Vegetation indices such as photochemical reflectance index (PRI) have been applied in tracing various biophysical and biochemical parameters within plants. However, less attention has been focused on chlorophyll (Chl) a/b ratio in arid environment. Quantitative estimation of Chl a/b ratio through remote sensing may provide important information regarding carbon balance and stress response in desert plants. We investigated the potential of PRI to trace Chl a/b ratio in Haloxylon ammodendron trees at Gurbantu ¨nggu ¨t Desert, China. The results revealed a seasonal and vertical variation in correlations. PRI was negatively correlated with Chl a/b ratio in all datasets, with pronounced sensitivity to changes in light across the canopy where low coefficient of determination (R 2 = 0.30) was obtained in bottom compared to the middle and top canopy assimilating branches. Seasonally, PRI-Chl a/b ratio relationship exhibited a higher R 2 for May-June dataset (R 2 = 0.62, P \ 0.001) and a lower RMSE of 0.18. A larger variability in both relationships was apparent when data from different heights were pooled together (total dataset). Although good correlation between seasonal PRI-Chl a/b relationship was found in top canopy branches, this relationship did not hold throughout the entire growing season. The identified normalized differential (ND) index (ND(510,570)) performed slightly better than the original PRI with R 2 improvement of 1 and 4 % for May-June and July-September datasets, respectively, and more than 5 % improvement in R 2 for within canopy datasets. In addition, both PRI and ND(510,570) performed poorly for pooled data. Therefore, within canopy variation and age-related changes must be considered when retrieving Chl a/b ratio through hyper-spectral remote sensing. Furthermore, both PRI and ND(510,570) must be applied with caution when dealing with desert species because of confounding factors related to strong erratic changes such as drought. Further studies on PRI mechanisms in different species under different environment are recommended.

No abstract