About Us | Help Videos | Contact Us | Subscriptions

Crop Science Abstract - FORAGE & GRAZINGLANDS

Estimation of Forage Production of Nilegrass Using Vegetation Reflectance


This article in CS

  1. Vol. 47 No. 4, p. 1647-1651
    Received: Oct 25, 2006

    * Corresponding author(s): cmyang@wufeng.tari.gov.tw
Request Permissions

  1. Chwen-Ming Yang *a,
  2. Yuh-Jyuan Leea,
  3. Kuo-Yuan Hongb and
  4. Fu-Hsing Hsub
  1. a Crop Science Division, Taiwan Agricultural Research Institute, Wufeng, Taichung Hsien 41301, Taiwan ROC
    b Forage Crops Division, Taiwan Livestock Research Institute, Hsinhua, Tainan Hsien 71246, Taiwan ROC. This work was supported by the research grants (92AS-8.1.1-CI-C1, 93AS-8.1.1-CI- C1, and 94AS-8.2.1.-CI-C1) from Council of Agriculture, Executive Yuan, Taiwan ROC


Changes in reflectance spectrum of a crop are known to follow the morphological development of vegetation, and thus spectral models combining spectral characteristics correlated with biomass production may be used for yield estimation. Field experiments were conducted to validate use of reflectance spectra (350–2400 nm) to estimate forage production (i.e., aboveground fresh weight) of nilegrass (Acroceras macrum Stapf) vegetation from June 2002 to May 2004. Correlation coefficients (r) between spectral reflectance and forage production varied across the spectral range of measurements. A linear relationship (P < 0.010) was found for several wavebands, with the highest r value located at 891 nm (r = 0.671; P < 0.010). Of the examined spectral indices, forage production was found to be best correlated with RGREEN/RNIR ratio (R 2 = 0.654, P < 0.001) where RGREEN was reflectance of green light (490–560 nm) maximum and RNIR was reflectance of the near-infrared (740–1300 nm) peak. Assessment of forage production was further improved by using a multiple linear regression (MLR) model. The best five-variable linear regression equation provided the best fit (R 2 = 0.726, P < 0.001, Mallows' Cp criterion = 6.000). When validating the MLR model with other datasets from different growing seasons, the model gave reasonable prediction values (r = 0.833; P < 0.001) with a slope of 1.086 and root mean square error of 3.891 (N = 21).

  Please view the pdf by using the Full Text (PDF) link under 'View' to the left.

Copyright © 2007. Crop Science Society of AmericaCrop Science Society of America