Using your own weather station in variety trials

Heading forecasts, GDD tables and G×E models all lean on daily weather. If the weather is wrong, they're wrong in ways that are hard to spot. A station on the research farm measures the field your plots are in. Most other sources measure somewhere nearby, or an average over a large area.

This article covers what to log, how to turn a logger's readings into clean daily numbers, the gaps and clock problems that catch people out, and how to use NASA POWER when your own station has nothing.

Why the station on the farm beats gridded data

Gridded weather products give a value for every point on a map. Each value stands for a grid cell that can be tens of kilometres across. That's fine for broad patterns. It misses what happens in one field.

• Frost pockets. Cold air drains into low ground. A trial in a hollow can get frost when the grid cell as a whole doesn't.
• Summer storms. Convective rain is patchy. A thunderstorm can drop heavy rain on one field and miss the next one over. A grid averages that away.
• Hills, lakes and shelter. Elevation, open water and shelterbelts all shift local temperatures.

For thermal time, small errors add up. A source that runs one degree warmer than your field adds about one GDD a day, or around 60 over a 60-day season. At 15 GDD a day, that's four days of development, enough to move a heading forecast noticeably.

What to log

You don't need a research-grade mast. You need a few things measured well.

• Air temperature and humidity, in a radiation shield, about 1.5 to 2 metres above short grass. An unshielded sensor in the sun reads too warm.
• Rain, from a gauge that's level, clear of trees and buildings, and has a clean funnel.
• Wind speed, gusts and direction, if you spray. Drift decisions depend on them.
• Optional extras: solar radiation, soil temperature at seeding depth, and leaf wetness for disease work.

Log every 5 to 15 minutes if you can. A reading once an hour can miss the true high or low between readings. Some loggers keep each interval's maximum and minimum, which avoids that.

Keep a simple station log: when it was installed, moved, cleaned or repaired, and when the battery died. You'll want it the first time a season looks odd.

From readings to daily numbers

Most crop models want one row per day: high, low, mean and rain. Getting there takes a few decisions.

Which day does a reading belong to?

Loggers usually stamp a reading at the end of its interval. A reading stamped 00:00 covers the last minutes of the day before, so put it there. Otherwise every day starts one reading early.

Which clock is the logger on?

This is the most common problem. A logger may be set to UTC, to local time with daylight saving, or to local standard time all year. Many are set once and never touched again.

If you read a logger that runs on UTC as if it were local time in Alberta, every "day" starts six or seven hours off. The overnight low can land on the wrong date, and so can an evening thunderstorm.

Write down the logger's clock when you set it up. If you inherit a station, check the file against something you know, such as the warmest hour of a sunny afternoon.

Rain totals

Some loggers record the rain since the last reading. Others keep a running total that resets at midnight or when a counter rolls over. For a running total, subtract each reading from the next one. A drop means the counter reset, so the new value is the rain since the reset.

After a long gap, the rain in between is unknown. Don't put it all on the first reading after the gap.

Gaps

A day with readings only until noon has a low but no real high. Using it as is puts a cold "day" into your heat sums. A simple rule works well: only count a day when its readings cover most of it, such as 20 of its 24 hours. Fill the other days from a backup source, and keep a note of which days came from where.

NASA POWER as a fallback

The NASA POWER project gives free daily weather for any point on Earth, with no account or key. Daily temperatures and rain go back to 1981, and the newest days arrive a few days late.

It's useful for three jobs:

1. Sites without a station, or seasons before you had one.
2. Filling a station's gaps, day by day.
3. Long history. Past years' weather at a site lets you play the rest of a season out for a forecast (see predicting heading date). Your station probably hasn't been there for decades.

Know its limits. POWER's values stand for a grid cell, not your field. It smooths out local extremes: a frost pocket can be colder than POWER says, and daily rain at a single point is where gridded data are weakest. POWER also asks to be credited wherever its data are shown.

A good habit for the first season is to run your station and POWER side by side and compare the daily highs and lows. A steady offset is normal. A big difference on a single day points to a local event or a station fault, and both are worth knowing about.

How Crossling does it

In Crossling you add your logger to a research station on the farm map (Farm map › Weather stations) and give it the station's time zone.

• Upload the logger's file. Crossling reads Campbell Scientific TOA5 files from LoggerNet, Onset HOBO exports, Davis WeatherLink exports (desktop and weatherlink.com), Ecowitt exports, and sheets of daily highs, lows and rain, such as Environment Canada or ACIS downloads. It recognises the columns and units, shows what it will import and what it left out and why, and lets you fix the column mapping, the units and the logger's clock (station time, standard time all year, or UTC) before anything is saved.
• Or let the logger send readings. A logger, or a script beside it (WeeWX, Home Assistant, Node-RED, or a Campbell logger), can send readings to one address with the station's own token. Sending the same readings twice changes nothing. The developer docs have the formats and a script to start from.
• Days are built carefully. A reading stamped at midnight closes the day before. A day counts when its readings cover at least 20 hours. Running rain totals are worked out, resets and all, and after a gap of more than six hours the rain in between is left unknown.
• Your station first, POWER for the rest. For each trial site at that station, a complete day from your station is used, and NASA POWER fills any day it missed (and the rain, if your station has no gauge). The site page says where its days came from, for example "Your station for 82 days, NASA POWER for 3".
• The same days everywhere. GDD, heat days, frost and dry spells on the site page, and the Outlook's heading forecasts and site models, all use these days. Today shows what your station measured in the last two hours beside the forecast.

The readings belong to your program, and only its members see them.

The demo program has a station at Lacombe with five weeks of summer readings made for the demo, so its site pages show station days and POWER days side by side. Try the demo, or create an account and add your own logger.