After the storm, which neighborhoods lost power — and who lives there?

When a hurricane knocks out the grid, the city goes dark — and from orbit you can actually see it. NASA’s Black Marble product (VNP46A2) turns the VIIRS day/night band into a daily, gap-filled, moonlight-corrected map of how much light each square kilometre is throwing off at night. Compare a clear night before landfall to a night after, and the blackout shows up as a plunge in radiance. Lay a free population grid on top, and you can start to ask the human question: not just where the lights went out, but roughly how many people were sitting in the dark.

Verified locally. Over Tampa Bay (27.6–28.1 °N), the VNP46A2 field Gap_Filled_DNB_BRDF-Corrected_NTL averaged 19.18 nW/cm²/sr on the clear night of 8 Oct 2024 — the eve of Hurricane Milton’s landfall — and fell to 9.35 nW/cm²/sr on the night of 11 Oct, a 51% drop in mean radiance across the AOI. The lights then crept back (13.8 on the 12th, 15.3 on the 13th) as crews restored power. That recovery curve is the grid coming back online, written in light.

What you can answer

• How much darker a city got after the storm — compare mean Gap_Filled_DNB_BRDF-Corrected_NTL over your bounding box, a clear night before vs. a night after (units: nW/cm²/sr)
• Which neighborhoods went dark — the grid is ~500 m, so you can map the per-pixel drop and see which districts lost the most light
• How fast the grid recovered — track the AOI mean night by night and watch the radiance climb back toward its pre-storm baseline
• Roughly how many people were in the dark zone — overlay WorldPop 1 km population on the pixels that lost light, and sum the population underneath
• Which counties to name — clip your AOI against geoBoundaries ADM2 so you can say “Pinellas and Hillsborough,” not just “a bright blob”
• A before/after story for any storm since 2012 — VIIRS has flown nightly since then, so the same recipe works for past hurricanes, ice storms, or wildfire evacuations

What you can NOT answer with these datasets alone

• Confirmed power outages — a darker pixel is consistent with a blackout, but clouds, smoke, storm surge, and evacuations also dim a city. Cross-check utility outage maps before claiming a grid failure.
• Who, exactly, lost power — WorldPop is a modeled ~1 km population estimate, not a customer list; it tells you the order of magnitude of people affected, not addresses or households
• The cause of any one dark patch — the gap-filled product blends nights to fill clouds, so a single dark pixel can reflect missing data, not a real outage. Check the quality flags.
• Economic damage or injuries — radiance is light, not dollars or lives; impact assessment needs ground reporting, insurance data, and damage surveys
• Indoor vs. outdoor light — VIIRS sees streetlights, signage, and spill from windows together; it can’t separate “houses dark” from “streetlights off”
• Demographic equity on its own — WorldPop gives counts, not income, age, or vulnerability; pair it with census data to ask who was least able to recover

Code template (Python, cloud-direct)

Verified locally. VNP46A2 is a daily HDF5 tile on a 2400×2400 grid. The brightness lives at HDFEOS/GRIDS/VIIRS_Grid_DNB_2d/Data Fields/Gap_Filled_DNB_BRDF-Corrected_NTL, with lat/lon arrays in the same group so you can subset to your bounding box before averaging. Read the real _FillValue and scale_factor from the dataset attributes — on the granules I checked the fill was -999.9 and the scale was 1.0, with units nW/cm²/sr.

import os, re, warnings, collections
warnings.filterwarnings("ignore")
import earthaccess, h5py, numpy as np

# load Earthdata creds from .env without `source` (passwords can break the shell)
for line in open(".env"):
    m = re.match(r'\s*(?:export\s+)?([A-Z0-9_]+)\s*=\s*(.*)\s*$', line)
    if m: os.environ.setdefault(m.group(1), m.group(2).strip().strip('"').strip("'"))
earthaccess.login(strategy="environment")   # free Earthdata Login

W, S, E, N = -82.8, 27.6, -82.2, 28.1       # Tampa Bay (Hurricane Milton)
GRP  = 'HDFEOS/GRIDS/VIIRS_Grid_DNB_2d/Data Fields'
NTL  = f'{GRP}/Gap_Filled_DNB_BRDF-Corrected_NTL'

def aoi_mean(granule):
    f  = earthaccess.open([granule])[0]
    h  = h5py.File(f, 'r')
    ds = h[NTL]
    lat, lon = h[f'{GRP}/lat'][:], h[f'{GRP}/lon'][:]
    fill  = float(np.array(ds.attrs.get('_FillValue', -999.9)).ravel()[0])
    scale = float(np.array(ds.attrs.get('scale_factor', 1.0)).ravel()[0])
    iy = np.where((lat >= S) & (lat <= N))[0]
    ix = np.where((lon >= W) & (lon <= E))[0]
    sub = ds[iy.min():iy.max()+1, ix.min():ix.max()+1].astype('float64')
    sub[sub == fill] = np.nan
    return np.nanmean(sub * scale)

res = earthaccess.search_data(short_name="VNP46A2",
        temporal=("2024-10-05", "2024-10-13"),
        bounding_box=(W, S, E, N))

by_date = collections.OrderedDict()
for g in res:
    d = g['umm']['TemporalExtent']['RangeDateTime']['BeginningDateTime'][:10]
    by_date.setdefault(d, []).append(g)

for d, gs in by_date.items():
    print(d, round(aoi_mean(gs[0]), 3), "nW/cm2/sr")
# 2024-10-08 -> 19.18  (clear night before landfall)
# 2024-10-11 ->  9.35  (after Milton)  == ~51% drop in mean radiance

# next: keep the dark pixels (big per-pixel drop), reproject WorldPop wpic1km
# onto them, and sum population to estimate how many people went dark.