Mapping Snow Cover from Satellites with Application to Numerical Weather Prediction

Abstract

This study has developed a method for retrieving snow cover over the South Island of New Zealand from satellite imagery for input into mesoscale modelling. The high resolution meteorological satellite data of the third generation Advanced Very High Resolution Radiometer (AVHRR/3) are used for this purpose. The snow cover over the South Island is retrieved for 2001 and 2002. Snow cover effects on the surface air temperature and surface wind are studied by assimilating the retrieved snow cover information into a mesoscale model. A new novel automatic method which can navigate AVHRR images to high accuracy of 1 km has been developed. First, this method uses an orbital model to obtain a first order navigation for an AVHRR image. Then, it uses image matching techniques and a digital elevation model to automatically correct the error of this first order navigation. This method is tested over National Oceanic and Atmospheric Administration (NOAA) polar orbital satellite NOAA 12 AVHRR images for New Zealand. It is also suitable to use for other NOAA polar orbital satellites. An example of application of this method to NOAA 16 AVHRR images demonstrates its usability. The Moderate-Resolution Images Spectroradiometer (MODIS) global snow cover mapping algorithm is adapted for AVHRR/3. Multi-spectral thresholds for classifying AVHRR images are derived and snow maps of the South Island of New Zealand are generated for 2001 and 2002. Thos snow cover mapping method is validated by using visual techniques. The accuracy of this snow mapping is higher than 96% in the South Island. Relating the snow cover extent maps with major snow events also verifies the reliability of this snow mapping method. Snow cover information retrieved using the above algorithm are assimilated into the Pennsylvania University/National Center of Atmospheric Research (PSU/NCAR) Fifth-Generation Mesoscale Model (MM5). MM5 is a numerical weather prediction model designed for simulating/Predicting mesoscale and regional-scale atmospheric circulation. The model was run for the period of 1200 UTC 14 July 2001 to 1200 UTC 20 July 2001 and the period of 1200 UTC 18 June 2002 to 1200 UTC 23 June 2002. For each period, the model was run twice, with and without snow cover respectively. The comparison of the results of MM5 run with snow cover and without snow cover indicates that the near surface air temperature and the surface wind are significantly influenced by the presence of snow cover especially in low altitude areas. The surface air temperature at 2 metres height can have difference of up to 8°C between runs with and without snow cover. Not only wind speed, but also wind direction is greatly affected. Results from the high resolution (3 km) domain for 2002 show location of wind-shift lines moved about 6 to 12 km towards snow free area in the run with snow cover. These results show the importance of assimilating accurate snow cover into mesoscale modelling for New Zealand.