Broadcast Meteorology in an AI World

While most of us probably haven’t realized, we have all been relying on artificial intelligence (AI) for quite some time now. As AI is becoming more mainstream, some people are feeling a range of emotions from fear to excitement. In the coming months and years, AI will significantly impact the day-to-day lives of broadcast meteorologists. AI will likely increase both the amount and accuracy of weather data and assist and streamline the completion of daily broadcast responsibilities while enhancing the content created.

iCast, The Weather Company’s proprietary forecast, uses AI to blend data from multiple sources to make the best possible forecast. It collects forecast data from numerous numerical weather prediction models (ECMWF, GFS, GRAF and many others) and compares them with subsequent observations to identify and correct systematic errors in each model. It also learns the best way to combine the information from the corrected forecasts to create a final consensus forecast that is more accurate than any of the inputs. AI is used in many other phases of the forecasting process in partnership with The Weather Company’s human forecasters “over the loop,” from identifying areas of spurious radar echoes to locating frontal boundaries.

The daily responsibilities of local on-air meteorologists have significantly increased over the years. Each day they must produce content specifically tailored for TV, mobile, websites, social, radio, Connected TV and more. It’s overwhelming!

Large Language Models (LLM), a form of generative AI, have the potential to assist you in completing daily tasks much faster. For example, an LLM can provide a cohesive summary of a longer forecast you write or even generate a written forecast. How does an LLM do that? The first thing an LLM needs is data. Without data, generative AI loses its ability to be effective. If the data available to the LLM is inadequate, the LLM will fill in the gaps with information which may not be true. Data for an LLM is something you can provide with your query or could be fed into it via other avenues. For example, you could submit a long set of text (or even a weather forecast) and ask an LLM to summarize it. Second, the LLM needs a proper prompt. If you don’t ask the right question, you won’t get the right answer. If you ask an LLM to summarize a long set of text, it will likely give you a short paragraph which is pretty good. However, if we alter the request to provide the summary as five bullet points, you will receive five bullet points based on your input which could be used to generate a headlines graphic or even provide a short summary for a web article. Need an attention grabbing headline? An LLM can help with that too!

Another way AI can help is to give a voice to your work. That voice could be someone you don’t know or even a synthetic version of your voice. Text to speech has been around for a long time. You’ve even used it within some of The Weather Company’s applications like LiveWire and Max Alert Live. AI can further enhance text to speech by making a voice-over which sounds very close to the voice provided to train the AI model.

Every single day meteorologists are required to provide a single forecast video for a DMA or maybe a sub-set of regions within the DMA. Then, on websites or mobile apps, users rely on API output to provide text and icons for the weather of the day. It is up to the user of the mobile app to analyze the hourly or daily page to extract the meaningful impacts of the weather forecast.

Enabling a meteorologist to provide a meaningful video forecast to viewers is where ReelSphere helps. ReelSphere, a new Weather Company product, leverages AI capabilities to analyze the weather in your area and create videos, based on your expertise, to provide a localized forecast to everyone within your DMA. AI can even provide a synthetic voice which sounds very similar to a viewer’s favorite meteorologist. How important is the local meteorologist when it comes to ReelSphere? The local meteorologist is essential. In a recent survey of broadcast viewers, The Weather Company learned that 75% trust a forecast created and presented by their favorite local meteorologist while 20% of those surveyed trust a forecast strictly generated by artificial intelligence. If the local meteorologist approves the forecast generated by AI, the acceptance of the video jumps to 50%.

With ReelSphere analyzing the weather conditions in the DMA and generating insightful videos with a voice-over, broadcast meteorologists can spend more time enhancing the television broadcast, gathering different perspectives on the weather story and thinking of new and exciting ways to tell the best-possible weather story, thanks to AI.