My adventures with a weather station

Phase 1. Fantasising. I'd always wanted a weather station, but I'd always thought that they were out of my price range. This distant love affair lasted several years. Indeed, even as a child I yearned for a maximum-minimum thermometer every Christmas. I always got clothes instead though. Fortunately this year my book royalties were slightly more than I expected so I decided to go for it. (Please buy The psychology of language if you haven't done so already, as now I'm dreaming of a sunshine recorded.) No more going away from home with heavy heart, desperately searching for someone to come and take my readings! I wasn't starting to get quite reluctant to go on holiday even, just in case I missed some memorable temperature or rainfall event.

Phase 2. Choosing what to buy. I found this relatively easy. My constraints were: (a) price - I couldn't afford thousands; (b) I wanted something semi-automatic - in particular I wanted to be able to collect data while I was away; (c) hence something with a computer interface - I have an Apple Macintosh; (d) otherwise, I wanted something that was as accurate as possible. All of these pointed to a Davis weather station of some sort. I also asked on, and many people there recommended the Davis.

Somehow I'd got a Davis catalogue (I think I'd got one some time ago off the company Skyview). What I chose was the Davis Weather Monitor II (the most expensive, but it has all the functions), with combined external temperature sensor and humidity meter, and raingauge. It comes complete with anemometer. I also got the weatherlink data logger plus Mac software. I also decided to go for the sensor mounting arm, for simplicity. The hardest part was working out what length of cable extensions to get. The extension cable is quite expensive, and the longer it is, the greater the likelihood of false readings. On the other hand, I wanted the sensors to be as exposed as possible.

Phase 3. Choosing a supplier. My original plan was to buy from Skyview. However, John Dann of Prodata introduced himself by email and gave me a quote. (His email address is: His price was much cheaper than the others so I decided to go with him. The only disadvantage is that Prodata don't take credit cards. I found the prospect of sending off a cheque a bit alarming, so I did a bit of research. They are a real company, and John Dann offered some letters of recommendation. I made the right decision.

In deciding what to buy, John Dann was extremely helpful, which is another reason to go with Prodata. Getting everything working correctly isn't as straightforward as it looks.

So I sent my cheque off and waited. There was a bit of a delay as one component was out of stock, but the big parcel arrived in three weeks. Usually, I'm told, it's much quicker.

Phase 4. Installation. This was the difficult bit for me. I am useless at, and hate, DIY. It was like second-year woodwork at school gone mad. I wanted to do it properly. Do not understimate the difficulty of this stage.

The parcel arrived on Monday 8 February 1999. First, I unpackied everything and checked that everything was present. This was a wonderful moment; just like your dream Christmas present of all time. Second, I connected up the temperature sensor to the Monitor II console and checked that that worked, and I also got the hang of the console. It's very straightforward to use.

A. Placing the station. This is nowhere near as easy as it sounds. I got the 12m extension cables, and had estimated where this put the station on the lawn. I drilled three holes in the edge of my study window frame so that the cable could run from the junction box on my window sill outside.

Connecting the cable to the cable extension is not straightforward either. The junction has to be completely watertight. The cables come with telephone-type sockets that are very convenient. These, however, are not waterproof. The recommended procedure is to cut the sockets off the cable and splice the ends together. This seemed alarmingly irreversible, so I made a temporary join and sealed the junction box within plastic bags and tape. I was going to seal it all with waterproof tape, but when I opened this packet it was horrible green goo. I dropped this idea.

So I dropped the sensor from the window and with the help of my wife decided on the final position. It would be terrible to place the sensor mount and then find the cable is two inches too short, wouldn't it? So we spent some time laying the real cable out.

Then the real horror of the cable struck us. What do you with it? It's black, thin, and ugly. It could also be dangerous, layed across the grass ripe to be cut with a lawnmower, or to trip someone up on the path. We minimised the run across the gravel patio area and then ran a little diversion across the edge of the lawn. This shortened the distance between the location and the house. The desired location is four times the distance from the height of objects around, but this is impossible. I aimed for two, but even with the extension cable this proved impossible. (Longer extension cables are available, but you might get signal loss then, apparently.) I did the best I could, but it isn't perfect. But nothing immediately overhangs it, it is pretty exposed, it's over grass, and is very exposed to the south and west. I now wonder if I should have got more extension cable. Think about this stage very carefully, as once everything is set up, it's going to be a real pain to change it.

B. Installing the mount. Next the post we had hanging around turned out to be too small. Eventually we decided to buy a special one from the garden centre. What's an extra £20? My wife was more alarmed by the aesthetics than I was, but now I see her point.

Having chosen the best position, I borrowed a mallet and made a hole. It's much easier if you make a small hole first, fill it with water, and leave it overnight. I learnt this tip off a farmer. Even I found this bit easy.

I got a special post holder for £6 at the local garden centre. This was easy to hammer into the ground. I bought a 3"x3" fence 7' post and got my wife to saw the end off so it would be 4' exactly above the ground (she's better at this than me). I fixed it but ran out of the right size screws. I had no difficulty in putting the post solidly into the ground, when I realised that these clever bracket designers had designed it so that the long screws wouldn't meet in the middle: they'd offset the holes! Splendidly ingenious. How do they think of these things?

C. Installing the pieces on the arm. This bit was more or less straightforward. The instructions are good. I hit two small snags. First, I assembled the radiation shield incorrectly the first time because I couldn't tell the difference between a closed and flat plate. This was wholely my fault. Second, I found a small problem with the rain gauge. I nearly didn't follow the instructions to test each component indoors, but I'm glad I did. Each tip of the bucket is supposed to be .2mm. When I tried it, it registered .3mm, then .2m, then .3, then .2, etc. Emails to John Dann suggested the answer: the console had come precalibrated for an imperial gauge, while of course the gauge is metric. It is very easy to change the calibration at the console (although I find the Enter key method a bit fiddly). Then all was fine.

D. Assembly. We decided to put the cables through a special hose used for putting cables through a pond, so I dissambled everything I'd done so far and went in search of the hose. Needless to say, the garden centre was out of stock of the right size. After a week my special hose it arrived.

And then it rained. After a dry month, just when we were ready to do the final outdoor assembly, it poured. A good excuse not to do anything.

Eventually we had a clear weekend. Then I looked at the screws that fix the arm mounting bracket to the post: these are special lag screws, and need a socket set. Needless to say, I didn't have a socket set. I didn't even know what one was. So another delay.

So in the following week I got a socket set. I went to one of those big DIY stores on the edge of town that I've managed to ignore for several years. I felt like a real mechanic. The next Sunday, my first free time, we fixed the mounting plate to the post. We argued about this; what was the best way to do it to ensure that it was level. We used a marker pen. The socket set made easy work of this. You do need two people for this bit, I found. Then the sensor arm is easy to fix. Then it poured with rain again. When it stopped, I put the rain gauge on. This bit was easy too. Then I left it with many layers of plastic bags protecting the telephone sockets at the end. The end was almost in sight!

In the week we tried putting the cables through the hose. This was much harder than it looked. We got a little way and then it was too dark to see. Fortunately my brother-in-law, who is good at DIY, bless him, visited. Things went smoothly after this. We cut the hose lengthways, and put the cable in. We put the junction boxes in another box, sealed it all up waterproof tape, and wrapped this in layers of plastic. I was very careful about this: whether this is sufficiently dry, we will see. Then we ran the cable up to my window where the monitor is, tidied everything up, connected all the wires, and switched it on...

Saturday 21 March 1999. It works! (slightly to my surprise). Calibrating the wind vane direction was much simpler than it sounded, much to my relief. (It does need two people, really, and a compass, of course.

E. Worry. Now the worry phase. My first impression was that the thermometer is very accurate, and the radiation shield seems to be very effective. The rain gauge seems to under-record: last night my old raingauge indicated 8mm, while the Monitor said 5mm, but I haven't really explored this yet. The wind speed is much less than I think it shoud be: the trees are swaying, yet it indicates just 10mph. Perhaps this is its location. I don't fancy putting it 10m above our roofline at all; I can live with it, as it is.

In the evening, I read the Weatherlink manual again (well, I do use computers a great deal, and we have this tradition ...), and discover that I should set the calibration information from the software, not the hardware console, as I've done. So I have to depower everything (I find the backup battery surprisingly fiddly to get in and out) and do this again. (Once the console is depowered this is relatively easy.)

On the other hand, the software is very easy to use, in spite of some alarmingly long pauses ...

On reflection, I think the rain gauge might be under-recording a bit; I'm still exploring this. I think the temperature sensor is about right; possibly a fraction high in the sun, but then again, may be not. The wind speed sensor definitely under-records, but this is because of location, and I'm willing to accept this compromise.

F. Disaster. Monday 17 May 1999: my wife is cutting the edge of the lawn with some long shears, and snip: right through the protective shielding we installed, and nicked some cables. I should have marked its location. The external humidity sensor and windpseed aren't giving any readings. I'm going have to effect some cable joins. In the meantime, I've made it waterproof, which is the most important thing.

... June 26th: At last a dry day with some free time. We repaired the severed cables. It was easier than I expected, but fiddly; again, two people are useful for manipulating the little juntions. (I used the connectors supplied by Davis.) Again discovered the importance of having the right tools: I now have a nice new shiny pair of wire strippers. The whole junction is sealed in a little plastic box with masses of plastic and waterproof tape. And now it all works again, hurrah.

G. New trouble. July 6. Overnight the monitor display goes funny. Some of the elements of the LCD become very faint, making the figures very difficult to read. I try depowering it and starting up again, but this makes no difference. It looks like it's back to the manufacturer under warranty.

.... Something funny happens for several hours on August 19th: the wind speed records zero, even though the anemometer cups are spinning round in a very brisk northeasterly breeze, and the windspeed direction is showing correctly. Looking at the data log, the measurement seemed to drop to zero sometime in the early morning. I wonder whether my reconnection has somehow become severed, but I can't see how. Then as mysteriously as it went, it comes back again, about midday. Odd.

By August 29, John Dann had sent out a replacement monitor. I installed this without difficulty and all seems to well once more ...

I've worked out that the wind speed problem is linked to high humidity (not temperature, which was my first guess). This is probably a connection somewhere - probably in my repair, which is now so sealed up I daren't think about opening it.

... October 27. I now have sincere doubts about the humidity being the cause of the wind speed problem. I just can't discern any real pattern - except it's worse at night. Fortunately, I've had an email from someone in the States who appears to be experiencing exactly the same problem. Fortunately because this suggests that it isn't my repair. One new possibility is that it's the mounting.

... November 4. Take the anemometer cups off the spindle and replace them. Find a dead earwig. When I put them back: hey presto! It starts working again! My joy lasts for several hours, but then when it gets dark, the wind speed goes down to zero (in a gale).... November 15. My latest hypothesis: the anemometer only works when the sun is shining on it.

... November 17. I tried unplugging the cable from the junction box as rapidly as possible and got a maximum reading of 4. This morning it was working, and I put my ear to the anemometer and could clearly hear the click of the switch from some way away. I've tried thi when it hasn't been working and I don't think I heard the same clicking. (It's a little difficult, it being windy and when you're not sure what you're listening for.) So my current theory: sometimes the switch isn't working. I find this strangely comforting.

... December 20. A new worry. It's a very cold night (-5.3C), but after sunrise the temperature starts rising slowly. But then it suddely plummits down to -10 - according to the display. The data stored in the Weatherlink don't reflect this, however: they show a steady, slow rise. Later I work out what's going on: the switch on a light plugged into the same extension cable sometimes affects the minimum temperature on the Monitor display, lowering the displayed minimum by several degrees.

... January 18, 2000. The anemometer is still playing up. I still worry about the rain gauge. And I have a new worry: should I have set the wind vane to magnetic north?

... January 24. Apparently magnetic north is fine; the difference here is just a few degrees. There was probably more error setting it up.

The anemometer has to go back to have the switch checked. Digging up the cable and dismantling the junction boxes would be a major undertaking, so I'm finding out it it's possible to disconnect the cable.

I've been experimenting with my rain gauge. I've been dripping 100 ml of water in. I've done it twice, and each time it registered 3.6mm. I make the diameter of the funnel 165mm, giving a surface area of 21,382.5 mm2. Assuming 100 ml has a volume of 100,000 mm3, I calculate that it should register as 4.67mm. This is consistent with my earlier observations against my old garden rain gauge. I had the impression that it was under-recording by about 30%.

... January 28. Although apparently 100ml needs to be dripped in over an hour rather than the five minutes or less that I did it in.

February 6. I try constructing a device for dripping 100ml over an hour into my rain gauge. I use a plastic milk carton in which I've made a little hole. Unfortunately, the first time I try it blows off. The second time the hole gets bunged up and when I try and remove the obstacle I get water all down my arm. Nevertheless, the reading is up to 4.0mm, which is more like it. I'll have to try again.

February 16. Disconnect the anemometer to send it back. It's easier to take to pieces than I expected. I decided to go for the "cutting the cable" option. But will it be so easy to put it back together again?

March 18. The anemometer must have been away longer than I realised. I got it back on Friday and on Saturday I reinstalled it. I reconnected the wires, and slightly to my surprise, it all seems to work perfectly. It can be windy again.

April 17. You know those spiked shoes you can wear that aerate your lawn? They have one-inch spikes on and you trudge up and down your lawn making little holes. If you've got cable buried beneath your lawn, be very, very careful. The horrible possibilities only struck me after I'd finished. Everything seems OK though.

March 22, 2001. Everything seems to be working fine still. What with this winter's snow, I occasionally wonder if I should have got a heater for the rain gauge, but I assume that it would need a power supply, which would be a big fuss. As it is, taking the snow out, defrosting it gently in the microwave, and pouring the ice melt SLOWLY back into the gauge seems to work fine.

My other worry is that the radiation shield isn't perfect, and that sunshine inflates the readings a little. If it does, I don't think that it can be very much. I particularly worry about low evening sunshine getting through the slats of the shield. But it's good to have something to worry about. I also wish I could have put the gauges in a more exposed site. If I were to do it again, I think I'd try to roof mount the anemometer. I still might, but it seems a big, and somewhat dangerous, tricky job. I'd also like a method of recording sunshine, but again the expense is prohibitive.

July 23 and 27, 2001. This web site generated interest from National Geographic TV channel, who filmed it and me in action on Monday 23 July. They also filmed in at the Hotel Botanico, Tenerife, on holiday, explaining how to take simple weather readings. More on this later, but look out for "One day, one world, one weather".

September 30, 2001. I get a new computer. A newer Mac, and it doesn't have a serial port. With some trepidation, I install the Keyspan software, plug in the serial-USB converter, copy the station details, plug in the WeatherLink ... and to my amazement it works first time.

March 26, 2002. For some reason I cannot download data from the Weatherlog. After prodding cables and restarts, I eventually decide more serious action is necessary. Everything worked fine after I: (1) unplugged the monitor and took out the battery; (2) reinstalled the software; (3) plunged in all the cables even more firmly. Of course, if I'd had more time I would have tried the more rigorous approach of trying each of these independently, but as I didn't, and didn't have much data to lose, I didn't.

October 19, 2002. My silky football skills let me down as the ball knocks the vane of my anemometer. It doesn't seem broken, and it's fiddly get the direction set again. Needless to say, I can't find the fiddly little alun key that you need to tighten it.

October 22, 2002. I realise I've probably put the anemometer vane back on the wrong way round, so the reading is exactly 180 degrees out.

December 19, 2003. We move! We move 9 miles from Pitroddie to Lundie. We just dig the post out of the ground, disconnect everything, put it in the removal van, and let the removal company take it. At the other end we strap the outside part to a fence post as a temporary installation, but it seems just fine. The monitor is in the garage. I wonder what the neighbours think I do at 9 o'clock every morning. In general moving the weather station seemed remarkably trouble free. I'm glad the new location is similar enough for the data to be considered as continuous.

January 1, 2005. Everything is almost working well. It is clear that the anemometer is just too low, and hence I am greatly under-recording the wind. Also the external humidity gauge no longer gives a reading; I suspect a loose connection outside somewhere.

January 20, 2005. I vaguely think I'd like to start all over again ... a wireless set-up would be ideal...

July 10, 2005. Things are starting to look shabby. The anenometer is too sheltered and almost covered in vegetation; the external humidity meter no longer works - there must be a break in the cable somewere.

October 27, 2005. Return from a trip abroad to discover my rain gauge has been grossly under-recording. The only thing that seemed wrong was that a couple of leaves were stuck to the bottom of the bucket. But surely all that water has to go somewhere - it can't evaporate that quickly, can it?

8 September 2006. The disadvantage of the location of the station is starting to show: it's inaccessible and the vegetation is growing densely around it. The anemometer records 1 mph gusts when it's very windy, 0 most of the time. I think the rain gauge is under-reading by about 30% because it's too sheltered. We prune around the station and clear out the rain gauge.

1 November 2006. A power surge when my wife unplugged the station and data logger midmonth caused the station to hang. But when I try to download the data at the end of the month, big problems; the whole logger has frozen.

2 November. I try again to download data, but it doesn't look good. My Mac just keeps freezing as it tries to download the archived data. Email John Dann. He too thinks it doesn't look very good. Start to think about a new, wireless station.

3 February 2007. I think the LCD display is breaking now. The final digit seems to be frozen at 7, unless I prod and manipulate the display.

8 February 2007. I decide I can take it no longer. I decide to replace the whole damned lot. So I order a brand new wireless Davis Vantage Pro 2. My, how weather stations seem to have come along in the last eight years. Smaller, more accurate, and no cables. After trawling the web, John Dann and Prodata again seem the best bet (note: unfortunately I don't earn anything for this free advertising). I phone up with my credit card details, and sit back and wait for delivery.

Some restraint is necessary though. I am tempted to add a solar sensor for sunshine, an fan-assisted shield, an extension for the anemometer, and a heated rain gauge for snow, but the cost would then be fantastic. They can always be added later, if necessary.

16 February 2007. The new weather station is supposed to be delivered today, but there's no sign by 5.07 pm. It's like Christmas when you're ten.

21 February 2007. After a short trip away my weather station is delivered. Unpack - it looks exciting and easy to install.

25 February 2007. We do the weather station. ItÕs remarkably straightforward. Hammering in the post holder is easy - the soil is good and relatively stone free. The only slight difficulty is getting it level. Hammer in the post, screw the post into the holder, try and get the top of the post level. It's deceptive on a slope. It does need two people. I had the anemometer on wrong by 180 degrees. But itÕs much easier than last time. Worry about the wind direction, whether itÕs level, and the rainfall recorder.

4 March 2007. Have the traditional concerns about the rain gauge under-recording. Is the gauge level enough? Is it making a difference?

20 March 2007. My first error: I open Weatherlink (for Mac) and get "Java error: Unable to open port". Unplugging the USB cable from my USB hub and plugging it in again seems to cure it though.

14 June 2007. Have a little bit of a scare as the weather station anemometer stops working. Go outside, chasing pheasants away, and itÕs just that one tall stem of grass has sprung up overnight and wrapped itself around the vane.

30 August 2007. Occasionally when opening WeatherLink on the Mac I get an error message along the lines "Failed - Java port exception". If this happens to you, don't worry - just unplug the USB cable from your computer (or hub, in my technology-laden case), wait a few seconds, and plug it in again. Annoying, but only 2/10 annoying.


1. Take it slowly.

2. Make sure you buy from somewhere that gives good after-sales advice. John Dann replied promptly to emails, even on a Sunday evening! I recommend Prodata most highly, although of course other suppliers might be as good.

3. Test each component, just as the manual says. I was tempted to skip this stage, but was really glad that I didn't. If I hadn't tested each component, I wouldn't have found out about the rain gauge.

4. Have the right tools. I found a socket set invaluable. I've used it since, too.

5. Many stages are much easier to do with two people.

6. Pay particular attention to your site. Once it's up, you won't want to take it down! The more exposed the location you choose, the better. Exposure is important.

7. Waterproofing: I find supermarket plastic bags and layers of electrical insulating tape invaluable.

8. Protect the cable and make sure everywhere knows where it is. Particularly the person who cuts the lawn.

9. I do recommend getting the computer interface. One of the greatest advantages is having continuity of data when you're away.

10. If I were to do it again, I'd go for a wireless version, regardless of cost. Perhaps one day I'll upgrade ...

So far, I'd say it's a great success, and I wish I'd done it earlier! So if you're thinking about it and wavering, go ahead. Don't delay!

A cheap alternative

But if you don't want to spend a large sum of money. You can keep a weather diary for nothing - you just note that day's weather. One of the best weather instruments is free - the sky. What sort of clouds are there? how cloudy is it? For years I got by with under £20 of stuff bought at the local garden centre. The minimum you need is a maximum-minimum thermometer and rain gauge. Both should be readily available (e.g. from your garden centre). The thermometers vary a bit so look through them on the shelves and pick one that appears to give an average reading. The key to accuracy is in the siting of the instruments. The rain gauge needs to be exposed as possible - mine's in the middle of the lawn. (If you do this and buy one that sticks in the ground you need to remember to remove it when mowing or playing croquet. I've gone through a couple in this way.)

The max-min thermometer needs to be exposed but on no account must sunshine fall directly on it at any time of day at any time of year. You can construct a screen, but it must be properly ventilated. For years I got by with my thermometer pinned to a north-facing wall. That's a pretty good compromise.

The other thing about the thermometer is the height. About 4' (1.1 m) is best, but on no account have it too close to the ground.

You can have an anemometer too, but for years I got by with a neighbour's weather vane and the Beaufort scale. Failing this, use the direction of the clouds.

You can also get cheap little humidity meters. They're fun and easy to use, but I've never really found them to be that informative. Make sure that you keep out of the rain though!

A barometer is very useful. For a long time I just used an old one of my mother's. You need to calibrate it for altitude, particularly if you're quite high up. Quite often the BBC give you a reading which you can use: look out for large areas of high pressure centred on the British Isles. (Look out for fine weather with light or zero winds.)

Easiest and cheapest of all - you can just note the weather. "Fine, sunny day".

I keep my data in an Excel spreadsheet. An example entry is below. Many of the things can be recorded without instrument (e.g. % of cloud, cloud type, whether there was any fog, thunder, gale, rain, hail, ground frost, air frost, snow, that day).

It is important to use approximately the same time for recording each day. The standard is about 9 am but I used to use 8 am. Don't forget that if you take readings at 9 am GMT/UT in the winter that's 10 am BST in the summer. (Automation means that I don't have to be there to take the readings, although it's most fun to do so.) The minimum recording is attributed to the day of measurement and the maximum and rainfall total is thrown back to the previous calendar day. In winter this sometimes leads to odd affairs like one day's minimum actually being recorded at 9.01 the previous (calendar) day, but it's always been done this way, and probably always will.

2000 October 18 Wednesday 13.9 5.6 4.5 9.7 0 0 0 0 0 0.0 nw 5 nw 4 14 100 70 1005 r 1008 s 1 clearing to mainly clear & sunny stratus, cumulus 3 1

Key: YEAR MONTH DATE DAY MAX TEMP MIN TEMP grass minimum 5 pm temp rain day (0.2 mm or +)? wet day (1.0 mm or +)? hail < 5mm that day? hail > 5mm that day snowfall? rain amount mm prep type wind direction am wind force am wd pm wf pm mx gust fog at 9am? airfrost? grfrost? thunder snowlay over 50% at 9am? sowdepth? humidity am humidity mmn pressam pdir (rise, static, fall) presspm pd no sun that day description of weather cloud type am 8 (cloud cover is traditionally recorded in how many 8ths of the sky are covered) cloud cover pm 8

The Beaufort wind scale

  1. Calm <1mph Smoke rises vertically
  2. Light air 1-3mph Smoke drifts
  3. Light breeze 4-7mph Smoke rises vertically
  4. Gentle breeze 8-12mph Leaves rustle
  5. Moderate breeze 13-18mph Paper blows about, small branches move
  6. Fresh breeze 19-24mph Small trees sway
  7. Strong breeze 25-31mph Large branches sway
  8. Moderate gale 32-38mph Big trees sway, difficult to walk against wind
  9. Fresh gale 39-46mph Twigs break off trees
  10. Strong gale 47-54mph Pots and slates blown off rooves
  11. Whole gale 55-63mph Trees uprooted
  12. Storm 64-72mph Smoke Widespread damage
  13. Hurricane 72+mph Smoke Disaster

  • Return to my weather front page