| Photosynthesis is a process that can harvest solar energy and transform it to biochemical energy. Through photosynthesis, the gross primary production (GPP) of terrestrial ecosystems constitutes the largest global land carbon flux and exhibits significant spatial and temporal variations. Due to its wide spatial coverage, remote sensing technology is shown to be useful for improving the estimation of GPP in combination with light use efficiency (LUE) models. Given accurate absorbed photo synthetically active radiation (APAR), LUE is a critical variable in the estimation of GPP, which is determined by a large number of factors constraining photosynthetic processes. Accurate estimation of LUE is essential for calculating GPP using remote sensing data and LUE models at regional and global scales. A promising method used for estimating LUE is the photochemical reflectance index (PRI= (R531- R570)/(R531+ R570), where R531 and R570 are reflectance at wavelengths 531 and 570 nm) through remote sensing. However, it has been documented that there are certain issues with PRI at the canopy scale. To utilize PRI in photo synthetically physiological studies, these issues need to be considered systematically.For this purpose, the following objectives are addressed in this thesis:1) To develop an automatic multi-angle hyperspectral observation system to continuously acquire PRI above a vegetation canopy over the hemisphere; 2) To investigate the ability of canopy-level PRI arithmetically averaged from multi-angle observations in tracking LUE variations throughout the growing season and external factors that influence this ability; 3) To develop a two-leaf PRI approach through differentiating the sunlit and shaded leaves based on a 4-scale geometric optical model; and 4) To evaluate this two-leaf approach for its effectiveness in reducing external effects on PRI. The major works and findings are summarized as follows:(1) An improved tower-based automatic canopy multi-angle hyperspectral observation system was established at the Qianyanzhou flux station in Jiangxi Province since January of 2013. In each 15-minute observation cycle, PRI was observed at four view zenith angles fixed at (37°,47°,57°) or (42°,52°,62°) in the azimuth angle range from 45° to 325° (defined from geodetic north). After data pre-processing, i.e. sensor calibration and dark currency correction, the spectral data then are used to compute canopy reflectance and PRI. The half-hourly canopy-level PRI is simply arithmetic averaged from all available multi-angle PRIobs when the view zenith angle is less than 63° within half an hour;(2) In this sub-tropical conifer forest, both half-hourly PRI and LUE exhibite detectable diurnal and seasonal variations, and decrease with increases of vapor pressure deficit (VPD), air temperature (Ta), and photosynthetically active radiation (PAR). Generally, PRI is able to capture diurnal and seasonal variations in LUE. However, correlations of PRI with LUE vary dramatically throughout the growing season. The correlation is the strongest (R2=0.6421, p< 0.001) in July and the poorest in May. Over the entire growing season, PRI relates better to LUE under clear or partially cloudy skies (clearness index, CI> 0.3) with moderate to high VPD (>20 hPa) and high temperatures (>31 ℃). Overall, PRI is most sensitive to variations in LUE under stressed conditions, and the sensitivity decreases as the growing conditions become favorable when atmosphere water vapor, temperature and soil moisture are near the optimum conditions;(3) To improve the ability of directional PRI observation to track canopy LUE, the canopy is treated as two-big leaves, i.e. sunlit and shaded leaves. On the basis of a geometrical optical model, the observed canopy reflectance for each view angle is separated to four components, i.e. sunlit and shaded leaves and sunlit and shaded backgrounds. To determine the fractions of these four components at each view angle, three models based on different theories are tested for simulating the fraction of sunlit leaves. Finally, a ratio of canopy reflectance to leaf reflectance is used to represent the fraction of sunlit leaves, and the fraction of shaded leaves is calculated with a four-scale geometrical optical model. Thus, sunlit and shaded PRI are estimated using the least squares regression with multi-angle observations. The retrieved PRI (PRIinv) calculated from the fractions of sunlit and shaded leaves and sunlit and shaded PRI can well capture variations of observed PRI at each view angle in every 15-minute observation period (>70%,p< 0.05, n= 5700);(4) In both the half-hourly and daily time steps, the canopy-level two-leaf PRI (PRIt) can effectively enhance (>60%) the correlation between PRI and LUE derived from the tower flux measurements over the big-leaf PRI taken as the arithmetic average of the multi-angle measurements in a given time interval. In the dry season from July to September, correlations of PRI with LUE at daily time steps are much stronger in the two-leaf case than in the big-leaf case. The correlation is the strongest (R2= 0.785, p< 0.001) in July. PRIt is very effective in detecting the low-moderate drought stress on LUE at half-hourly time steps, while ineffective in detecting severe atmospheric water and heat stresses, which is probably due to alternative radiative energy sink, i.e. photorespiratioa Overall, the two-leaf approach well overcomes some external effects (e.g. sun-target-view geometry) that interfere with PRI signals. |
PDF Full Text Request