Quick look at a frame¶

Open a frame of DISP granules, apply the product corrections and quality mask, chain the ministacks, re-reference, and look at a map and a time series. The granules are synthetic (disp_xr.tutorial), written with the layout of real DISP-S1 files and a known truth, so the result can be checked. Replace folder with your own granules.

In [1]:

import tempfile
from pathlib import Path

import matplotlib.pyplot as plt
import numpy as np

import disp_xr as dxr
from disp_xr.tutorial import make_disp_stack

folder = Path(tempfile.mkdtemp()) / "granules"
truth = make_disp_stack(folder, "three_ministacks", shape=(128, 128))
print(len(truth.paths), "granules in", folder)

import tempfile from pathlib import Path import matplotlib.pyplot as plt import numpy as np import disp_xr as dxr from disp_xr.tutorial import make_disp_stack folder = Path(tempfile.mkdtemp()) / "granules" truth = make_disp_stack(folder, "three_ministacks", shape=(128, 128)) print(len(truth.paths), "granules in", folder)

15 granules in /tmp/tmpvm4wbg78/granules

Open, correct, mask, chain¶

open_disp reads metadata only. Corrections and the mask apply per granule, before chain chains the three ministacks into one series from the first reference date.

In [2]:

ds = dxr.open_disp(folder, corrections=True)
ds = ds.disp.correct().disp.mask("recommended").disp.chain()
ds = ds.disp.rereference("auto")
print("relative to", ds.attrs["reference_date"], "| epochs:", ds.sizes["time"])

ds = dxr.open_disp(folder, corrections=True) ds = ds.disp.correct().disp.mask("recommended").disp.chain() ds = ds.disp.rereference("auto") print("relative to", ds.attrs["reference_date"], "| epochs:", ds.sizes["time"])

relative to 2020-01-01 00:01:53 | epochs: 16

Map¶

plot_map shows one epoch (the last by default) in millimeters on a scale symmetric about zero. Masked pixels and water are blank.

In [3]:

fig, (left, right) = plt.subplots(1, 2, figsize=(11, 4.5), layout="constrained")
ds.disp.plot_map(ds.time[len(ds.time) // 2].values, ax=left)
ds.disp.plot_map(ax=right)
plt.show()

fig, (left, right) = plt.subplots(1, 2, figsize=(11, 4.5), layout="constrained") ds.disp.plot_map(ds.time[len(ds.time) // 2].values, ax=left) ds.disp.plot_map(ax=right) plt.show()

Time series¶

plot_point takes a longitude and latitude (or row/col) and marks each reference-date change, where one ministack is joined to the next.

In [4]:

row, col = 64, 90
ax = ds.disp.plot_point(row=row, col=col, buffer_m=60)
plt.show()

row, col = 64, 90 ax = ds.disp.plot_point(row=row, col=col, buffer_m=60) plt.show()

Check against the truth¶

rereference records the pixel it chose in attrs["reference"]. The synthetic truth is known, so the chaining itself can be checked: without corrections, masking or re-referencing, the chained series must equal the displacement the granules were built from.

In [5]:

print("reference:", ds.attrs["reference"])
chained = dxr.chain(dxr.open_disp(folder))
error = chained.displacement.values - truth.cumulative
print(f"largest difference from the truth: {np.nanmax(np.abs(error)) * 1000:.4f} mm")

print("reference:", ds.attrs["reference"]) chained = dxr.chain(dxr.open_disp(folder)) error = chained.displacement.values - truth.cumulative print(f"largest difference from the truth: {np.nanmax(np.abs(error)) * 1000:.4f} mm")

reference: {"method": "auto", "row": 31, "col": 40, "x": 501015.0, "y": 4113555.0}

largest difference from the truth: 0.0000 mm