Generative AI and Tempest/Weatherflow

Hey devs! Curious to see if there’s any AI integrators here. I’m currently an AI researcher/prompt librarian and enjoying analyzing the data I have access to for my unit.

I bought the Tempest System this weekend because I thought they were super neat in general. No intention of monetization, still in the exploratory phase. I really, really enjoy prompt engineering (in regards to GAI), so I made a weatherman role. Right now I’m logging all data by the minute and then feeding it manually into a GPT-4 32k model w the role, for analysis. Will automate this process soon.

Role response is attached.

Let me know what you think and if you’ve been working on anything similar!

GPT-4 32k w/ custom role:

:cloud: Well, let’s dive into these weather data observations from your station. We have a series of readings taken one minute apart, from 07:42 to 08:06. Let’s break down the key observations and what they mean for your weather forecast.

Wind Speed and Direction: The wind speed is relatively low, averaging between 0.0 and 0.85 m/s, with a maximum gust of 1.63 m/s at 08:00. This suggests calm conditions. The direction is primarily from the west-southwest (200-240 degrees), which in many regions could suggest a weather system moving in from the west.

Pressure: The atmospheric pressure is stable at around 25.76 inHg, which is a standard atmospheric pressure level. The slight increase from 25.761 to 25.767 inHg within this short period could suggest a stable or improving weather situation.

Temperature: The air temperature is mild and steadily increasing from 63.68°F at 07:42 to 65.84°F at 08:06. This is typical as the day warms up after sunrise.

Humidity: The relative humidity is decreasing from 64% at 07:42 to 56% at 08:06, but then it increases back to 60% at 08:06. This is a normal fluctuation during the day as temperature rises.

Illuminance, UV and Solar Radiation: All three are gradually increasing, which is expected as the morning progresses. No sudden spikes indicate normal weather conditions.

Rain Accumulation & Precipitation Type: Both are consistently at zero, indicating no rain during this period.

Based on these observations, I would predict a calm and dry day ahead with a slow increase in temperature. The stable atmospheric pressure and decreasing humidity suggest little chance of rain. The wind from the west could bring changes later in the day if this is a common weather pattern in your area, so it’s worth keeping an eye on that.

In meteorological terms, we’re looking at a stable weather system. The temperature is affected by the solar radiation, which also affects the humidity - as the temperature rises, the air can hold more water vapor, which can decrease relative humidity. The stable pressure suggests that the current system (whether it’s a high-pressure or low-pressure system) is staying in place, which means the weather is likely to stay consistent.

Remember, this is a prediction based on the limited data available and actual weather can be influenced by a wide range of factors. It’s always a good idea to check with a comprehensive local forecast for the most accurate information. :cloud:


definitely interesting route to follow. But the current output is pretty useless. No one would claim a raise of 0.06 inHg could suggest a stable or improving weather situation, based on 24 minutes of data.

1 Like

I am not currently working on anything similar but in the past I tried to use my different weather station to forecast my local winds.
And wind readings are a valuable input which if your Tempest is not clear of obstructions will be confused.
You could consider including more inputs which are not in your Tempest data. Such as where is the moon and where is the sun. It is simple to calculate where they are based on your location and date and time.
An interesting topic thank you.
Cheers Ian :slight_smile:

1 Like

I agree. After five or seventy days of data it’ll be quite a bit more interesting to see the analysis.

In the past what I found most useful for my forecast was comparing the diurnal cycle of barometric pressure with my history. And removing the daily and lunar cycles from the current change in barometric pressure to get the current change in barometric pressure due to other factors such as the weather systems and the air temperature warming making it lighter. If I was creating a new setup I would also input the 3 second winds to my AI because they indicate what the thermals are doing at your location.
Cheers Ian :slight_smile:

1 Like

How often are you ‘confirming’ the actual weather?