scales of motion in the summer marginal ice zone

This repository contains code, data, and notebooks used for the paper "Observing floe-scale sea ice motion in the Greenland Sea marginal ice zone during summer," submitted to Annals of Glaciology. We examine the spatial and temporal scales of sea ice motion at floe scale using floe trajectories derived from optical satellite imagery. The analysis focuses on observations from the East Greenland Sea from 2003-2020 during the sunlit months (April-August).

structure

The repository contains folders for data, scripts, figures, and Jupyter notebooks. The processed imagery is stored externally in the Brown Digital Repository. Github limits the size of data, so we cannot save the satellite imagery here, however the processed floe data should be small enough to work. In general: if github lets you upload it, place the data in the data folder and give it a descriptive name. I prefer lower case names with no spaces but I won't fight you if you have different preferences. Code to reproduce each step of the analysis and to reproduce each figure should go into the scripts folder. Where possible, use relative pathnames so that the code in scripts reads data from the data folder. If the data is too large, place it in a directory outside the Git repository and provide instructions on how to obtain it. Analysis code and plotting code should be separated, so that the figures can be re-generated without additional computation time. This also lets us make minor changes to figures easily.

python environment

The python environment used to run the scripts here can be recreated using the ift_annals.yml file. After installing miniconda, open a terminal, navigate to the project folders, and run conda env create -f ift_annals.yml

scripts

New data processing pipeline: - process_fram_strait_v0 Github repo processes the matlab data and combines it with the truecolor and falsecolor satellite imagery. This is where initial data processing and cleaning is taking place. - 01_add_external_data.py Interpolate ERA5 wind and sea level pressure data to floe positions in the interpolated data; add ice edge distance to the interpolated and to the full dataset.

After activating the conda environment, the scripts can be run by navigating to the scripts folder in terminal and running e.g. python 01_extract_ift_data.py The function adding the NSIDC ice motion and sea ice concentration data requires these datasets to be downloaded separately.

01_extract_ift_data.py This script reads that matlab output Rosalinda generated, then produces CSV files with the time stamps, floe properties, and floe ids. (TBD: currently inside a notebook, needs to placed into a script) 02_interpolate_ift_data.py Makes a best estimate of the IFT floe locations at exactly 12 UTC each day to enable velocity estimation. Also compiles floe properties and rotation rates. Adds a qc flag based on speed z-score (must be less than 6), circularity (must be greater than 0.6), total path length, and a speed threshold (maximum speed must be larger than 1 pixel/day). 03_add_nsidc_info_ift.py Interpolates NSIDC ice motion and climate data record sea ice concentration to floe locations, and calculates the distance to the sea ice edge. TBD: add ERA5 wind data as well. Requires external datasets. 04_finding_polygons.py Searches all possible combinations of three sea ice floes, and makes a list of all the combinations with minimum interior angle greater than 20 degrees. The polygon data is large - 1.15 GB - so needs to be created locally rather than shared on github.
05_calculate_deformation.py Calculates strain rates.

Figure 1: Map with sea ice extents for each year

scripts (updated)

1. 00_plot_study_region.py Generates Figure 1, a map with sea ice extent for each year and the bounding boxes for other figures.
2. 01_pull_ift_data.py Accesses data on the Brown server. Data can be downloaded directly from the Brown Digital Repository instead.
3. 02_merge_external_data.py Interpolates the NSIDC ice motion and sea ice concentration data to the IFT positions.
4. 03_plot_data_summary.py Generates Figure 2 (Example of algorithm steps and addressing the area bias), Figure 3 (tracked floes), and Figure 4 (Data availability, maps)
5. velocity_figures.py Generates Fig 11 and Fig 12. TBD: Add scale vector to Fig 11.
6. rotation_figure.py Generates Figure 13

TBD: get scripts for FSD figures sorted out.

Figure 5: Estimating the minimum for the FSD Figure 6: Examples of polygons for deformation Figure 7: Results of the likelihood ratio tests Figure 8: Snapshots of FSD pdfs and ccdfs Figure 9: Seasonality of the FSD Figure 10: Tracked vs all floes FSD pdfs and ccdfs Figure 11: Comparison with NSIDC ice motion vectors Figure 12: Distributions of perturbation velocities Figure 13: Rotation rate distribution Figure 14: Slope of the deformation rates as a function of polygon length scale

data

floe_tracker folder contains ift_with_era5.csv which has the interpolated tracked floes with all the floe properties with wind speeds from ERA5 added in. ift_with_nsidc.csv is produced first; ift_with_era5.csv contains the same data but with the addition of the winds. The subfolder interpolated has the files for each year as produced by script 2. The subfolder parsed has the full IFT data including non-tracked floes produced by script 1. | --- | --- | --- | | Column name | definition | units | 'year' | Year (YYYY) | | 'datetime' | Date and time (YYYY-mm-dd HH:MM) | | 'floeid' | Floe ID | | 'xstere' | meters | 'ystere' | meters | 'longitude' | decimal degrees | 'latitude' | decimal degrees | 'area' 'perimeter' 'majoraxis' 'minoraxis' 'zeta', 'zetaest', 'u' | meters per second 'v' | meters per second 'bearing' 'speed' 'circularity', 'qcflag', 'seaiceconcentration', 'edgedist', 'unsidc', 'vnsidc', 'u5Dnsidc', 'v5Dnsidc', 'u15Dnsidc', 'v15Dnsidc', 'u31Dnsidc', 'v31Dnsidc', 'ualong', 'valong', 'uacross', 'vacross', 'Ufluctuating', 'Ualong', 'uwind', 'vwind', 'wind_speed'

figures

Figures should be print-ready for Annals of Glaciology. They should be eps, tif, or pdf if possible. (Some figures need to be compressed into jpg first and then converted, such as scatterplots with huge numbers of points). The figure width should be 85 mm for single column figures and 179 mm for two-column figures. Fontsize should be 9. Font should be Optima, or if that is not available, Arial.

TBD items

• Check font size in figures
• Update methodology for deformation section
  • Effect of reusing floes - how can we avoid it?
  • Selecting
• Description of updated FSD results
• Comparison of FSD results to prior work
• Description of updated deformation results