Operational Simulation of at Sensor Radiance Sensitivity Using the Modo / Modtran4 Environment

The MODTRAN radiative transfer code is a well established standard for simulating the at sensor radiance for optical instruments and imaging spectrometers from the UV to the thermal infrared. However, its efficient use is a task not easily accomplished. This situation has led to various developments for improving the efficiency as well as for the inversion of the model for atmospheric correction. The MODO user interface to MODTRAN is a tool for the forward modeling task which so far has been in use by various expert users. This paper describes workflows and examples for simulating the at sensor radiance for standard remote sensing situations. An interface design is proposed and implemented as part of MODO which improves the reliability of simulations for hyperspectral instruments. This end-to-end solution starts with inclusion and selection of surface reflectance functions from spectral libraries. Second, the atmospheric parameters most critical to the radiative transfer are to be defined and third, the components of the at sensor radiance shall be produced directly for specific sensor response functions. The selection of relevant parameters describing the situations is done on experience in various application area. The integration of the given principles leads to a comprehensive graphical user interface design which is proposed for setting up MODTRAN runs in an efficient manner. INTRODUCTION MODTRAN-4 has been established for many years as de-facto standard for simulating the at sensor radiance for imaging spectrometers (i, ii). It has been used for sensititvity analysis, sensor definition studies (iii) but also for the development of hyperspectral applications, and atmospheric correction procedures (iv, v, vi). The MODO user interface to MODTRAN is a tool which has been created to ease the use of the original radiative transfer code (vii). Based on the experience of the past years, an update of the tool in view of a more efficient sensitivity analysis has become critical. The capabilities of current and upcoming instruments for selected standard applications and measurement situations shall be simulated in an efficient manner. Specifically, the impact of surface reflectance signatures or of the spectral calibration on the at-sensor signatures is typically analysed. Furthermore, the comparison of simulated spectra to field measurements leads to a better understanding of the performance of an instrument. The most important tasks to be modeled are: sensitivity of signal to variations in response functions and spectral bandwidths/positions, surface reflectance signatures propagations (including the propagation of HDRF field measurements), intercomparison of signatures at various sensor systems for cross-calibration, support to vicarious calibration procedures by intercomparison of ground measurements to the respective sensor signal, creation of sensitivity series for retrieval of atmospheric gases and aerosols, and fast derivation of at-sensor radiance levels for adjustments of sensor gain and performance settings. © EARSeL and Warsaw University, Warsaw 2005. Proceedings of 4th EARSeL Workshop on Imaging Spectroscopy. New quality in environmental studies. Zagajewski B., Sobczak M., Wrzesień M., (eds) The related software design has lead to new workflows and interfaces as described hereafter. SIMULATION WORKFLOW AND IMPLEMENTATION The MODTRAN code as provided by the Air Force Geophysics Laboratory (AFGL) through ONTAR Inc. (http://www.ontar.com) is written in the FORTRAN computing language. It is handled by rigidly formatted ASCII input files. The handling of these files is very sensitive and requires experience with the code. This also bears the danger of introducing errors in at-sensor data simulations. MODO provides the graphical user interface for the creation of the input files as well as for the treatment of the outputs with respect to hyperspectral remote sensing. The updates now include streamlined workflows for the ‘daily work’. The whole workflow requires a library of sensor response functions which has compiled in collaboration with DLR, Munich. It contains most currently known sensor response functions, both as explicite functions or as gaussian approximation (as often used in high resolution spectrometers). The library lets you easily select a sensor of choice for direct simulation of the response-dependent at-sensor signal. At-Sensor Signals Simulation The core interface of the MODO procedure is a tape5 editor window. It allows to set most of the input parameters using pull-down menus instead of manually editing the rigidly formatted ASCII file. However, the various input options to MODTRAN may be misleading if a fast result of atsensor signals is to be calculated. Thus, a streamlined version of this window has been created. It uses four standard processing options, which allow the trade-off between processing accuracy and speed (the indicated approximative time is given for the radiance simulation of one hyperspectral standard situation on a 1.5 GHz machine). Low resolution: (4 seconds) High resolution: (1 minutes) High resolution with DISORT multiple scattering algorithm (5 minutes) High resolution with DISORT and correlated-k approach (3-4 hours not to be recommended). Despite the differences in speed, this four standard options exhibit significant differences of the simulated radiance values, specifically within or at the edges of atmospheric absorption features. A non-respresentative example is given in Figure1, where the deviations of the first three methods from the most accurate option is shown. Differences, inherent to the MODTRAN radiative transfer code are found which are at up to 5% in standard cases but may even be higher when strong absorption is present. Furthermore, the parameters most often used for simulations have been selected from the standard options. All cloud options have been omitted as they usually are not required – nor desired – for imaging spectrometry applications. The respective workflow from standard situations to atsensor radiance is depicted in Figure 2. It includes the extraction of the at-sensor radiance/irradiance or transmittance and a convolution to the selected sensor response function.