def get_differences(band):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
RrsR = [
band.convolve(wavelengths_data, R_rs)
for wavelengths_data, R_rs in zip(wavelengths, R_rss)
]
RrsL = [
band.convolve(wavelengths_data, Lw) /
band.convolve(wavelengths_data, Ed)
for wavelengths_data, Ed, Lw in zip(wavelengths, Eds, Lws)
]
difference_absolute, difference_relative = zip(*[
calculate_differences(reflectance_space, radiance_space)
for reflectance_space, radiance_space in zip(RrsR, RrsL)
])
# Get all differences into one array
difference_absolute = np.array([x for y in difference_absolute
for x in y]) * 1e6
difference_relative = np.array([x for y in difference_relative for x in y])
difference_combined = np.stack([difference_absolute, difference_relative])
print(band)
return difference_combined
print(f" {band}")
with warnings.catch_warnings():
warnings.simplefilter("ignore")
RrsR = [
band.convolve(wavelengths_data, R_rs)
for wavelengths_data, R_rs in zip(wavelengths, R_rss)
]
RrsL = [
band.convolve(wavelengths_data, Lw) /
band.convolve(wavelengths_data, Ed)
for wavelengths_data, Ed, Lw in zip(wavelengths, Eds, Lws)
]
difference_absolute, difference_relative = zip(*[
calculate_differences(reflectance_space, radiance_space)
for reflectance_space, radiance_space in zip(RrsR, RrsL)
])
# Convert to 10^-6 sr
difference_absolute = [1e6 * diff for diff in difference_absolute]
short_name = band.label.replace('\n', '_').replace(" ", "_")
boxplot_single(band,
difference_absolute,
difference_relative,
labels,
saveto=f"results/per_band/OLI_Green.pdf",
sensor_name="OLI")
wavelengths_central = np.arange(330, 810, 1)
FWHMs = np.arange(6, 66, 1)
# Arrays for storing results - per FWHM, per central wavelength, absolute/relative
medians = np.tile(np.nan, [2, len(FWHMs), len(wavelengths_central)]) # Median
perc5s = np.copy(medians) # 5th percentile
perc95s = np.copy(medians) # 95th percentile
for i,center in enumerate(wavelengths_central):
print(f"Central wavelength: {center} nm")
for j,fwhm in enumerate(FWHMs):
boxcar = load_synthetic_sensor(sensor_type, center, fwhm)
reflectance_space = boxcar.band_average(wavelengths_data, R_rs)
radiance_space = boxcar.band_average(wavelengths_data, Lw) / boxcar.band_average(wavelengths_data, Ed)
difference_absolute, difference_relative = calculate_differences(reflectance_space, radiance_space)
difference_combined = np.array([difference_absolute, difference_relative])[:,0]
medians[:,j,i] = np.median(difference_combined, axis=1)
perc5s[:,j,i] = np.percentile(difference_combined, 5, axis=1)
perc95s[:,j,i] = np.percentile(difference_combined, 95, axis=1)
results_stacked = np.stack([perc5s, medians, perc95s])
results_absrel = np.moveaxis(results_stacked, 1, 0)
results_absrel[0] *= 1e6 # Convert to 10^-6 sr^-1
quantities = ["P5", "Median", "P95"]
for results_combined, absrel in zip(results_absrel, ["abs", "rel"]):
synthetic_sensor_contourf_combined(wavelengths_central, FWHMs, results_combined, sensor_type=sensor_type, absrel=absrel, label=label, quantities=quantities)
inds = np.searchsorted(FWHMs, [6, 10, 15, 20, 30, 40, 50])
from sba.bandaveraging import calculate_differences
from sba.io import load_data
from sba.chla import satellite_algorithms, satellite_algorithm_labels, satellite_algorithm_colours
from matplotlib import pyplot as plt
label, wavelengths_data, Ed, Lw, R_rs = load_data()
difference_absolute, difference_relative = zip(*[
calculate_differences(*func(wavelengths_data, Ed, Lw, R_rs))
for func in satellite_algorithms
])
fig, ax = plt.subplots(figsize=(7, 1))
bp = ax.boxplot(difference_relative,
whis=[5, 95],
showfliers=False,
labels=satellite_algorithm_labels,
patch_artist=True)
for patch, colour in zip(bp["boxes"], satellite_algorithm_colours):
patch.set_facecolor(colour)
ax.set_ylabel("$\Delta$ Chl-a [%]")
ax.grid(ls="--")
ax.axhline(0, ls="--", c="k")
ax.locator_params(axis="y", nbins=5)
ax.set_title(f"Convolution error in retrieval algorithms, {label} data")
ax2 = ax.twinx()
ax2.set_ylabel("$\Delta$ TSM [%]")
ax2.boxplot(difference_relative,
whis=[5, 95],
showfliers=False,
from sba.bandaveraging import calculate_differences
from sba.io import load_data_file
from sba.chla import HydroColor
from matplotlib import pyplot as plt
from pathlib import Path
data_files = Path("data").glob("*processed.tab")
labels, wavelengths, Eds, Lws, R_rss = zip(
*[load_data_file(file) for file in data_files])
difference_absolute, difference_relative = zip(*[
calculate_differences(*HydroColor(wavelengths_data, Ed, Lw, R_rs))
for wavelengths_data, Ed, Lw, R_rs in zip(wavelengths, Eds, Lws, R_rss)
])
for difference, absrel in zip([difference_absolute, difference_relative],
["abs", "rel"]):
unit = "NTU" if absrel == "abs" else "%"
plt.figure(figsize=(7, 2.5), tight_layout=True)
bp = plt.boxplot(difference,
whis=[5, 95],
showfliers=False,
labels=labels,
patch_artist=True,
medianprops={"c": "k"})
for patch in bp["boxes"]:
patch.set_facecolor("xkcd:tan")
plt.ylabel(f"$\Delta$ Turbidity [{unit}]")
plt.grid(ls="--")
plt.axhline(0, ls="--", c="k")
