Impact of improved calibration of a NEO HySpex VNIR-1600 sensor on remote sensing of water depth

This paper investigates at the example of bathymetry how much an application can profit from comprehensive characterisations required for an improved calibration of data from a state-of-the-art commercial hyperspectral sensor. A NEO HySpex VNIR-1600 sensor is used for this study, and the improvements are based on measurements of sensor properties not covered by the manufacturer, in particular detector nonlinearity and stray light. This additional knowledge about the instrument is used to implement corrections for nonlinearity, stray light, spectral smile distortion and nonuniform spectral bandwidth and to base the radiometric calibration on a SI-traceable radiance standard. Bathymetry is retrieved from a data take from the lake Starnberg using WASI-2D. The results using the original and improved calibration procedures are compared to ground reference data, with an emphasis on the effect of stray light correction. For our instrument, stray light biases the detector response from 416 - 500 nm up to 8 % and from 700 - 760 nm up to 5 %. Stray light induced errors affect bathymetry mainly in water deeper than Secchi depth, while in shallower water the dominant error source is the calibration accuracy of the light source used for radiometric calibration. Stray light correction reduced the systematic error of water depth by 19 % from Secchi depth to three times Secchi depth, while the relative standard deviation remained stable at 5 %