Mapping of the Growing Season on Svalbard Based on Satellite Data for the 1985 to 2009 Period

The aim of this study is to map trends, extreme years, and mean date in onset, end, and length of the growing season on Svalbard for the 1985-2009 period.

Phenological observations along tracks close to the village Longyearbyen and close to the research station Ny Ålesund has been established. These tracks were designed to detect the phenological variation at a scale that can be compared with medium resolution satellite data, as they capture the spatial and temporal variation in onset and end dates covering an area of about 15 km2 and 2 km2, respectively. The onset and end of the growing season is defined as flowering catkins and colouring of leaves of the small willow Salix polaris, respectively.

To map the growing season for the 2000 to 2009 period we used Terra MODIS satellite data with 500 meters spatial resolution and 8-days temporal resolution. To map the onset of the growing season we used a pixel-specific threshold method on Normalized Difference Vegetation Index (NDVI) values. At autumn the use of NDVI values showed less correspondences with phenological field observations. Hence, based on analyses of spectral properties of different vegetation types during the growing season we developed a "Normalized Different Autumn Index (NDAI)". NDAI is based on combinations of a band in the visible part with a band in the short-wave infrared part. This improved the mapping of the end of the growing season.

The results based on MODIS data show large variations in onset of growing season from year-to-year during the 2000 to 2009 period. In some fjords, and on the northeast coast, the onset of the growing season clearly corresponds with the timing of the melting of the sea-ice. At autumn several years could not be mapped due to too long periods with clouds or fog, and for the remaining years the results show only small variations among the years.

To prolong the time period we also processed more than 7000 NOAA-AVHRR satellite images for the 1985-2009 period. All of the images are from the period May 1 to September 30, and the highest NDVI value from twice-monthly composite periods were retained. This results in fewer, but more reliable NDVI values, representing the time series. Due to the 1 km spatial resolution of the dataset, only the warmest and most vegetated areas have been analyzed. The preliminary result shows a linear trend of much higher NDVI values in the second half of June, which could indicate much earlier onset of the growing season.