cd gsi/MissionSource/WGS84
source environment_config
module load gdal

SetLimits.sh MMtest -0.85 53.1 0.15 53.7

CreateProject.sh MMtest

GetAWSinfo.sh MMtest 30

#  <Use WinSCP to copy all annotated QuickLooks to local windows machine to visualise>
#  Download /lustre/w23/mattgsi/satdata/RF/AWS/QL/New_MMtest*.tif to local windows machine
#
#  Download /home/w23/mattgsi/gsi/MissionSource/WGS84/AWS_MMtest_Scenes.txt to local windows m/c
#
#  On local windows machine, cycle through all downloaded quicklooks, and delete unusable
#  scenes (too cloudy or only small portion of AoI covered) from AWS_MMtest_Scenes.txt.
#
#  Upload edited AWS_MMtest_Scenes.txt to /home/w23/mattgsi/gsi/MissionSource/WGS84


GetAWSdata.sh MMtest

#  Files for AoI extracted into dir: /lustre/w23/mattgsi/satdata/RF/AWS/MMtest/NDVI
# (only the bands needed for NDVI are downloaded)
#  [This also does the equivalent of "FourierAWS.sh MMtest", which woudl only need
# to be run if additional dates are downloaded later]
#
#  <optional - to get all bands for selected scenes>

GetAWSbands.sh MMtest

CreateAWS.sh MMtest

#  Now set up with AWS data remapped to AoI, with per-year Fourier(*4) stats,
# ready for running RFtrain etc
#  BUT only if we have (default) params for Landcover, Species & Structure (currently over Canada).
#  If not, we need to set up some other Target data to use for training, by editing the file:
#     /lustre/w23/mattgsi/satdata/RF/Projects/Models/AoI_Target_MMtest.txt
#  For example, add a replacement/new line:
#     Target/WheatYield,/home/w23/mattgsi/gsi/New_Target_Data/Crop_Data/processed/rejigged/wheat_YieldDensity_1000.tif
#  Then run the following:
#
#  <optional to extract Fourier-filtered images from multiple dates per band>

FourierBandsAWS.sh MMtest

AddNewTargetData.sh MMtest

#  This creates the new Target file for the AoI:
#     /lustre/w23/mattgsi/satdata/RF/Projects/MMtest/Target/WheatYield_MMtest.tif
#  We now may need to create a new Target Parameter set to specify which params to
# use in the subsequent RF processing:
#  This requires creating two new files (by editing other example files), for example:
#     /lustre/w23/mattgsi/satdata/RF/Projects/Models/Paramset_Wheat.txt
#     /lustre/w23/mattgsi/satdata/RF/Projects/Models/RFparams_Wheat.txt
#  (Note that these can now be used for any other AoIs which want to model with the same parameters)
#  We can now run RF:

RunProject.sh MMtest AWS Wheat

#  This generated Trained & Scored Wheat Yield at 30m (using 10km input Wheat Yield Density data)
#
#  As an experiment, then used the previous scored 1km WheatYield data
# (mean over 14 years) as new target,
#  so added a new line to AoI_Target_MMtest.txt:
#      Target/Wheat1kmYield,/lustre/w23/mattgsi/satdata/RF/1km/Wheat_Test_2/Scores_Historic/mean_ConditionalMean_wheat_YieldDensity_100.h17v03.tif
#  Then re-ran:

AddNewTargetData.sh MMtest

#   This created the new Target file for the AoI:
#     /lustre/w23/mattgsi/satdata/RF/Projects/MMtest/Target/Wheat1kmYield_MMtest.tif
#   Then created two nes fiels for the new Wheat1km paramset: Paramset_Wheat1km.txt & RFparams_Wheat1km.txt
#   The re-ran RF with new 1km target data (instead of previous 10km data)

RunProject.sh MMtest AWS Wheat1km

#   This generated Trained & Scored Wheat Yield at 30m (using 1km input Wheat Yield Density data)


#   Then using Sentinel-2 data (10m):

GetS2AWSInfo.sh MMtest 30

#  <Use WinSCP to copy all annotated QuickLooks to local windows machine to visualise>
#  Download /lustre/w23/mattgsi/satdata/RF/S2AWS/MMtest/QL/New_*.tif to local windows machine
#
#  Download /home/w23/mattgsi/gsi/MissionSource/WGS84/S2AWS_Lidar_good_scenes.txt
#  If any of the pre-selected "good" scenes are no good, then delete from
# S2AWS_Lidar_good_scenes.txt, and upload to original location.

GetS2AWSdata.sh MMtest

#  This downloads the Sentinel-2 imagery for the "good" scenes, just for the bands needed for NDVI
#  The resulting imagery is extracted/remapped to the specified AoI,
# and placed in /lustre/w23/mattgsi/satdata/RF/S2AWS/<AoI>/NDVI