Expert article
Modern precipitation observation: The pine needles don't get a vote
The 2026 WMO Statement of Guidance confirms what every network operator already knows: precipitation observation has gaps, and they cluster where the worst weather happens. What's new is that the money to close them is finally moving. So is the technology.
Somewhere in a mountain catchment right now, a tipping bucket is full of pine needles.
It hasn't reported precipitation in eleven days. The hydrologist looking at the data assumes it's been dry. It hasn't been dry. Two storms have passed through, but the funnel is clogged, and the next service window is in March.
This isn't a story about a careless agency. The team has quality control (QC) routines, cross-checks, automated flagging. The gauge will get cleaned eventually, the two missed storms will be marked as no-observation, and nobody will notice. That's just the arithmetic of running hundreds of mechanical instruments across difficult terrain with a small field crew.
In different forms and on different continents, this is the operational reality of precipitation observation in 2026. It's also why the World Meteorological Organization's latest Statement of Guidance (SOG) reads the way it does.
What the WMO actually said
The Statement of Guidance (SOG) for Atmospheric Applications, published in May, is not a sensational document. Statements of Guidance never are. But buried in the careful, committee-edited language is a finding that stopped me when I first read it.
In a 2025 assessment of 62 member states, half had only basic hazard monitoring and forecasting capability, and sixteen percent had less than basic. Look at the hazards in question: flash floods, riverine floods, drought, tropical cyclones, thunderstorms, heatwaves. Every one of them except the heatwaves is a precipitation hazard in some form. The global early warning ambition rests on a precipitation observing system that, in large parts of the world, struggles to reliably detect the rain.
The SOG rates surface precipitation coverage as marginal, and it names the regions: Africa, the oceans, High Mountain Asia. It also acknowledges an awkward circularity: in remote regions, the satellite estimates we lean on to cover surface gaps lack accuracy precisely because the surface measurements that should calibrate and validate them aren't there either.
None of this is new. The Rolling Review of Requirements has documented it for years. What's new is the momentum behind fixing it. The UN's Early Warnings for All initiative targets universal coverage by the end of 2027, and the SOG counts more than 60 projects worth over 140 million Swiss francs already being implemented, many targeted at the most vulnerable regions. This changes the question facing network directors. It's no longer whether to expand coverage, but how to do it without expanding the field organization that normally comes with it. Worth saying plainly: WMO doesn't endorse vendors or technologies, and shouldn't. What follows is my read of the operational implications, with the evidence cited so you can weigh it yourself.
The three-hundred-year-old problem
The rain gauge is one of the oldest instruments in continuous scientific use, and it hasn't fundamentally changed in over three centuries. You catch the water, then you measure how much you caught. Everything since, from tipping mechanisms and weighing cells to heated funnels and double fences, has iterated on the how. The fundamental act is still collection, and three structural problems come with it that no mechanical refinement has solved.
The wind takes the data. Wind deflects raindrops away from the gauge opening, and the harder it blows, the more you miss. The classic literature (Sevruk, Yang, the WMO Solid Precipitation Measurement Intercomparison) established this decades ago, and if you suspect modern gauges have fixed it, a 2025 review in the Royal Meteorological Society's Weather (Dunn et al.) says otherwise: wind-induced undercatch still costs modern tipping buckets 5 to 46 percent of rainfall depending on gauge design, siting and conditions, and for frozen precipitation at exposed sites the gauge may record less than a third of what actually fell. The error is unavoidable in principle, because the gauge body itself deforms the airflow that carries the drops, and the review concludes that existing correction methods are still not robust enough for wide adoption. The losses peak during high-wind events, which is exactly when you most need the number to be right.
The maintenance scales with the station count. A well-maintained tipping bucket takes anywhere from a couple of service visits per year to a dozen, depending on climate, siting and gauge type. Your network may sit at the low end; many well-run temperate networks do. The math still binds: across 200 stations, even mid-range assumptions put you north of a thousand visits a year. Every 50 stations you add means a bigger field team, and stations on a glacier mean a helicopter contract.
The failure mode is silence. A clogged funnel and a dry sky produce the same signal: no data. Quality control catches much of this, but the cases it misses tend to be in remote single-station catchments where there's nothing to cross-check against.
None of this is a complaint about operations. It's physics and arithmetic, and it's why extending the surface network into underserved regions can't be done by buying more of the same gauges.
Counting precipitation without catching it
Vaisala’s PRECICAP® Radar Precipitation Sensor RM60 doesn't catch any water.
There's no funnel or collection vessel, nothing for water to land in. It's a sealed unit that sits on a mast and points a narrow radar beam straight up. When a raindrop passes through the beam, the radar measures its size and fall speed, then does the same for the next one. Tens of thousands of individual particles pass through in a moderate rain event, each characterized independently.
That sounds like a research instrument. It isn't. It's a sealed industrial sensor designed to live on a pole for a decade or more in any climate, running on 1.5 watts and talking to your data logger over Modbus or ASCII. Set the patented PRECICAP technology against those three problems and you can see why we built it this way.
That sounds like a research instrument. It isn't. It's a sealed industrial sensor designed to live on a pole for a decade or more in any climate, running on 1.5 watts and talking to your data logger over Modbus or ASCII. Set the patented PRECICAP technology against those three problems and you can see why we built it this way.
Wind undercatch is something you model and correct on a conventional gauge. Here there's simply no orifice for the wind to deflect drops away from, so that whole error mode has nothing to act on. I won't claim the sensor is error-free. No instrument is, and a new measurement principle carries its own uncertainty budget, which is why we publish how it's characterized. But the bias that grows with wind speed, during exactly the events that matter, is absent by geometry. During Atlantic Storm Bert in November 2024, an RM60 measured through sustained 14 m/s winds and gusts above 23 m/s with no shield, no fence, no correction. I wouldn't hang an argument on one storm, but it sits inside a larger record: over 700,000 sensor-hours since 2020, from tropical rainforest to continental subarctic to high-altitude convective sites, compared against double-fenced and pit-mounted reference gauges.
Maintenance is mostly a story about what isn't there. No collection surfaces, no moving parts, no optics, no heating. There's nothing to clog or freeze or wear out. In our subarctic deployments, RM60 has measured through several winters while being exposed to snow and icing. No scheduled maintenance over the service life, and at low power consumption, a small solar panel runs the remote site your service van was never going to reach anyway.
The silence problem gets turned around: instead of failing without a notice, the sensor reports its own health, and it re-calibrates against physics during precipitation events. "Autocalibration" gets thrown around loosely in this industry, so here is what it means in this case: each measured raindrop is referenced back to base SI units (the measurement volume in lenghts, the Doppler frequency in time) and to the known physical properties of water and falling drops. A documented, unbroken chain of traceability, verified every time it rains. The full metrological argument is in a paper we're presenting at the WMO technical conference on instruments (TECO-2026) later this year.
And because every particle is measured individually, you get more than accumulation: real-time intensity, precipitation type, drop size distribution across 22 size classes and reflectivity factor. All from measured distinct particles rather than modeled bulk returns. The sensor that removes your scheduled maintenance visits happens to bring disdrometer-class data to every station.
The interesting question
Whether maintenance-free radar precipitation measurement works is, honestly, not the interesting question anymore. That's what the 700,000 hours and the reference intercomparisons are for.
The interesting question is what a precipitation network looks like when you're not designing it around a maintenance schedule. Stations go where the science says they belong, not where the service van can reach. Mountain catchments get densified without a helicopter contract. Rich microphysical measurement available from every station instead of a few reference sites.
For most networks the right answer isn't immediately replacing the tipping buckets you have. It's looking hard at the next 50 stations you haven't built yet: the gaps, the densification targets, the previously infeasible locations, the manual stations where volunteers are getting harder to find every year. That's where the next decade of precipitation observation is being decided: in expansion, not replacement.
One caveat from someone who sits in a lot of procurement conversations: tenders often gate on purchase price, even when the 10-year cost picture points the other way. We're happy to have either conversation. They tend to land in different places.
The pine needles
About that clogged gauge.
The funnel will eventually get cleaned, but the data won't come back. The two missed storms stay marked as no-observation, and the regional total for the month is biased low without anyone flagging it. Multiply that across hundreds of stations and decades of operation and those small biases add up to something real.
This isn't a failure of the team. It's what running a 400-station network with a field crew of twelve looks like, using instruments designed for a different century. The point of RM60 is that this story doesn't have to keep happening.
The WMO has named the gap. The technology to close it already exists.
The pine needles, for once, don't get a vote.
Juuso Pokkinen works in Sensors Product Management at Vaisala, where he has spent more than a decade taking weather and environmental products from concept to market. He works closely with national meteorological and hydrological services, aviation, maritime, and road weather customers. His background is in chemical engineering and environmental technologies.