Is it feasible to estimate LAI for high species-diversity forests with NDVI?

• Main Authors: Chung-Chun Chang, 張仲鈞
• Other Authors: 宋國彰
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/cgi-bin/gs32/gsweb.cgi/login?o=dnclcdr&s=id=%22107NCHU5080043%22.&searchmode=basic

碩士 === 國立中興大學 === 水土保持學系所 === 107 === Normalized difference vegetation index (NDVI) is widely used to estimate leaf area index (LAI) for the vegetation of large areas because numerous foreign studies have shown that NDVI is highly correlated with LAI. Nevertheless, most studied objects of these studies are low species-diversity forests. In recent years, the technology development of unmanned aerial vehicle (UAV) enables researchers to acquire real-time and high-resolution aerial images easily. The main purpose of the present study was to confirm whether NDVI generated with the UAV technique can be used to estimate LAI in forests with high species diversity. One of the two study sites was the 1.2-ha campus plantation (hereafter campus plot; 25°07’24” N, 121°40’26” E) located in National Chung Hsing University, in which there were more than 10 coniferous species. The other was the 25-ha Lienhuachih Forest Dynamics Plot (hereafter forest plot; 23°54’49” N, 120°52’43” E) in central Taiwan, which was a naturally-regenerated broad-leaved forest harboring 144 tree species. In a period of nearly one year, three image acquisition experiments were conducted for each of these two plots. In every experiment, aerial photographs were taken 80 m in the hereafter campus plot and 100 m in the hereafter forest plot above the plots with multi-spectral band cameras mounted on unmanned aerial vehicles (UAVs). These photographs were processed to NDVI orthographic images through reflectance calibration, image stitching, and image analysis. In the same weeks of aerial photograph acquisitions, canopy photographs were taken with the technique of hemispherical photography 1.3 m above the ground for every simple location in the plots (19 and 26 sample locations respectively for the campus and forest plots; the minimum distance between locations was 20 m). LAI for each location was obtained through analyzing these canopy photographs with a computer program. Results showed that LAIs of the three experiments in the same plots were highly correlated. Correlation analyses showed that there were significantly positive linear relationships between LAIs and NDVIs in three out of the six experiments (two experiments of the campus plot and one of the forest plot). The comparison between experimental conditions of these three experiments and those of the other three with poor LAI-NDVI correlations indicated that weather condition, the spectral band recorded by cameras, and the number of aerial photographs were not factors determined LAI-NDVI correlations. The present study confirmed that NDVI can be used to estimate LAI for high species-diversity forests, but not all NDVIs generated with the UAV technique are suitable for the LAI estimation. Further studies are needed to raise the suitableness of UAV-generated NDVI on the LAI estimation.