Spectral Characteristics of Forest Vegetation in Moderate Drought Condition Observed by Laboratory Measurements and Spaceborne Hyperspectral Data

Although there have been several studies on the spectral characteristics related to leaf water content, it remains unclear whether the spectral property of leaves can be extended to the canopy-level. In this study, we attempt to compare the spectral characteristics of forest vegetation in moderate drought condition observed by laboratory measurement and satellite hyperspectral image data. Spectral reflectance data were measured from detached pine needles and oak leaves in the laboratory with a spectroradiometer. Canopy reflectance spectra of the same species were collected from temperate forest stands with dense canopy conditions using EO-1 Hyperion imaging spectrometer data obtained during the moderate drought season in 2001, and then compared with those obtained in the normal precipitation season of 2002. The relationship between leaf-level spectral reflectance and leaf water content was the clearest at the shortwave infrared (SWIR) regions. However, the canopy-level spectral characteristics of forest stands did not quite correspond with the leaf-level reflectance spectra. Further, four water-related spectral indices (WI, NDWI, MSI, and NDII) developed mainly with leaf-level reflectance were not very effective to be used with the canopy-level reflectance in dense forest condition. Forest canopy spectra under moderate drought status may be more influenced by canopy foliage mass, rather than by canopy moisture level. Introduction Symptoms of forest canopy stress vary with the length and magnitude of a drought, here drought is a combination of meteorological and hydrological drought (McVicar and Jupp, 1998), resulting in a physiological response. Lack of water content in vegetation can be major limitation to primary productivity, and as it reflects a drier than normal ecosystem, environmental problems such as wild fires (Roberts et al., 2003; Marod et al., 2004) can occur at such times. Monitoring of canopy drought condition is crucial for predicting vulnerability to forest fire and diseases and for estimating changes in forest productivity as a result of climate changes (Tian et al., 1998). Whether remote sensing can detect vegetation drought stress in forests depends on the spectral characteristics of leaf, canopy, and stand levels. Spectral Characteristics of Forest Vegetation in Moderate Drought Condition Observed by Laboratory Measurements and Spaceborne Hyperspectral Data Kyu-Sung Lee, Min-Jung Kook, Jung-Il Shin, Sun-Hwa Kim, and Tae-Geun Kim To detect such drought-stressed forest vegetation, remote sensor data should provide proper sets of spectral, spatial, and temporal resolutions. Spectral characteristics of leaf water have attracted research attention in the remote sensing community (Thomas et al., 1971; Tucker, 1980; Ripple, 1986). It has been relatively well known that the leaf reflectance decreases with increasing leaf water content (Hunt and Rock, 1989; Aldakheel and Danson, 1997; Ceccato et al., 2001). In particular, the wavelengths between 1,400 nm and 2,500 nm (shortwave infrared: SWIR) spectrum are known to be very sensitive to leaf water content (Tucker, 1980; Danson et al., 1992). Several studies also found that leaf water stress affects spectral reflectance in visible and nearinfrared (NIR) wavelengths, which might be related to the change of leaf pigments and internal cell structure as a result of the leaf moisture stress (Knipling, 1970; Harris et al., 2005). Several ratio-based spectral vegetation indices have also been developed and tested for correlating with plant water content (Cohen, 1991; Stimson et al., 2005). Although forest canopy water content may be an indicator to assess drought stress, most studies have focused on the spectral characteristics at leaf-level, and few have investigated canopy-level moisture content, in particular for forests with dense canopy closure (Bowyer and Danson, 2004). Remote sensing studies of canopy moisture stress have been primarily conducted over agricultural crop and semiarid vegetation areas (e.g., Ustin and Roberts, 1998; Jackson et al., 2004, Van Niel et al., 2003). In relatively dense forest remotely sensed estimation of the seasonal drought stress may be limited by small changes in the canopy’s physical structure. In a closed canopy, the spectral resolution of multispectral data may not be sensitive enough to assess the canopy water content. Recent developments in the technology of hyperspectral sensor data have provided the capability to detect minute variation of vegetation spectra (Ustin et al., 2004). Hyperspectral sensing may be an alternative to overcome such limitations by providing very narrow spectral bands enabling the observation of particular spectral features PHOTOGRAMMETRIC ENGINEER ING & REMOTE SENS ING Oc t obe r 2007 1121 Inha University, Department of Geoinformatic Engineering, 253 Yonghyun-dong, Nam-gu, Incheon 402–751, South Korea ([email protected]). Photogrammetric Engineering & Remote Sensing Vol. 73, No. 10, October 2007, pp. 1121–1127. 0099-1112/07/7310–1121/$3.00/0 © 2007 American Society for Photogrammetry and Remote Sensing PMSRS-01.qxd 9/14/07 11:17 PM Page 1121