Patterns in the yearly trajectory of standard deviation of NDVI over 25 years for forest, grasslands and croplands across ecological gradients in Alberta, Canada

Abstract

Interannual variability of vegetation represented by the temporal standard deviation (STD) of Normalized Difference Vegetation Index (NDVI), temporal STD, was calculated using a 25-year semi-monthly sequence of 8 km resolution NDVI images (Advanced Very High Resolution Radiometer (AVHRR)-Global Inventory Modeling and Mapping Studies (GIMMS) data). The seasonal trajectory of temporal STD versus date was graphed to identify land-cover (LC) and location-related patterns. AVHRR-derived LC data and ecosystem mapping represented by Natural Subregions (NSRs) provided geographical stratification for data extraction. Finer subdivision came from separately considering >80% of pure LC areas within each NSR. The research was carried out over the Province of Alberta, Canada, which spans 11° of latitude and includes non-coinciding gradients in day length, temperature, precipitation and elevation. Two patterns occur. The ‘spring-variable’ pattern shows higher temporal STD in mid-spring and early autumn, with decreased values in summer. Slightly elevated temporal STD in early April and late October is interpreted to be due to snowmelt and snowfall. Spring-variability is typical of forested LCs and of cooler and wetter areas, primarily related to vegetation response to temperature variability at a given date from one year to the next. The ‘summer-variable’ pattern shows a single midsummer temporal STD peak. This pattern is typical of dry areas where grassland and cropland LCs predominate. Where these two LCs occur in cooler areas, their temporal STD approaches the spring-variable pattern. Summer-variability typifies moisture-dependent ecosystems. The temporal resolution is sufficient to demonstrate that forests in the extreme cool and wet areas of the province have a later spring temporal STD peak. Alberta has already warmed over the 1960–1990 period; because of the dates involved, warming before the 1980s is unrecoverable from satellite data. The data used here provide a baseline of variability over a period as long as that used for calculating climate normals. It is the longest available spatially complete satellite record to date. In future, with longer image time series, it will be possible to observe any changes in the degree and pattern of variability for the same location/LC. This will allow testing if interannual variability of vegetation changes in its temporal pattern or increases in magnitude for a given place.

Acknowledgements

I thank Elisabeth Beaubien for detailed reading and discussion of this article, especially of its phenological aspects and implications. Thanks go to two anonymous reviewers for their insightful comments and sharp eyes.