6. Quantitative analysis¶

Once the depth rasters are on disk, EuroFlood turns them into numbers — per-event and aggregated — so you can rank events, estimate exposure, and compare flooding across years, all without leaving Python.

This tutorial downloads real rasters, so it needs network access.

In [1]:

from pathlib import Path

import euroflood as ef

ef.settings.output_dir = Path("out")

# Download the depth rasters for the six largest events — enough to compare, and quick.
# (Drop the `.head(6)` to analyse every event; a download progress bar shows while it runs.)
cat = ef.floods("Zutphen, Netherlands", shape="bbox")  # bbox ROI (tutorial 02)
dl = cat.sort_values("area_km2", ascending=False).head(6).download()

from pathlib import Path import euroflood as ef ef.settings.output_dir = Path("out") # Download the depth rasters for the six largest events — enough to compare, and quick. # (Drop the `.head(6)` to analyse every event; a download progress bar shows while it runs.) cat = ef.floods("Zutphen, Netherlands", shape="bbox") # bbox ROI (tutorial 02) dl = cat.sort_values("area_km2", ascending=False).head(6).download()

In [2]:

dl

dl

Out[2]:

EuroFlood catalogue: 6 flood events · 2018-01-01 … 2024-01-01 · 73.7 km² total · ⬇ 6/6 downloaded

	collection	event_id	date	year	end_date	cluster_id	filename	download_url	area_km2	geometry	path
2	historic	1811386473	2018-01-22	2018	2018-03-05	358	WD_MERGE_2018-01-22---2018-03-05_duration_35_d...	https://jeodpp.jrc.ec.europa.eu/ftp/jrc-openda...	29.103	POLYGON ((6.14039 52.08873, 6.27499 52.08873, ...	out/flood_2018-01-22_id1811386473_8277ba1c.tif
19	historic	3623893553	2018-01-01	2018	2018-01-29	351	WD_MERGE_2018-01-01---2018-01-29_duration_28_d...	https://jeodpp.jrc.ec.europa.eu/ftp/jrc-openda...	9.277	POLYGON ((6.14039 52.08873, 6.27499 52.08873, ...	out/flood_2018-01-01_id3623893553_8277ba1c.tif
20	historic	3211422323	2023-12-11	2023	2024-01-01	358	WD_MERGE_2023-12-11---2024-01-01_duration_21_d...	https://jeodpp.jrc.ec.europa.eu/ftp/jrc-openda...	9.234	POLYGON ((6.14039 52.08873, 6.27499 52.08873, ...	out/flood_2023-12-11_id3211422323_8277ba1c.tif
28	historic	3351340227	2020-02-24	2020	2020-04-06	380	WD_MERGE_2020-02-24---2020-04-06_duration_35_d...	https://jeodpp.jrc.ec.europa.eu/ftp/jrc-openda...	9.097	POLYGON ((6.14039 52.08873, 6.27499 52.08873, ...	out/flood_2020-02-24_id3351340227_8277ba1c.tif
11	historic	3215720203	2023-10-09	2023	2023-11-27	346	WD_MERGE_2023-10-09---2023-11-27_duration_35_d...	https://jeodpp.jrc.ec.europa.eu/ftp/jrc-openda...	8.516	POLYGON ((6.14039 52.08873, 6.27499 52.08873, ...	out/flood_2023-10-09_id3215720203_8277ba1c.tif
3	historic	2217388040	2024-01-01	2024	2024-02-19	334	WD_MERGE_2024-01-01---2024-02-19_duration_35_d...	https://jeodpp.jrc.ec.europa.eu/ftp/jrc-openda...	8.474	POLYGON ((6.14039 52.08873, 6.27499 52.08873, ...	out/flood_2024-01-01_id2217388040_8277ba1c.tif

Per-event statistics¶

.stats() returns one row per downloaded event with, for each: the number of wet pixels, the max / mean / p95 depth (metres), the flooded area (km²), and the water volume (m³ and million m³). EFAS centimetres are scaled to metres automatically.

In [3]:

stats = dl.stats()
stats

stats = dl.stats() stats

Out[3]:

	event_id	date	wet_pixels	max_depth_m	mean_depth_m	p95_depth_m	flooded_area_km2	volume_m3	volume_Mm3
0	1811386473	2018-01-22	56816	6.78	0.509	1.730	22.7264	11561892.0	11.5619
1	3623893553	2018-01-01	18384	6.38	0.933	2.308	7.3536	6859016.0	6.8590
2	3211422323	2023-12-11	18279	4.27	0.730	1.980	7.3116	5336436.0	5.3364
3	3351340227	2020-02-24	18024	5.58	0.880	2.150	7.2096	6343796.0	6.3438
4	3215720203	2023-10-09	16145	5.57	0.794	2.240	6.4580	5125944.0	5.1259
5	2217388040	2024-01-01	14846	2.24	0.181	0.540	5.9384	1072996.0	1.0730

It is an ordinary DataFrame — rank the events by how much water they moved:

In [4]:

stats.sort_values("volume_Mm3", ascending=False)[
    ["event_id", "date", "max_depth_m", "flooded_area_km2", "volume_Mm3"]
]

stats.sort_values("volume_Mm3", ascending=False)[ ["event_id", "date", "max_depth_m", "flooded_area_km2", "volume_Mm3"] ]

Out[4]:

	event_id	date	max_depth_m	flooded_area_km2	volume_Mm3
0	1811386473	2018-01-22	6.78	22.7264	11.5619
1	3623893553	2018-01-01	6.38	7.3536	6.8590
3	3351340227	2020-02-24	5.58	7.2096	6.3438
2	3211422323	2023-12-11	4.27	7.3116	5.3364
4	3215720203	2023-10-09	5.57	6.4580	5.1259
5	2217388040	2024-01-01	2.24	5.9384	1.0730

A quick bar chart of the peak depth per event:

In [5]:

ax = stats.sort_values("date").plot.bar(
    x="date",
    y="max_depth_m",
    legend=False,
    title="Peak flood depth per event — Zutphen",
)
ax.set_ylabel("max depth (m)")

ax = stats.sort_values("date").plot.bar( x="date", y="max_depth_m", legend=False, title="Peak flood depth per event — Zutphen", ) ax.set_ylabel("max depth (m)")

Out[5]:

Text(0, 0.5, 'max depth (m)')

The aggregate envelope¶

.summary() combines every downloaded raster into a per-pixel maximum composite (so overlapping floods are counted once) and returns the same measures plus n_events — the "worst case seen" across the whole record:

In [6]:

dl.summary()

dl.summary()

Out[6]:

{'wet_pixels': 38920,
 'max_depth_m': 6.33,
 'mean_depth_m': 0.523,
 'p95_depth_m': 1.74,
 'flooded_area_km2': 23.52,
 'volume_m3': 12291067.2,
 'volume_Mm3': 12.2911,
 'n_events': 6}

The depth distribution of one event¶

Read a single raster and look at where the water actually was — most flooded cells are shallow, with a thin tail of deep water:

In [7]:

import matplotlib.pyplot as plt
import numpy as np

depth = dl.depths()[0]
array, *_ = depth.read()
wet_m = array[array > 0] * 0.01  # centimetres -> metres

_fig, ax = plt.subplots()
ax.hist(wet_m, bins=30)
ax.set(
    xlabel="water depth (m)", ylabel="pixels", title="Depth distribution of one event"
)
print(
    f"{wet_m.size:,} wet pixels · median {np.median(wet_m):.2f} m · max {wet_m.max():.2f} m"
)

import matplotlib.pyplot as plt import numpy as np depth = dl.depths()[0] array, *_ = depth.read() wet_m = array[array > 0] * 0.01 # centimetres -> metres _fig, ax = plt.subplots() ax.hist(wet_m, bins=30) ax.set( xlabel="water depth (m)", ylabel="pixels", title="Depth distribution of one event" ) print( f"{wet_m.size:,} wet pixels · median {np.median(wet_m):.2f} m · max {wet_m.max():.2f} m" )

41,525 wet pixels · median 0.25 m · max 6.34 m

In [ ]: