Effect of calibration set size on prediction at local scale of soil carbon by Vis-NIR spectroscopy

详细信息    查看全文

• 作者：Federica Lucà ; ^a ; ^{federica.luca@isafom.cnr.it} ; Massimo Conforti^a ; Annamaria Castrignanò ; ^a ; ^b ; Giorgio Matteucci^a ; Gabriele Buttafuoco^a
• 关键词：Soil spectroscopy ; Soil carbon ; Calibration set size ; Principal component regression ; Partial least square regression ; Support vector machine regression
• 刊名：Geoderma
• 出版年：2017
• 出版时间：15 February 2017
• 年：2017
• 卷：288
• 期：Complete
• 页码：175-183
• 全文大小：2297 K
• 卷排序：288

文摘

Calibration size effect on carbon prediction by soil spectroscopy and PCR, PLSR, SVMR Calibration size is reduced to 20%, 50% and 80% by PCR, SVMR and PLSR, respectively. PCR is less sensitive than PLSR and SVMR to calibration set size. SVMR outperforms PLSR and PCR with higher R2, RPIQ and lower RMSEP. The best combination between method and sample size is SVMR with 72 samples.