Expedition weather forecasts.

We rely more and more on good weather forecasting in the mountains. Free and paid internet resources are incredibly valuable these days. But we also go well beyond internet signal (for now…) with our mountain travels. I’ve long used someone at home, with good internet signal, to relay weather forecasts via satellite text messaging. I’ve been on the other end, sending forecasts from home to friends and colleagues in the wild. Someday this relay won’t be necessary, but it is currently very valuable. To transcribe complex weather forecast data to into a concise, useful satellite text message format is laborious; a lot to ask of those back at home in civilization. I’ve got a tight set of instructions for Rosie to follow, but it is still a task requiring many minutes each time I want info in the field. But I just worked out a short cut. ChatGPT will transcribe and shorten weather forecasts to satellite text format. It takes a little training and feedback at first, but then it can work out really well. Saving lots of time on the home front. Here’s the instructions that I had ChatGPT write after working out the formatting matters and such (bonus points to anyone that can guess where my example weather forecast is for this week? i.e. Where am I off to?):

Instructions:

  1. Take a Screenshot of the Weather Forecast
    Capture the weather forecast from your source (e.g., NOAA, MeteoBlue, or others). Ensure the forecast includes relevant data such as temperature, wind, precipitation, sun hours, and predictability.

  2. Upload the Screenshot to ChatGPT
    Upload the screenshot into the chat. Ensure the text in the image is clear and legible.

  3. Provide Necessary Context
    In your message, clarify that you would like ChatGPT to generate a satellite-ready weather forecast based on the screenshot. For example, you can write:

    • "Please generate a satellite-ready weather forecast from this screenshot."

    • Optionally, specify the elevation if it’s not obvious in the forecast, like “6529 ft” for a high mountain point or “2075m” for a different elevation.

    • Add prompts for how to handle data fields that aren’t in the given forecast screenshot. “Omit data that is not available”

  4. ChatGPT's Response
    Once you upload the image and provide context, ChatGPT will:

    • Extract the data from the screenshot.

    • Format it into a concise, standardized satellite-ready forecast based on the fields you specified.

Output Format Example:

"4/21, 2075m, 4/22 MB, 17–31 F, W11mph, 0in, 9h sun, 4/4 pred." This includes:

  • Today's Date: 4/21 (the current date of the query)

  • Forecast Point Elevation: 2075m (or the relevant elevation you specify)

  • Forecast Day's Date: 4/22 (the date for the forecast)

  • Source Abbreviation: MB for MeteoBlue (or another abbreviation based on your screenshot’s source)

  • Temperature Range: 17–31 F (the forecasted low and high temperatures)

  • Wind Speed and Direction: W11mph (West at 11 miles per hour)

  • Precipitation: 0in (total expected precipitation, including type like snow, rain, etc.)

  • Sun Hours: 9h sun (number of hours with sunlight)

  • Predictability: 4/4 pred (forecast confidence on a scale of 4)

  1. Verify the Output
    Once ChatGPT formats the output, verify that all the fields are present and accurate based on the original screenshot. If any information is missing or unclear, provide additional context to ensure the forecast is accurate.

Interestingly, ChatGPT offered to extrapolate to fill in the gaps in what the given forecast product did not provide. This suggests that the AI tool could be used to make further advice or suggestions. I guess I’m not ready for that yet… All I want is the raw forecast information… for now…

Jediah Porter