I was also kicking around the idea of an "open source" alternative to the consumer stations on the market by in 08'. It is a good idea for sure. Here is the list wish list I posted back then:
---
1) Real-time. There is no sense producing a custom weather station when
you can buy a complete La Crosse for under $80.
2) Accurate. The station sensors would need to be better than the low
end stations and comparable (if not better) than a Davis VP.
3) Logging. It would need to support PC-less operation using on board
memory. The greater the logging capability the better.
4) Expandable. Users should be able to add whatever sensors they want in
the future like UV, solar radiation, multiple temperature and humidity
sensors.
5) Wireless capability. Ideally the PIC controller could have it's own
Wifi transceiver. If it has Wifi, it might be able to upload directly to
weather severs without the need for a PC. There is a unit out there that
can do this with existing weather consoles. I forget the name of it but
can find it if you are interested. They use an open source router to do
this.
---
Back then I found an interesting board on Sparkfun. I forget exactly why we did not move forward with it but I think the board was lacking some capabilities.
Looking at their site today, it appears they made a new revision of the board recently and it seems to support most critical functions:
http://www.sparkfun.com/products/10586While it is $124 for the board, and you would need to buy the rain bucket and wind speed and direction sensors:
http://www.sparkfun.com/products/8942It could be a nice starter platform for a DIY station. The interesting part is that it has a programmable Arduino and open source firmware that can be customized for whatever purpose (open source firmware). Real-time capability is not a problem since the code can be completely customized. It can be connected via mini-USB or there are wireless options (X-Bee, BlueSMiRF). Perhaps there is even a way to use a Wifi transceiver and impelment a tiny TCP/IP stack so that data could be uploaded directly to the Internet for PC-less operation like the Davis WeatherLinkIP. There is even a LCD serial display that can be connected according to the comments via a TX line.
On the software side, maybe the real idea is to make a protocol / capability specification such that the hardware chosen is completely optional. For example, the protocol could ask the station what types of sensors it supports, whether it has a logger and how many records it can store. Someone on the site at Sparkfun also suggested mimicking existing protocols like Peet Bros which is not a bad idea. Perhaps all that is needed is a few extra packets that detail the station capability, logging requests, and maybe some utility functions like setting the time, etc.
If there is interest in this board let me know. I can also support this station in WUHU, including logging capability.