Estimation of water content for short vegetation based on PROSAIL model and vegetation water indices

Abstract

As the dominant component by weight of live vegetation, vegetation moisture is one of the main factors determining plant photosynthesis, respiration, and biomass. Based on the type of remote sensed data used, retrieval algorithms for vegetation moisture retrieving algorithms fall into two categories, i.e., microwave-data-based methods and optical-data-based methods. However, microwave-based methods are always characterized by low spatial resolutions and often have difficulty in separating out vegetation and soil signals. On the contrary, because of the high spatial resolution and good sensitivity to green vegetation, optical remote sensing techniques have been the baseline method for estimating Vegetation Water Content (VWC) of short vegetation (i.e., Canopy Water Content, CWC). Here, we try to set up a universal, accurate and easy-to-apply way of retrieving CWC/VWC of short vegetation based on simulations from the PROSAIL model and generalized normalized Difference water index (NDWI), i.e., spectral indices taking the form of the NDWI formula. The new proposed method is based on PROSAIL model and four NDWI variants, i.e., NDWI(860, 970), NDWI(860, 1240), NDWI(860, 1640) and NDWI(1240, 1640). First, the parameter sensitivity analysis is carried out to determine their different influence mechanisms on the output reflectance and to optimize the PROSAIL model's input parameters. After that, canopy reflectance simulations are generated for short vegetation. According to the simulated reflectance, simulations of the four NDWI variants are derived, which were used to construct relationships with the simulated CWC and VWC of short vegetation. It is found that, instead of the linear relationship derived in previous studies, the simulated CWC/VWC is best approximated as an exponential function of NDWI. Following the analysis of the PROSAIL-generated results, a newly NDWI-based scheme is proposed for estimating CWC for short vegetation. Furthermore, VWC can also be estimated by combining the empirical relationship between VWC and CWC. Results derived from simulations show that the four NDWI variants are all linear related to ln(CWC), which were further used as CWC retrieving models. Moreover, the CWC retrieving models can also be used for VWC retrieving by combining the empirical relationship between VWC and CWC. Results derived from simulations also indicate that since NDWI(860, 1640) and NDWI(1240, 1640) are highly correlated (R2=0.99), both of the two variants can provide similar and relatively good CWC estimation accuracy. The validation results based on ground measurements show good consistency with simulated results, i.e., the VWC estimates from NDWI(860, 1640) and NDWI(1240, 1640) variants have high accuracy with both R2=0.88 and RMSE respectively of 0.4558 kg/m2 and 0.4380 kg/m2. The validation results based on Landsat 5 TM datasets also show that the R2 between CWC estimates and CWC ground measurements is 0.84, with a corresponding RMSE of 0.1342 kg/m2, while the RMSE between VWC estimates and VWC ground measurements is 0.5651 kg/m2. The proposed NDWI-based scheme for retrieving CWC/VWC of short vegetation is easy to implement and highly accurate. It can also be applied to agriculture for crop growth monitoring and drought indication. The estimation framework is also useful for CWC/VWC estimation of other short vegetation types. Moreover, since crop cover remains a challenging land cover for satellite-based soil moisture retrieval, this method can also be used to improve the quality of cropland vegetation information available as an ancillary input data for microwave-based soil moisture retrieval algorithms.