying to identify late successional forest types, however, sometimes texture inherent in the imagery due to these shadows is utilized for successional forest class identification (Xxxxxx & Xxxxxxxx, 2002). The TDA underestimates the number of trees in a plot by an average error of 9 trees per plot. This occurs because many trees in the image are masked by the crowns of others. Furthermore, 1-meter resolution is too course (not detailed enough) to capture some of the younger, smaller (less than 1 m crowns) trees within the inner plot, even if they were not covered by larger growth. Therefore, it might be necessary to adjust the variables of the algorithm depending on the age and structure of relative tree stands. Some of these results do not depend strictly on tree size, however. A few plots that were dominated by Pacific Silver Firs showed poor results of tree crown delineation, compared to a landscape mostly dominated by Xxxxxxx Firs and Western Hemlocks. Because of this, the threshold pixel values within t...
Limiting Factors. At this time, it is not known which DEM resolution, flow accumulation algorithm, or flow accumulation threshold produce the most accurate stream channel locations. Flow accumulation algorithms and digital culvert development can take a large amount of processing time, especially for large areas.
Limiting Factors. LIDAR is expensive per unit area. LIDAR data is not available statewide; it is only available in relatively small spatial extents where acquisitions have taken place. LIDAR point data requires a large amount of storage space and processing capability. All models presented in this report are specific to the LiDAR data and geographic location, thus the models cannot be and should not be extrapolated to other landscapes and watersheds without understanding that the errors and uncertainties will change and cannot be evaluated.
Limiting Factors. Development of the canopy height model, and measuring the diameters of the individual tree objects is very time consuming. This processing took several weeks for the Mashel watershed. The inclusion of the Canopy Height Model (CHM) radius and diameter information as possible metrics for the model did increase the accuracy, but not by a substantial amount. The additional processing time may not be worth the accuracy improvement. This will need to be determined based on available project budget and goals.
Limiting Factors. There are different ways to estimate canopy closure or canopy cover (which are not the same) in the field, but they can be time consuming and the estimates they produce don’t necessarily agree with one another or with LIDAR. None of these methods can be said to represent the precise field condition, thus while some are considered by researchers to have greater or lesser accuracy none can really serve as a check on the performance of the others.
Limiting Factors. Illumination angles, sensor viewing angles, species, stand density, and image quality can all affect the feasibility of identifying individual tree canopies for both automated and manual methods. Imagery must be of a high enough resolution for individual crowns to be visible. Differing image quality over an area can cause results to vary across a study site. Field estimations of percent canopy cover may have subjective biases because of the methodology used. Moreover, it may be possible to only estimate upper levels of tree canopy coverage using aerial/satellite imagery because of its inability to offer some of the vertical information capable in LiDAR.
Limiting Factors. Evaluating all plots together does reveal the need for model adjustment, else existing outliers cannot be accurately estimated. The highest margin of error was 62 in plot 2123, which was previously discussed under the Limiting Factors of the Crown Diameter section. Examining the results of all plots, our model’s predictive capability decreases substantially.
Limiting Factors. Developing the canopy height model and ITOs used in the abundance model is very time consuming, taking several weeks for the Mashel watershed. The methods used to develop these datasets are described in Appendix D - Individual Tree Segmentation.
Limiting Factors. The major limiting factor in this prediction again resides within the initial TDA. The majority of tree species collected from the field are Western Hemlocks and Xxxxxxx Firs, which contribute to much of the TDA’s accuracy. Deciduous trees are not captured within individual tree crown objects as accurately since their overall disproportion in number influenced the design of the algorithm much less. Still, this technique was chosen because it builds much of the provisional work for calculating tree species. A different approach that would circumvent the dependence on the TDA is to perform a similar spectral analyses but on the entire plot-level image instead of ITOs. This might inherit more complexities with grass, shrubs, and open ground, but these features could also be suppressed. Even in a rudimentary analysis the general proportions of deciduous and conifer presence might be detected fairly accurately, as current results suggest.
