Epic 1: Publishing short communication with correlation analysis of Sentinel 1, Sentinel 2 Indices ~ GEDI_LAI (:alarm_clock: 16.08.2020 computation complete and finished draft).
[ ] no spatial prediction only plot basis for error estimation
[ ] building solid base for further spatial/temporal predictions
[ ] Submit to: Remote Sensing of Environment - Short Communications (9.085)
[ ] Alternative: Environmental Research letters - Special GEDI Issue (6.096) but: might be too focused on applications)
Story 1 Dataset with LAI and Corine Information (:alarm_clock: until 19.7.2020)
[x] create sf with GEDI and Corine
[x] chosen area needs to
[x] have at least 500 points in each of the classes (12 - non-irrigated arable land, 18 - pastures, 23 - broad leaved forest, 24 - coniferous forest, 25 - mixed forest)
[x] fit within 1 Sentinel tile to avoid mosaiiking, etc.
Story 2 Extended dataset from Story 1 with Sentinel 1 and 2 data (:alarm_clock: until 31.7.2020)
[ ] choose sentinel 2 bands and indices
[ ] use Bae et al. (2019) for Sentinel 1 Preprocessing
[ ] output: gamma with 2 bands (vv, vh)
[ ] buffer and exclude cloud mask
[ ] extract values at GEDI loacations
[ ] use footprint of 25 m diameter with mean value
[ ] plotting for quality control
Story 3 Regression Analysis (:alarm_clock: until 7.8.202 0)
[ ] choose model (lm, gam, gamm, stepaic)
[ ] separated by: all variables, only Sentinel 1, only Sentinel 2
[ ] visualisation (t-value, r², slope, standard error)
[ ] separated by: landuse and variables
Story 4 Present (:alarm_clock: 16.8.2020)
[ ] Combine, interpret, conclude
[ ] smooth over writing
Backlog
predict for Germany and examine how to deal with missing GEDI data north of 56° (Spatial Upscale, temporal Downscale of GEDI-LAI with Sentinel 1 and 2 for Germany)
mosaiiking Sentinel 1&2 (might be a challenge)
possibly: matching GEDI Orbit to Sentinel 1 & 2 with different dates