Geospatial Time Series Analysis

July 30, 2023 · View on GitHub

Methods to stack geospatial rasters and run memory-efficient computations.
DOI

Motivation

Sometimes your data is larger than your local memory and not everyone has access to high-performance computing environments. This excludes individuals from participating in studies with the objective to process larger-than-memory datasets.

When data cubes are too large to load into memory, Zarr and dask enable chunked storage and parallelized data retrieval. GTSA streamlines this process for geospatial raster datasets, enabling memory efficient comptutations along both the spatial and temporal axes.

Installation

Download and install Miniconda

After installing Miniconda set up Mamba (optional but recommended)

$ conda install mamba -n base -c conda-forge

Clone the repo and set up the conda environment

$ git clone https://github.com/friedrichknuth/gtsa.git
$ cd ./gtsa
$ mamba env create -f environment.yml
$ conda activate gtsa
$ pip install -e .

Python examples

See Jupyter Notebooks for example Python code.

Command Line examples

See command --help for more information about each command listed below.

Processing

Stack single-band rasters and chunk along the analysis dimension

 create_stack --datadir data/dems/south-cascade \
             --date_string_format %Y%m%d \
             --date_string_pattern _........_ \
             --date_string_pattern_offset 1 \
             --outdir data/dems/south-cascade \
             --dask_enabled \
             --overwrite

Run memory-efficient time series analysis methods using dask

Basic --compute options include count, min, max, mean, std, median, sum, and nmad.

computation=count
gtsa --input_file data/dems/south-cascade/temporal/stack.zarr \
     --compute $computation \
     --outdir data/dems/south-cascade/outputs \
     --workers 8\
     --dask_enabled \
     --overwrite \ 
     --test_run

Linear regression

gtsa --input_file data/dems/south-cascade/temporal/stack.zarr \
     --compute polyfit \
     --degree 1 \
     --frequency 1Y \
     --outdir data/dems/south-cascade/outputs \
     --dask_enabled

Higher-order polynomial fits

degree=3 
gtsa --input_file data/dems/south-cascade/temporal/stack.zarr \
     --compute polyfit \
     --degree $degree \
     --frequency 1Y \
     --outdir data/dems/south-cascade/outputs \
     --dask_enabled

Apply a custom function you define in gtsa.custom.func.

gtsa --input_file data/dems/south-cascade/temporal/stack.zarr \
     --compute custom \
     --outdir data/dems/south-cascade/outputs \
     --dask_enabled

Visualization

Convert single-band rasters to Cloud Optimized GeoTIFFs (COGs)

create_cogs --datadir data/orthos/south-cascade \
            --outdir data/orthos/south-cascade/cogs \
            --workers 8 \
            --overwrite

Create interactive folium map for efficient visualization

create_cog_map --pipeline notebooks/visualization/pipeline.json \
               --output_file map.html \
               --zoom_start 11 \ 
               --overwrite

See rendered interactive map examples here.

Sample data

Download sample Digital Elevation Model (DEM) data

download_data --site south-cascade \
              --outdir data \
              --product dem \
              --include_refdem \
              --workers 8 \
              --overwrite

Download sample orthoimage data

download_data --site south-cascade \
              --outdir data \
              --product ortho \
              --workers 8 \
              --overwrite

Data citations

Knuth. F. and D. Shean. (2022). Historical digital elevation models (DEMs) and orthoimage mosaics for North American Glacier Aerial Photography (NAGAP) program, version 1.0 [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7297154

Knuth, F., Shean, D., Bhushan, S., Schwat, E., Alexandrov, O., McNeil, C., Dehecq, A., Florentine, C. and O’Neel, S., 2023. Historical Structure from Motion (HSfM): Automated processing of historical aerial photographs for long-term topographic change analysis. Remote Sensing of Environment, 285, p.113379. https://doi.org/10.1016/j.rse.2022.113379

Similar software

Complimentary software