Limitations of Remote Sensing Surface Reflectance and Derived Data Products

The quality of RS data products can be limited by multiple factors including the spatial-temporal resolution of the input data, sub pixel clouds, and by the ‘local’ applicability of the artefact corrections (i.e. gap filling, smoothing, and curve fitting) in the phenology algorithm. In Australia, poor quality data due to cloud cover occurs more frequently at tropical latitudes near the coast during the monsoon season, at higher latitudes in winter, as well as in the west coast of Tasmania and high elevation all year round. Similarly, poor quality data also occurs with very high frequency in highly reflecting surfaces (e.g. salt crusts) in the interior of the Australian continent; in these areas, surface reflectance data (and its derivatives) are not reliable.


   In Australia, seasons only ‘fit’ within a calendar year in certain regions. Therefore, typically the first and last year in a time series of RS data will not capture complete seasons. This should be taken into account when computing derived metrics, particularly in phenology studies.


   MODIS Land Products are distributed by NASA’s Land Processes Distributed Active Archive Center (LPDAAC) as tiles from a global sinusoidal projection. Tiles are mosaiced, remapped, and reformatted using the MODIS Reprojecting Tool (MRT; Paget and King 2008). In TERN’s case, tiles are typically reprojected to a geographic latitude/longitude projection for the Australian extent. The mosaicing, remapping and reformatting processes do not alter either the data or the quality fields. However, not all tiles for a particular mosaic necessarily become available from the LPDAAC on the same day. It is therefore possible that the most recent mosaic for any given data product will be updated over several days, each time containing progressively more actual data within the same (standard) geographic extent. In addition, occasionally file corruption can prevent users to un-gzip the mosaic files.


   The Advanced Very-High-Resolution Radiometer (AVHRR) on-board NOAA’s satellites is calibrated before launch. Its calibration changes in orbit, typically with a rapid change immediately after launch and then drifting during their mission lifetime of several years. Moreover, the AVHRR carries no on-board calibration system for the reflective channels. To achieve calibration consistency over the length of time series measurements, the procedure developed for Australia by the Environmental Resources Information Network (ERIN) is typically used. This procedure assumes the reflectance stability of a set of Australian arid sites for the NOAA-14 satellite, and then it detrends the calibration of other satellites to match to NOAA-14. Therefore, time series of data (e.g. NDVI time series, see Table 1) can be compared in relative terms to other AVHRR measurements, but the data cannot be relied upon to make absolute statements about long-term trends.


REFERENCES