On the Synergic Use of Sentinel-1 and CoreH2O SAR Data for the Retrieval of Snow Water Equivalent on Land and Glaciers
Macelloni, Giovanni1; Brogioni, Marco1; Montomoli, Francesco1; Lemmetyinen, Juha2; Pulliainen, Jouni2; Rott, Helmut3; Voglmeier, Karl3; Hajnsek, Irena4; Scheiber, Rolf5; Rommen, Björn6
1IFAC - CNR, ITALY; 2FMI, FINLAND; 3ENVEO, AUSTRIA; 4DLR- ETH, GERMANY; 5DLR, GERMANY; 6ESA - ESTEC, NETHERLANDS
The potential of multi-frequency multi-polarization SAR observation for the study of the Earth's bio- and geo-physical variables has been proven in the past by using airborne and shuttle-based SAR systems. The use of a multi-frequency, multi-polarization and multi-angular approach can help in separating these mechanisms and relating SAR measurements to geo/bio-physical quantities. However, in spite of its widely recognized potential, wavelength diversity remains out of reach of the research community due to the lack of space-borne data that operate at a single frequency on a given platform. A possible improvement is expected by the Cold Regions Hydrology High-Resolution Observatory (CoReH2O) mission, selected for scientific and technical feasibility studies (Phase-A) within the Earth Explorer Programme of the European Space Agency, which combines a X- and Ku-band SAR operating on a single platform. The mission aims to close gaps in spatially detailed observations of key parameters of the global snow and ice masses for applications in climate research and hydrology. The identified primary geophysical parameters of the mission are the snow extent and snow water equivalent (SWE) and winter snow accumulation on glaciers.
While the CoReH2O mission is a candidate mission ESA is developing the Sentinel-1 European Radar Observatory, a polar-orbiting satellite system for the continuation of SAR operational applications. Sentinel-1 is a C-band imaging radar mission consisting of a pair of satellites aimed at providing an all-weather day-and-night supply of imagery for GMES user services. The first Sentinel-1 satellite is envisaged to launch in 2013 and will be followed by the second satellite within two years.
Whereas previous works were devoted to the development of retrieval algorithms for snow and land ice applications for both missions separately the possibility to combine data from these two missions have been recently studied in an ESA project called ALGOSNOW.
As a first result it was demonstrated that the C-band frequency range could provide useful information on soil status along the snow season which improves the SWE retrieval capabilities of CoReH2O. In fact the CoReH2O baseline retrieval algorithm relies on the snow background contribution to remain constant between a reference image in snow free or shallow snow conditions and the subsequent images used for retrieval of SWE. However, for some surface types the background signal may still evolve during the snow season due to incomplete freezing of soil. The use of a soil freeze-thaw detection algorithm based on C-band in the retrieval allows in principle to the separation of SAR images into frozen and thawed areas. The X-band channels may allow detection of soil freezing also directly from CoReH2O observations; however, the enhanced penetration depth of Sentinel-1 SAR can potentially provide more accurate segmentation, as well as fill out temporal gaps between CoReH2O acquisitions. The evolution of the freeze/thaw state of soil, derived from both instruments, can be considered in the segmentation of SWE retrieval. The historical database of soil state can be used to select the reference image for later SWE detection (ideally the reference image should be the first available image with stable soil conditions). Areas with thawing soil can be excluded from the retrieval, or at least flagged for additional uncertainty. Preliminary tests were performed using (1) real SAR data acquired over the Sodankylä test site in Finland and (2) using simulated data demonstrating the potential of the developed approach. The possible contribution of C-band for the retrieval of SWE in vegetated areas (in particular sparse forests) has also been demonstrated for this test site.
For glaciers, the contribution of C-band data could be useful to improve the CoReH2O retrieval capabilities. In fact, the sensitivity of different radar frequencies to measure snow accumulation for a specified period depends on the morphology of the snow/ice medium, which is related to the diagenetic glacier facies. These are distinct zones in the upper layers of glaciers and ice sheets exhibiting characteristic morphological features of the snow and ice volume determined by accumulation, ablation and metamorphism. The diagenetic facies on glaciers and ice sheets can be subdivided into: the dry-snow facies, the percolation facies, the superimposed ice zone, the glacier ice zone. In order to assess the feasibility of retrieving snow and ice physical properties from SAR data, it is important to consider the dominating scattering mechanisms in the different glacier zones.
The developed algorithm is based on the classification of glacier facies. The segmentation of the dry snow zone is based on the ratios [X-band/C-band] and [Ku-band/C-band], using SAR data of the winter season. For discriminating the percolation zone and glacier ice zone, summer and winter data are needed. The discrimination is based on multi-temporal thresholds. Multiple frequencies offer improvements for the segmentation due to reduction of noise (related to sensor and target characteristics). Once the glacier facies are identified the algorithm for retrieving winter snow accumulation in the percolation zone and glacier ice zone developed for CoReH2O is applied. The add-on of Sentinel-1 offers some improvement during the transition period from melting to dry snow (autumn) and in spring during the start of the melting period. In particular for CoReH2O Mission Phase 2 (15 day repeat observations) the improved time sequence is of interest to detect short-term melt events which may modify the metamorphic state of upper snow layers. For the dry zone the retrieval algorithm is based on empirical or semi-empirical relations between backscatter at the radar at different frequencies and accumulation measured in snow pits and shallow firn cores. Such algorithms have so far been based on single frequency backscatter data. Also in this case preliminary tests were carried out using simulated and real SAR data confirm the potential of this approach.
Within the project the specific error sources for the combined use of Sentinel-1 and CoReH2O data are discussed. Unlike data acquired by a single satellite platform during one overpass, the combined use of data from two or more platforms is subject to additional error sources. In this case absolute errors of one sensor add up (sometimes compensate) as relative errors too each other. Three possible errors sources are identified and discussed : co-registration, geo-coding and radiometric calibration respectively.