Methodologies and Quality 1

Surface albedo is one of the essential climate variables (ECV) and a key parameter for the energy balance of the Earth.  Albedo retrievals are usually performed utilizing only a single instrument or an instrument family. This limits the total amount of available data per time period per terrestrial scene. Combining multiplatform observations yields improvements in both the accuracy and temporal resolution of surface albedo retrievals. This study is carried out in the World Meteorological Organization (WMO) Sustained and coordinated processing of Environmental Satellite data for Climate Monitoring (SCOPE-CM) project SCM-02.  It is focused on polar optical imagers, whose strengths are in high data acquisition rates over the high latitudes of Earth, which have a key role in the climate change.

An established intercalibration approach of two satellite instruments is to use simultaneous nadir observations (SNO). As the goal is to find a method applicable to random polar orbiting satellites, one has to take into account the possibility of satellites that never observe the same place simultaneously, so that the use of SNOs impossible. Our solution is to derive the top of atmosphere (TOA) reflectance distributions of a large area covering the entire range of global TOA reflectance. At this phase we concentrate on bands for which the wavelength range is similar enough to produce essentially the same reflectance for the same target. Then the assumption is that for a large enough statistical sample the reflectance distributions (with the same sun and satellite angle configuration) should be equal, as the instruments are observing the same target.

As an example we start with two satellites, at first using two separate data sets of MODIS in order to test the intercalibration method. Then we apply the method to fitting MODIS and AVHRR TOA reflectance data.

When fitting the distribution means of the two independent TOA reflectance values of the same period, one has to first convert the data sets to the same resolution. Then one has to take into account the uncertainty of both data sets. When the instruments are identical, one can use standard orthogonal regression. However, when there is a marked difference between the instrument accuracy of the two data sets, one has to carry out the linear regression minimizing the deviation from the regression line in the horizontal and vertical directions according to the respective instrument uncertainties of the two data sets. In addition, one has to take into account that the uncertainty of the TOA reflectance distribution means decreases as the inverse square root of the number of points in the distribution. Finally, the obliquely viewed pixels have larger uncertainty than the nadir pixels, because the true pixel size is larger than the nominal, and ascending and descending pixels do not cover exactly the same area. This heteroskedastic character of the points of the linear regression is taken into account by using as individual weights for the points the inverse of the pointwise variance.

The intercalibration method is first presented using simulated data with known uncertainty statistics. Then the results of this intercalibration approach are shown for two independent MODIS data sets with similar and simulated dissimilar instrument accuracy. The effect of the heteroscedasticity of the data on the regression is demonstrated. The advantage in first using simulated data and MODIS vs. MODIS intercalibration is that it is known in advance that one should obtain a 1:1 relationship for the linear regression. Hence, the goodness of the proposed method can be assessed separately before applying it to real data (AVHRR vs. MODIS). Finally, sample MODIS and AVHRR red channel data sets are intercalibrated using the regression method developed.

Accurate spectral calibration of imaging spectrometers is essential for the scientific exploitation of radiance measurements of atmosphere and land surface. In particular the retrieval of chlorophyll fluorescence and atmospheric trace gases concentrations requires highly accurate knowledge of center wavelength positions (CW) and instrument slit function parameters (SFP).

 We present a new algorithm for in-flight spectral calibration of imaging spectrometers with high spectral resolution and wide spectral range in the visible and near-infrared (VNIR). Our algorithm determines CWs and SFPs by fitting a forward model to the measured radiance spectrum. This forward model is based on the Lambert-Beer law and includes a high-resolution solar reference spectrum (Δλ=0.01 nm) with its sharp Fraunhofer lines, atmospheric absorbers (NO₂, O₄, O₂ and H₂O), and a cubic Hermite spline (C-spline). The spline approximates surface reflectance and atmospheric scattering, which vary only slowly with wavelength. The calculated spectrum is convolved with the slit function and compared to the measured spectrum to compute the fit parameters. CWs and SFPs can be described either by scaling their sensitivities to other environment parameters (e.g., temperature and pressure) or alternatively a more flexible C-spline. The CWs, SFPs and all other model parameters are determined by finding the maximum a posteriori solution of the least square problem.

 We tested our algorithm using simulated Airborne Prism Experiment (APEX) spectra (375-1015 nm) for different surface reflectances (“green grass”, “asphalt paving” and “metal roofing”). A Monte-Carlo simulation was conducted using APEX radiance noise specifications (signal-to-noise ratio of 625 for 50% surface reflectance) and a spectral shift corresponding to a pressure change of 7.5 hPa. The wavelength shift was modelled by a C-spline with a knot distance of 50 wavelength bands. We find that our method can be used to retrieve CWs with high accuracy. The performance is best for the bright surface “metal roofing” (< 0.01 nm at 500 nm) while for “green grass” and “asphalt paving” the accuracy is still better than 0.03 nm at 500 nm. The accuracy can be further improved by reducing the noise for instance by applying a spatial binning of several ground pixels.

 We conclude that our new approach can improve the in-flight spectral calibration of APEX and other imaging spectrometers. The accuracy is significantly better than earlier proposed approaches and better suited for the retrieval of e.g., chlorophyll fluorescence and atmospheric trace gases. The new approach is currently applied to APEX spectra and will be used to improve the APEX Level 1 product.

We present our generic model that can be employed to describe and correct for degradation of (scan) mirrors and diffusers in satellite instruments that suffer from changing optical Ultraviolet to visible properties during their operational lifetime.  Our hypothesis is that mirrors in flight suffer from the deposition of a thin absorbing layer of contaminant, which slowly builds up over time. Demonstration of application to different instruments will be shown, e.g. SCIAMACHY on ENVISAT and GOME on ERS-2.

The importance of measuring water vapor is obvious by its influence on e.g. the water cycle, the radiation budget and atmospheric chemistry. The microwave radiometer (MWR) on board of ESA’s ERS1, ERS2 and ENVISAT as well as the GOME (Global Ozone Monitoring Experiment) instrument on board the ERS-2 satellite allow for the retrieval of total column water vapor (TCWV). To verify those measurements, the NCAR GNSS (Wang et al., 2007) data base is used.
 
The NCAR GNSS data base is a 2-hourly data set measured by the ground-based Global Positioning System (GPS) using the zenith path delay between the transmitting GPS satellite and the receiving ground stations for the calculation of TCWV. It includes data since 1995 and is still ongoing, which is why this data base is predestined for an evaluation of measurements from ERS 2 and ENVISAT satellites, whose time series are starting October 1995.  
 
After a short introduction of the NCAR GNSS data base and pre-processing steps, the diurnal cycle of TCWV on global scales is described. The analysis is extended by describing the sampling bias if NCAR GNSS data area subsampled to approximate satellite overpasses. The resulting behavior of bias and standard deviation are analyzed. Mean, amplitude and maximum deviation as well as the local times of minimum and maximum of the diurnal cycle of TCWV were investigated with respect to their influence on bias and standard deviation. All investigations were done using the whole climatology, as well as separating them into seasonal and zonal studies. Since the altitude of the receiving station is additionally influencing the results, an additional study considers only stations lower than 500 meters.  
 
Preliminary results show that the bias is almost independent from latitude. Instead the standard deviation correlates with the amplitude of the diurnal cycle and is a function of latitude.  
 
Finally first results from the evaluation of the GOME Evolution and EMiR MWR products are shown.

The TROPOspheric Monitoring Instrument (TROPOMI) is the satellite instrument on board of the Copernicus Sentinel-5 Precursor satellite. The Sentinel-5 Precursor (S5p) is the first of the atmospheric composition Sentinels, to be launched in 2016 for a mission of seven years. Tropomi is a spaceborne nadir viewing spectrometer with bands in the ultraviolet and visible (270 to 495 nm), the near infrared (675 to 775 nm) and the shortwave infrared (2305 to 2385 nm). 

The Instrument Spectral Response Function (ISRF) of a pixel describes its (normalized) signal as a function of wavelength. An accurate ISRF is needed in the retrieval of Methane and CO.

The instrument spectral response of nearly all individual detector pixels of the SWIR spectrometer has been measured in a way not earlier done for Earth-observing instruments. During the on-ground calibration campaign at CSL (Liege, Belgium) measurements for both radiance and irradiance geometry were taken while scanning a tuneable SWIR laser over the full TROPOMI-SWIR spectral band.  From these measurements the ISRF has been mathematically modelled and calibration key data (CKD) have been generated. Various algorithm steps have been included to exclude effects of imperfect behaviour of the scan laser.

In this paper we will present measurement methods, the data analysis, results and CKD, as well as the validation of the results.