New Computing Environments: MicroComputers in Large-Scale Scientific Computing
New Computing Environments: MicroComputers in Large-Scale Scientific Computing
Quantitative Remote Sensing of Land Surfaces
Quantitative Remote Sensing of Land Surfaces
| Hi-index | 0.00 |
Inversion is an important process in remote sensing. To estimate the accuracy of retrieved parameters before inversion, we defined an Uncertainty and Sensitivity Ratio in the inversion of remote sensing models. It is a ratio of the priori uncertainty and sensitivity of one input parameter in the forward model to the sum of all parameters' uncertainty and sensitivity. USR was shown to reflect the variance of the retrieved parameter by error propagation theory. Subsequently, we took numeric experiments to illustrate the property of USR for typical vegetation cover scenes. Forty-eight multiangular datasets of Bidirectional Reflectance Factors were generated for these scenes by adding random noise to the Light Scattering by Arbitrarily Inclined Leaves model output. Seven parameters were retrieved for each scene. The results suggested that the Mean Square Error of the inversion results is highly correlated with USR. Based on this ratio, we selected two different angular observations in red and near-infrared (NIR) band to get a new Normalized Difference Vegetation Index (NDVI), which could produce less uncertainty to retrieve the Leaf Area Index (LAI) than using the conventional single-angular observations. The new NDVI was tested by retrieving LAI using the simulated multispectral and multiangular datasets. It was found that the inversion results of LAI are more accurate by using this new NDVI than using the traditional one.