کد مقاله کد نشریه سال انتشار مقاله انگلیسی نسخه تمام متن
6543655 159212 2013 14 صفحه PDF دانلود رایگان
عنوان انگلیسی مقاله ISI
On the evaluation of competition indices - The problem of overlapping samples
ترجمه فارسی عنوان
در ارزیابی شاخص های رقابت - مشکل نمونه های همپوشانی
کلمات کلیدی
افتراق افقی لبه، موجودی جنگل ملی، اسکنر لیزری هواپیما، همبستگی فضایی، شاخص های رقابت،
موضوعات مرتبط
علوم زیستی و بیوفناوری علوم کشاورزی و بیولوژیک بوم شناسی، تکامل، رفتار و سامانه شناسی
چکیده انگلیسی
We discuss statistical concerns regarding evaluation of three types of individual tree competition indices (non-spatially, spatially explicit and based on airborne laser scanning), with special attention to the method of selection of competitors, and the spatial dependency and smoothing caused by overlapping samples of competitors. We quantify the effect of spatial autocorrelation on the effective sample size for various search methods, to reveal potential type I statistical error, for a sample of 557 plots of the Norwegian National Forest Inventory located in the Hedmark Country. Our results show that spatial autocorrelation mostly appears when competitors are selected within short search radii (3-4) m of the subject tree. However, when simultaneously accounting for the impact of spatial autocorrelation on the effective sample size between individual tree growth at breast height and competition, the effect appears to be neglect-able. This result is verified by testing if the change in the effective degrees of freedom in the Spearman rank correlation t-test for the Clifford et al. correction and a spatial bootstrap method, relative to the classical t-test effective degrees of freedom, are correlated with different measures of stand structure. This ratio showed no systematic variation across measures of plot micro and macro-scale variation like Loreyś mean height, the Gini-coefficient of tree basal area or volume per hectare. The conclusion seems indifferent to plot edge bias correction. A linear mixed model with spatial covariance structure confirmed that sample overlap does not cause serious spatial dependence. Moreover, a median based statistical test revealed a significant smoothing effect, with increasing search radii of competitors, which causes loss of variation. However, the smoothing does not decrease the ability of the competition indices to correlate with individual tree growth at breast height within search radii of 12 m, and thus it does not represent any problem for prediction.
ناشر
Database: Elsevier - ScienceDirect (ساینس دایرکت)
Journal: Forest Ecology and Management - Volume 310, 15 December 2013, Pages 120-133
نویسندگان
, , , ,