Continued Evolution of the ESA DUE GlobSnow Data Records of Snow Extent and Snow Water Equivalent
Luojus, Kari¹; Pulliainen, Jouni¹; Metsämäki, Sari²; Solberg, Rune³; Nagler, Thomas⁴; Wiesmann, Andreas⁵; Derksen, Chris⁶; Malnes, Eirik⁷; Wunderle, Stefan⁸; Hüsler, Fabia⁸; Foppa, Nando⁹; Schoener, Wolfgang¹⁰; Pinnock, Simon^11
1Finnish Meteorological Institute, FINLAND; ²Finnish Environment Institute, FINLAND; ³Norwegian Computing Centre, NORWAY; ⁴ENVEO IT GmbH, AUSTRIA; ⁵GAMMA RS AG, SWITZERLAND; ⁶Environment Canada, CANADA; ⁷Norut, NORWAY; ⁸University of Bern, SWITZERLAND; ⁹Meteoswiss, SWITZERLAND; ¹⁰ZAMG, AUSTRIA; ¹¹ESA/ESRIN, ITALY

The efforts of the European Space Agency (ESA) Data User Element (DUE) funded GlobSnow project has resulted in two hemispherical-scale records of snow parameters intended for climate research purposes. The datasets contain satellite-retrieved information on snow extent (SE) and snow water equivalent (SWE) extending more than 15 and 30 years respectively. The dataset on snow extent is based on optical data of Envisat AATSR and ERS-2 ATSR-2 sensors covering Northern Hemisphere between years 1995 to 2012. The record on snow water equivalent is produced using a combination of passive microwave radiometer and ground-based weather station data.
The SE processing system applies optical measurements in the visual-to-thermal part of the electromagnetic spectrum acquired by the ERS-2 sensor ATSR-2 and the Envisat sensor AATSR, covering Northern Hemisphere in a latitude-longitude grid with a 0.01 degree spatial resolution (approximately 1 km).
The SWE retrieval is based on the time-series of measurements by two different space-borne passive radiometers (SMMR and SSM/I) combined with ground-based weather station observations. The time series spans from 1980 to present day at a spatial resolution of approximately 25 km.
The GlobSnow-2 project, initiated in May 2012, is a direct continuation to the GlobSnow-1 project that was active from 2008 to 2012. The objective of the GlobSnow-2 project is further enhancement of the retrieval methodologies for SE and SWE products and a re-processing of the long term datasets utilizing the improved retrieval algorithms of GlobSnow-2. In addition to the further development of methodologies for the legacy sensor families of GlobSnow-1, the consortium will investigate the utilization of AVHRR and NPP Suomi VIIRS data as gap fillers before the launch of the Sentinel-3 SLSTR-sensor. Also the development of a new product combining the high resolution SE data with the lower resolution SWE product will be a topic for GlobSnow-2.
The current GlobSnow SE and SWE data records have been released and are available through the GlobSnow web-pages (www.globsnow.info). The ESA GlobSnow project was initiated in November 2008, and is being coordinated by the Finnish Meteorological Institute (FMI). Other project partners involved are NR (Norwegian Computing Centre), ENVEO IT GmbH, Central Institute for Meteorology and Geodynamics Austria (ZAMG), GAMMA Remote Sensing AG, University of Bern, MeteoSwiss, Finnish Environment Institute (SYKE), Environment Canada (EC) and Northern Research Institute (Norut).

- Enhancing the SWE retrieval methodology
The efforts on-going within the GlobSnow-2 project include improvement of the current SWE retrieval methodology [1] and an ultimate reprocessing of the full product dataset, covering the years 1979 to 2012. The main efforts for the SWE algorithm development include 1) the implementation of an improved multi-layer HUT snow emission model, 2) implementation of climatologically based spatially and temporally varying snow density values and 3) homogenization of the weather station input data.
In order to improve the SWE retrieval over heterogeneous landscapes and in regions with significant lake coverage, an improved version of the HUT multi-layer snow emission model will be incorporated into the existing SWE retrieval methodology. Improvements will result from the forward snow emission modelling component of the SWE retrieval scheme through snow grain size estimates that are physically more reasonable than the current estimates of the effective snow grain size.
The SWE retrieval methodology from GlobSnow-1 utilizes a fixed snow density value to convert snow depth to SWE within the retrieval scheme. A significant improvement to the retrieval accuracy may be obtained by treating the snow density values within the algorithm with spatially and temporally varying values based on climatology (temporal and spatial variability according to land cover and climate zone).
An important issue affecting the overall quality of the long term SWE data record is the snow depth observation input data from the global synoptic weather station network. These data are used for two purposes: (a) to derive a background field of snow depth by applying kriging interpolation and (b) to derive the value of snow grain size at the locations of the weather stations through forward snow emission model simulations. Any issues with the weather station data therefore directly affect the SWE time series. Efforts for detecting and removing of erroneous input data and spatio-temporal filtering of the weather station data will be investigated within the GlobSnow-2 project.
Efforts for improving the SWE retrieval will be assessed by producing and evaluating a prototype SWE data set that extends 10 years for non-alpine areas of the northern hemisphere. After the evaluation of the enhancements the full long-term SWE time series will be reprocessed using the new improved methodology.

- Enhancing the SE retrieval methodology
After the end of the Envisat mission (in 2012) and the end of AATSR data delivery the consortium decided to start producing the GlobSnow near-real time (NRT) snow extent product by utilizing the data from the NPP Suomi VIIRS sensor, with Metop-AVHRR as a backup data source. The operational production of NRT SE maps was initiated in October 2012, currently covering the pan-European domain. There are two different production chains, one utilizing the NPP Suomi VIIRS data and another relying on Metop-AVHRR data. The production chains are independent and provide daily fractional snow cover extent for the pan-European domain. NPP Suomi VIIRS production chain will be expanded to cover the whole Northern Hemisphere during the winter 2012-2013 and global coverage is planned for the winter 2013-2014.
In addition to migration of the NRT SE retrieval to new sensors, efforts are on-going to improve the retrieval algorithms. Efforts are on-going to improve the transmissivity map utilized as auxiliary information in retrieval, utilization of spatially varying bare-ground reflectance information and possible utilization of additional sensors, such as SPOT VGT for retrieval of hemispherical scale snow extent information.

- Rerefences
[1] Takala, M., Luojus, K., Pulliainen, J., Derksen, C., Lemmetyinen, J., Kärnä, J.-P, Koskinen, J., Bojkov, B., "Estimating northern hemisphere snow water equivalent for climate research through assimilation of space-borne radiometer data and ground-based measurements", Remote Sensing of Environment, Vol. 115, Issue 12, 15 December 2011, Pages 3517-3529, ISSN 0034-4257, DOI: 10.1016/j.rse.2011.08.014.