Reprocessed all MOD15A2GFS and MOD15A2PHN products
for years 2000 through 2005 to address the following issues...
MOD15A2GFS (Gap-Filled, smoothed LAI-fPAR):
- Scaled weights can be outside range of 16 bit integer. This causes
incorrect QC categories to be assigned, which can lead to misclassified
high quality and low quality data.
- After adjustment for second iteration TIMESAT run, small LAI values get
smaller weight and high LAI values get higher weight due to up-fold
weighting scheme. The weight is multiplied by 10000 and saved in short
integer (16-bit) format. The highest possible weight is 4, which is out
of valid range of 16-bit integer after scaling by 10000. Also, pixels
with weights smaller than 0.249 are not classified due to a bug in the
program. The pixels falling into these two situations are marked with
original LAI values in QC layer. Therefore, both high and low quality
data have a chance to slip into MODIS_LAI_FPAR_QC layer with original LAI
values. Four SDS layers are affected: COMPOSED_LAI, COMPOSED_FPAR,
1.Use a weight of 1000 instead of 10000. This avoids the highest weight
out of valid range of 16-bits integer after scaling.
2.Added trap to handle pixels with an adjusted weight between 0.01 to 0.249.
MOD15A2PHE (Phenology derived from Gap-Filled, smoothed
- Improve temporal resolution of phenology dates from 8 days to 1 day.
- In previous version, the greenup (browndown) date is marked at the
fitted point, which is closest to the time for which the left (right)
edge has increased (decreased) to 10% of the seasonal amplitude measured
from the left (right) minimum level. Therefore, the temporal resolution
of the phenology dates is 8-day.
- In current version, the greenup (browndown) date is calculate as a
fractional time of an 8-day period by a linear interpolation from
neighbor fitted points before and after 10% of the seasonal amplitude.
The interpolation parameters are calculated from the LAI value at 10%
of the seasonal amplitude and the LAI values from the neighbor fitted
points before and after it.