The ESA-ESRIN contract no. 21719/08/I-OL covered the activities performed between April 2008 and July 2014 by the L2 sub-group of the MIPAS Quality Working Group. The main objective of this project was to monitor the quality of MIPAS L2 products, to improve the performances of the L2 ESA processor, used for both the near real time analysis during ENVISAT mission and for the off-line analysis, and to enlarge the list of retrieved species. In particular, three different versions of the MIPAS operational processor were finalized: IPF V5, ML2PP V6 and ML2PP V7, leading to significant improvements in the quality of the products and in their diagnostics, to an increase of the number of processed scans, and to an enlargement of the list of provided products. With respect to IPF V4.1&2, used to process only the first two years of MIPAS measurements (2002-2004, Full Resolution (FR) measurements) and validated with respect to correlative measurements (Ridolfi et al., 2006, Cortesi et al., 2007, Wang et al., 2007, Payan et al., 2009, Wetzel et al., 2007, Wetzel et al., 2013, all contained in the MIPAS special issue), IPF V5: o performs the analysis of the full mission (both FR and Optimized Resolution (OR) measurements, characterized by a reduced spectral resolution but finer spatial resolution) after the introduction of an a posteriori regularization with self-adapting threshold aimed to reduce instabilities of the retrieval with measurements of the OR phase (due to sampling intervals finer than the instantaneous field of view of the instrument and the choice of using a retrieval grid aligned with the vertical sampling of the measurements); o provides as output Averaging Kernel (AK) for each scan. Products of ML2PP V6 are very similar to the ones of V5 in terms of bias, but are characterized by: o an improved diagnostics, since the Covariance Matrix and the AK matrix are computed taking into account all iterations performed during the retrieval (this is needed if the final Marquardt parameter is different from zero). Furthermore, horizontal AKs are provided for each latitude band and each season. o additional four new species, namely CFC-11, CFC-12, N2O5 and ClONO2. MIPAS ESA processor products V7 are characterized by: o reduced time dependent calibration error due to non-linearities (coming from using new L1 data, V7); o availability of retrieval altitude grid characterized by a very small bias after correction with ECMWF pressure/altitude relationship; o new microwindows for FR measurements with an increased information content and leading to a reduction of the detected bias in the FR measurements for CH4, N2O and CFC-12; o larger stability, given by the new approach for retrieving continuum (the continuum transmission is fitted instead of the continuum cross-section), and new regularization with altitude dependent regularization strength; o lower bias of CFC-11, due to proper handling of COCl2 interfering species in the retrieval o better representation of the measurements, proven by the fact that a smaller chi-square is obtained for most of the species; o better handling of diurnally varying interfering species, coming from the use of diurnally varying Initial Guess profiles; o products of five new species (HCFC-22, CFC-14, COF2, CCl4 and HCN) available. Concerning V7 products, preliminary verifications have been performed only on the diagnostic dataset V7, including more than 4000 orbits: the Non-Linearity correction performed in the L1 files introduces a drift in V7 profiles with respect to previous versions, that corrects for the instrumental drift present in previous datasets. Some problems have also been identified: V7 FR Temperature profiles in polar and midlatitude regions presents a negative bias of about 0.5-1K with respect to both V6 products and correlative measurements coming from the Non-Linearity correction. In the tropics in general V7 Temperature has a positive bias with respect to previous versions, for both FR and OR measurements, and this seems to come from the new approach for handling continuum. A new version of the spectroscopic database has been finalized, with improvements in H2O, HNO3, COCl2 and C3H8. The new version of the spectroscopic database will be used for the final reprocessing of the whole MIPAS mission in the frame of the subsequent contract. A large effort was spent for improving the quality of the MIPAS L2 products, but further improvements have been identified as needed to reduce some systematic errors of the products and to fully exploit the information contained in MIPAS measurements. One of the main problems identified so far is the error due to the horizontal inhomogeneities that are not currently taken into account by the algorithm. The quality of the retrieval is affected by the horizontal gradients but from the analysis of single scans no enough information is contained in the measurements to retrieve the gradients. Considering that the code will be used only for a reanalysis, the choice has been to take advantage of the information on horizontal gradients of both temperature and target species from previous reprocessing or ECMWF reanalysis and to modify the forward model in order to handle the externally provided gradients. Another important objective is the full exploitation of MIPAS measurements that implies the retrieval of weaker and weaker species, whose interference from other species cannot be disentangled. To this purpose the possibility of performing a multi-target retrieval is crucial. Furthermore, we have to consider that the vertical profile retrieval of weak species requires the use of some constraint for the attainment of a fast and useful product and it can be useful to have the possibility of tuning or removing it as needed during the average procedure. To this purpose Optimal Estimation seems to be more apt than the current approach where the constraint is determined with a self-consistent method, difficult to rebuild. For very weak species a posteriori averages of retrieved profiles after removing a priori constrain, as studied in the frame of this project, can be also considered. Validation of ML2PP V7 and further investigations aimed to solve these open issues will be completed in the frame of the subsequent contract.
Final Report of the ESA-ESRIN Contract no. 21719/08/I-OL Support to MIPAS Level 2 Product Validation The 'MIPAS Quality Working Group' project
P Raspollini;S Ceccherini;B Carli;M Ridolfi;E Castelli;B M Dinelli;L Sgheri
2014
Abstract
The ESA-ESRIN contract no. 21719/08/I-OL covered the activities performed between April 2008 and July 2014 by the L2 sub-group of the MIPAS Quality Working Group. The main objective of this project was to monitor the quality of MIPAS L2 products, to improve the performances of the L2 ESA processor, used for both the near real time analysis during ENVISAT mission and for the off-line analysis, and to enlarge the list of retrieved species. In particular, three different versions of the MIPAS operational processor were finalized: IPF V5, ML2PP V6 and ML2PP V7, leading to significant improvements in the quality of the products and in their diagnostics, to an increase of the number of processed scans, and to an enlargement of the list of provided products. With respect to IPF V4.1&2, used to process only the first two years of MIPAS measurements (2002-2004, Full Resolution (FR) measurements) and validated with respect to correlative measurements (Ridolfi et al., 2006, Cortesi et al., 2007, Wang et al., 2007, Payan et al., 2009, Wetzel et al., 2007, Wetzel et al., 2013, all contained in the MIPAS special issue), IPF V5: o performs the analysis of the full mission (both FR and Optimized Resolution (OR) measurements, characterized by a reduced spectral resolution but finer spatial resolution) after the introduction of an a posteriori regularization with self-adapting threshold aimed to reduce instabilities of the retrieval with measurements of the OR phase (due to sampling intervals finer than the instantaneous field of view of the instrument and the choice of using a retrieval grid aligned with the vertical sampling of the measurements); o provides as output Averaging Kernel (AK) for each scan. Products of ML2PP V6 are very similar to the ones of V5 in terms of bias, but are characterized by: o an improved diagnostics, since the Covariance Matrix and the AK matrix are computed taking into account all iterations performed during the retrieval (this is needed if the final Marquardt parameter is different from zero). Furthermore, horizontal AKs are provided for each latitude band and each season. o additional four new species, namely CFC-11, CFC-12, N2O5 and ClONO2. MIPAS ESA processor products V7 are characterized by: o reduced time dependent calibration error due to non-linearities (coming from using new L1 data, V7); o availability of retrieval altitude grid characterized by a very small bias after correction with ECMWF pressure/altitude relationship; o new microwindows for FR measurements with an increased information content and leading to a reduction of the detected bias in the FR measurements for CH4, N2O and CFC-12; o larger stability, given by the new approach for retrieving continuum (the continuum transmission is fitted instead of the continuum cross-section), and new regularization with altitude dependent regularization strength; o lower bias of CFC-11, due to proper handling of COCl2 interfering species in the retrieval o better representation of the measurements, proven by the fact that a smaller chi-square is obtained for most of the species; o better handling of diurnally varying interfering species, coming from the use of diurnally varying Initial Guess profiles; o products of five new species (HCFC-22, CFC-14, COF2, CCl4 and HCN) available. Concerning V7 products, preliminary verifications have been performed only on the diagnostic dataset V7, including more than 4000 orbits: the Non-Linearity correction performed in the L1 files introduces a drift in V7 profiles with respect to previous versions, that corrects for the instrumental drift present in previous datasets. Some problems have also been identified: V7 FR Temperature profiles in polar and midlatitude regions presents a negative bias of about 0.5-1K with respect to both V6 products and correlative measurements coming from the Non-Linearity correction. In the tropics in general V7 Temperature has a positive bias with respect to previous versions, for both FR and OR measurements, and this seems to come from the new approach for handling continuum. A new version of the spectroscopic database has been finalized, with improvements in H2O, HNO3, COCl2 and C3H8. The new version of the spectroscopic database will be used for the final reprocessing of the whole MIPAS mission in the frame of the subsequent contract. A large effort was spent for improving the quality of the MIPAS L2 products, but further improvements have been identified as needed to reduce some systematic errors of the products and to fully exploit the information contained in MIPAS measurements. One of the main problems identified so far is the error due to the horizontal inhomogeneities that are not currently taken into account by the algorithm. The quality of the retrieval is affected by the horizontal gradients but from the analysis of single scans no enough information is contained in the measurements to retrieve the gradients. Considering that the code will be used only for a reanalysis, the choice has been to take advantage of the information on horizontal gradients of both temperature and target species from previous reprocessing or ECMWF reanalysis and to modify the forward model in order to handle the externally provided gradients. Another important objective is the full exploitation of MIPAS measurements that implies the retrieval of weaker and weaker species, whose interference from other species cannot be disentangled. To this purpose the possibility of performing a multi-target retrieval is crucial. Furthermore, we have to consider that the vertical profile retrieval of weak species requires the use of some constraint for the attainment of a fast and useful product and it can be useful to have the possibility of tuning or removing it as needed during the average procedure. To this purpose Optimal Estimation seems to be more apt than the current approach where the constraint is determined with a self-consistent method, difficult to rebuild. For very weak species a posteriori averages of retrieved profiles after removing a priori constrain, as studied in the frame of this project, can be also considered. Validation of ML2PP V7 and further investigations aimed to solve these open issues will be completed in the frame of the subsequent contract.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
