In v0.5.5 and previous versions, if timeseries monitoring in CCDCesque ended on an observation that was screened as noise in the monitoring period (using remove_noise=True), the last record's end attribute would be recorded as the date of this screened observation.
Nobody seems to have noticed before because it's only a 8-16 day offset when using Landsat data and CCDCesque models end on the last observation before the consecutive check so one more observation behind this number wasn't noticeable. I only found it when trying to validate some performance optimizations occuring within v0.6.0 (see #70).
This bug was resolved in the optimization implementations (per #70), specifically in 31e21f905adf9dce0b521a588b2cbf488dbbec24
In
v0.5.5and previous versions, if timeseries monitoring inCCDCesqueended on an observation that was screened as noise in the monitoring period (usingremove_noise=True), the last record'sendattribute would be recorded as the date of this screened observation.Nobody seems to have noticed before because it's only a 8-16 day offset when using Landsat data and
CCDCesquemodels end on the last observation before theconsecutivecheck so one more observation behind this number wasn't noticeable. I only found it when trying to validate some performance optimizations occuring withinv0.6.0(see #70).This bug was resolved in the optimization implementations (per #70), specifically in 31e21f905adf9dce0b521a588b2cbf488dbbec24