How wind speed affects methane emission rate estimates from satellite imagery
Every quantification method built on imaging spectrometry reduces, eventually, to one ratio: how much excess methane is sitting in the air column, divided by how fast the wind is carrying it away. The mass side of that ratio is the part reviewers rarely argue about. The wind side is where most of the disagreement lives.
Why wind sits in the denominator
The two methods used to turn a retrieved plume into a kilogram-per-hour number, the integrated mass enhancement (IME) approach and the cross-sectional flux method, both need a wind term to convert a static snapshot of concentration into a rate. IME multiplies the total excess mass in the plume by an empirical decay constant tied to wind speed. Cross-sectional flux integrates concentration across a line perpendicular to the plume axis and multiplies by the wind component crossing that line. In both cases, double the wind speed and you roughly double the reported emission rate, holding the retrieved mass constant.
That means a wind input that's off by 30% doesn't shave 30% off your confidence interval somewhere in the middle of the error budget. It sits at the front of it. Published uncertainty analyses on airborne and satellite plume quantification consistently name wind speed and direction as the largest single contributor to retrieved rate uncertainty, ahead of retrieval noise, plume detection threshold, and background subtraction combined.
Satellite imagery gives you a concentration field. The wind number has to come from somewhere else, almost always a reanalysis product.
ERA5, 10 m wind, and the gap it doesn't close
ERA5 is the default because it's global, hourly, and free. For methane work it's typically the 10 m wind speed field, sometimes blended with boundary layer height, interpolated to the plume's time and location. That's a reasonable starting point and a known compromise.
The compromise: ERA5's native grid runs around 31 km, downscaled to a quarter-degree product for distribution. A single pixel can cover a pad, the tank battery next to it, and a stretch of open range, each with its own surface roughness and its own local wind. The reanalysis wind describes the regional flow. The plume is responding to the wind 10 to 50 meters off the ground at one specific pad, shaped by terrain, vegetation, and nearby structures in ways a 31 km grid cell can't see.
Some workflows apply a log-law or power-law correction to step ERA5's 10 m estimate up to the effective transport height of the plume, since near-surface wind is slower than the air a few tens of meters up where the plume is advecting. Others pull higher-resolution products like HRRR where coverage allows, trading global availability for finer grid spacing over the continental US. A smaller number of operators still have ground-truth anemometer data at the facility, which is the gold standard and also the least available, since you're usually quantifying a plume precisely because you didn't have eyes on that pad already.
None of these fixes eliminates the error. Each just narrows it. Wind normalization in this context means picking a wind source, documenting why, and then stating the emission rate as a point estimate with an honest band around it rather than a single confident number. Attaching a wind source to a reported rate shows where its uncertainty actually sits.
What this means for reading a quantified plume
If you're comparing two detections of the same facility on different days and the reported rate swings by 40%, check the wind input before you assume the source changed. A calm day with light, variable wind is also the hardest day to get a clean cross-section, so low-wind detections deserve an extra look rather than extra trust.
This is also why a single annual flyover with one wind snapshot tells you less than a cadence that lets you watch the same pad across multiple wind conditions. One frame might land on a day with unusually strong or unusually calm flow, and you'd never know it without a second look. Methane Plume Map attributes daily, tasking-dependent spectrometer passes back to a specific site boundary, which is what lets you track one pad across enough wind conditions to tell a real rate change from a wind artifact.
If your draw area has a site you need watched past a single flyover, that's the use case this map is built for.