Add global MODIS LAI (MCD15A3H) smoothing pipeline in preprocessing

[zyj8881357]

Overview

This PR introduces a parallelized workflow for processing global MODIS LAI (MCD15A3H) data from 2003 to 2025. It is specifically designed and optimized to run efficiently on the NCCS Discover supercomputer, adding scripts for robust Quality Control (QC), temporal smoothing, and data compression under the @GEOSgcm_GridComp/GEOSagcm_GridComp/GEOSphysics_GridComp/GEOSsurface_GridComp/Utils/Raster/preproc/veg/lai_grn/. Please see the readme.txt for details.

Key Features & Technical Details

• QC & Smoothing (process_tile_array.py):

  • Implements QC filtering to pick "perfect" time points and applies baseline filling for snow/ice coverage.
  • Utilizes a TIMESAT-style Savitzky-Golay upper-envelope smoothing algorithm to recover true LAI time series from cloud/snow/shadow/aerosol noises.
  • Works on a 2.5x2.5 degree intermediate sub-tile.

• Tile Stitching & Compression (stitch_tiles.py):

  • Dynamically merges sixteen 2.5x2.5 degree intermediate sub-tiles into unified 10x10 degree tiles (following standard MODIS H/V naming conventions).
  • Employs optimized compression (ubyte + scale_factor + zlib) for global outputs.

• HPC run (submit_global_array.sh):

  • Executes the scripts in parallel across 20 compute nodes simultaneously.

@gmao-rreichle