Potential of Hyperion imagery for simulation of MODIS NDVI and AVHRR-consistent NDVI time series in a semi-arid region

Long time series of satellite remotely sensed Normalized Difference Vegetation Index (NDVI) as an indicator of vegetation greenness can be used in a broad range of research topics such as climate change, vegetation dynamics and desertification monitoring. The first phase of this study tests a practical method using Hyperion imagery onboard on Earth Observing-1 (EO1) spacecraft for simulation of MODerate resolution Imaging Spectroradiometer (MODIS) NDVI and prediction of Advanced Very High Resolution Radiometer (AVHRR) NDVI data series in a semi-arid region in New Mexico, USA in 2007. Bo-Cai (2000) introduced the main concepts of this method. He used Airborne Visible/Infra-Red Imaging Spectrometer (AVIRIS) dataset and an atmospheric radiative transfer model based on the look up table procedure for simulation of AVHRR-consistent NDVI. Here, Hyperion data and MODTRAN Organizer software (MODO) as an interface of the MODTRAN radiative transfer code for performing the atmospheric gaseous absorption corrections on the AVHRR NIR channel are instead applied. The second part, evaluates the consistency of the simulated MODIS NDVI and the predicted AVHRR NDVI time series (derived by the practical method) with time series of the real MODIS/Terra and MODIS/Aqua NDVI (16-day composites, 250 m) and the real AVHRR NDVI (biweekly, 1000 m) in two sites with different land covers (shrubland and grassland) located in the study area, in 2007. Before performing the consistency evaluation, both the simulated MODIS and predicted AVHRR NDVI data sets (with 30 m resolution) are scaled down to the resolution of the corresponding real data sets i.e., 250 m and 100 m, respectively. The results of applying Hyperion data and MODO software in the tested method demonstrate that the MODIS NDVI and AVHRR NDVI time series can be simulated at image level (with 30 m resolution) promisingly. For example, the results of employing one of the Hyperion images over the study area (on 19 July 2007) into the method show that the bias in the prediction of AVHRR NDVI is about 27% lower compared to the case that it is estimated by MODIS NDVI. In addition, the results of the consistency evaluation based on the correlation and linear regression analysis for the time series of simulated MODIS NDVI and predicted AVHRR NDVI in a shrubland site and a grassland site in 2007, indicate that (1) the simulated MODIS NDVI values are highly correlated (0.81-0.98) with the real MODIS NDVI values, with Terra more than Aqua platforms and the root mean square error (RMSE) of the simulated values is in the range of 0.13-0.20 ndvi unit and (2) the predicted AVHRR NDVI time series is correlated with the real AVHRR NDVI, in the shrubland site higher than in the grassland site (0.87 and 0.66, respectively) while RMSE in the grassland site (0.12-ndvi unit) is lower than that in the shrubland site (0.18-ndvi unit). However, two main limitations in the Hyperion data (disordered temporal resolution and not globally coverage) make it useless in the generating of a reliable NDVI time series by the tested method for large and long period applications. Optimistically, if some kind of hyper spectral data with high temporal resolution and global coverage in future are applied in this ii method, results will have the capability of integration with real AVHRR NDVI time series for making long time series for global application.