Data, lab
What does the model do to a single fire?
Every coefficient Emberline uses, exposed as a calculator. Pick a region, move the numbers, and see exactly how a VIIRS detection becomes a MODIS-equivalent count.
Detection probability
How likely is it that Aqua MODIS also detected a fire that VIIRS saw, given the fire's radiated power? Fitted on 2012–2018.
P(Aqua also detects), Australia
—
FRP at 50% (95% CI)
—
CI MW. Shaded: 9–11 MW “thresholds of FRP detection” (Laurent et al. 2019, citing Roberts & Wooster 2008 and Schroeder et al. 2010) — a sanity check, not the same quantity.
P(MODIS detects) = 1 / (1 + exp(−(b₀ + b₁ · ln FRP)))
= 1 / (1 + exp(−(undefined + undefined · ln 10)))VIIRS → MODIS-equivalent converter
Paste the FRP of any set of VIIRS detections (MW, comma- or space-separated). Emberline weights each by its detection probability and scales by κ.
10 valid values.
VIIRS detections
10
Σ P (expected MODIS sightings)
0.00
κ (pixels per MODIS detection)
—
MODIS-equivalent count
—
Aqua-equivalent count = κ · Σ P(FRPᵢ) = undefined × 0.000 = —
Step-ratio calculator
How much did yearly detections change between two periods? Try the default (2009–11 vs 2012–14) to see the sensor jump, then flip to harmonized.
Mean before
—
Mean after
—
Ratio
—
ratio = mean(raw, 2012–2014) / mean(raw, 2009–2011) = — / —
Overpass sampling
What share of a day's peak fire activity does a satellite see at a given local overpass time? Published diurnal cycles (Andela et al. 2015, Table 1), for illustration.
A sees, of peak
93%
B sees, of peak
25%
B ÷ A
×0.27
share(t) = exp(−(t − h_peak)² / 2σ²) h_peak = 13.08 h, σ = 1.09 h (baseline ignored)
Trend calculator
Ordinary least squares on the harmonized yearly totals for any range of years.
Change per year
—
Slope (detections / yr)
—
R²
—
|t| > 2.1 ≈ significant
—
need ≥ 3 years
Anomaly calculator
Where does one month sit against the same month in every other year, on the harmonized record?
Jul 2023
—
Median of other years
—
× median
—
Percentile
—
no harmonized value for this month