A Weather Forecast Specific to your Back Yard?

When one looks at a weather forecast on their mobile device, they don’t assume that the forecast they are reading is for their specific location (such as their house) – it’s based on synoptic data gathered for an entire region. However, one company says they’re changing the game – through the application of machine learning and historic data, they say that they can produce accurate, on-spot weather forecasting, even down to the individual wind farm turbine.

Meet Meteo-Logic, a company that is bringing the big data approach to the weather modeling industry, aiming at industrial spot applications where weather is a critical component to field operations. Think about farmers having weather forecasts for specific fields with specific crops; Or ski resorts that can forecast weather on specific runs. Anywhere that weather is a factor in business, Meteo-Logic says they can measure and predict it.

First the system is fed with several years of historic synoptic data for that area. Using Bayesian-based machine learning algorithms, the data forms the basis of a weather model that is then linked to measured values for the specific spot in question. As the system collects data, it’s able to look for patterns to match with past data, and determine how the weather will react based on the patterns found in the data.

“According to the bulletin of American Meteorological society, weather variations excluding extreme events and disasters, cost US alone $485 billion a year” said Igal Zivoni, Founder and CEO of Meteo-Logic “Much of these costs could have been saved if there was an effective way to accurately predict weather and weather dependent factors, on spot, on time. Now there is”.

According to Dr. Baruch Ziv, a senior advisor to Meteo-Logic who served previously for 20 years as a meteorological officer for the Israeli Air Force, there are three widely used methods used to make weather forecasts:

• What was is what will be – The forecaster examines what the weather was yesterday and assumes that today’s weather will be similar. Statistically, it works fairly well, but the chances of success drop during periods of variable weather and it’s clear that it’s an unacceptable method for the discerning consumer.
• Use of models– This method takes data from a widely used international model, the American GFS model. This type of model offers calculated temperature values for specific locations that are dozens of kilometers equidistant from one another. The disadvantage of this model is that there is no data for areas between the points, and to make a forecast for a specific spot one needs to make weighted calculations using several different points. This method misses temperature jumps in areas where weather conditions change sharply. There are times when it’s cooler along the shore because of winds blowing from the sea, while the country’s internal regions are suffering heat-wave conditions.
• Using sophisticated models – There are some very advanced models, in which the distance between the reporting points is small, from one to three kilometers apart. But, this method is very expensive and takes a lot of time to get up and running.

Meteo-Logic offers a fourth method combining historic data with number crunching analytics that includes location specific data. Through this synthesis, they say that they’ve been able to reduce the standard deviation between the Meteo-Logic and the American GFS model by half. “This is a very significant improvement,” said Dr. Ziv.

Last week, the company announced that it has raised $3 million dollars in Series A funding from Hong Kong-based, Horizons Ventures, which manage the technology investments of one of the most prolific investors in the big data arena, Mr. Li Ka-Shing. Recently, the knighted “Sir Ka-Shing Li” was seen dropping money bags on the University of Oxford to fund the creation of the Big Data Institute at the Li Ka Shing Center for Health Information and Discovery.

With an investor on board, the company is going after specific industries as they work to monetize their system: the agriculture industry, the solar power industry, and the wind farm industry, where the company has recently launched a wind mill power forecasting solution that delivers accurate predictions for individual turbine power production for up to 5 days in advance. 

It will be interesting to see how well the solution gets adopted, not to mention how flooded the weather modeling space could eventually get. While the credentials that Meteo-Logical has are impressive, there are some whales in this ocean. Recently IBM released a predictive analytics package aimed at wind farms of their own.

In the meantime, we’ll use our sensors to keep an eye on the developments in this space and look forward to the day when we’ve all got smart sensors on our decks to tell us what to expect for the weekend BBQ.

Related items:

Predictive Analytics Prevailing on the Wind Farm 

The Five Types of Hadoop Data 

Huawei, CCTV Partner to Dev on High Density ‘Big Data’ Storage