How hyperlocal forecast differs from generic forecasts

Have you ever heard any of these popular sayings?

“Red sky at night, sailors delight. Red sky in morning, sailors take warning.”, “The higher the clouds, the finer the weather.”, “Rainbow in the morning gives you fair warning.”, “Ring around the moon? Rain real soon.”, “Rain foretold, long last. Short notice, soon will pass.”

Humanity has long sought to observe and predict the behavior of nature. Popular sayings are evidence of this need that has lasted through generations.

The great importance of having reliable weather forecasts, whether to plan planting and harvest, issue disaster risk alerts or plan navigation and port operations, has led us – with advances in technology – to increasingly sophisticated forecasting systems.

A lot is used forecasts in the most diverse situations, but what are they and how do sea and weather forecasts work?

What are sea and weather forecasts?

Current weather forecasting systems are powerful numerical models, a kind of program that operates on supercomputers, which generate an approximate mathematical representation of reality.

These models use data on air temperature, rainfall, winds, waves, among others, to solve complex mathematical equations that describe the behavior of ocean and atmosphere phenomena and the interaction between them to predict what will happen in the coming days.

It takes a lot of science, technology infrastructure and computing power to get the right data and information to generate a forecast. But how does this all happen?

How weather and ocean forecasting is done?

1. The initial condition

Forecasting starts with observing the current state of the atmosphere and ocean.

Sensors on land, at sea, in the air and in orbit in space - such as satellite, weather stations, ocean buoys, weather balloons, flow meters, aircraft and radar - measure a range of weather conditions to get a complete picture of current environmental conditions. and establish the initial condition of the model.

The quality of the forecast is totally dependent on the weather condition in which the model is started, being defined in meteorology as an “initial condition problem”. The greater the amount of data and the more accurate they were, the better.

2. Predicting the future

Then the already tested and calibrated models solve a large number of complex equations for various locations on the surface of the land and ocean and in different layers of the atmosphere. This allows meteorologists and oceanographers to simulate how the ocean and atmosphere are currently behaving and predict what will happen in the coming days.

A curious fact is that atmospheric and oceanic phenomena are chaotic in nature, so small deviations in the first hours of the forecast can result in large errors days ahead. That's why today's forecast is much more accurate than tomorrow's forecast, and the more days ahead, the more difficult.

3. Global forecasts

The results of these models are analyzed and performed, and generate global sea and weather forecasts that aim to represent oceanic and atmospheric events around the entire globe.

4. The use of forecasts

Global forecasts can be broadcast directly or used as a basis to generate higher resolution regional or local forecasts.

Once the forecast is complete, effectively communicating the forecast message and understanding the impacts of weather events becomes as important as the details of the forecast itself.

Where are global weather forecasts made?

All this structure necessary to generate forecasts on a global scale – which goes through satellites and other sensors distributed across the planet, supercomputers and a large team – is only possible for a few large government institutions around the globe. The most notable, with widely used products, is NOAA in the United States. In Brazil, we have the CPTEC/INPE.

It is institutions like these that generate publicly available global climate models, which are re-analyzed and re-worked by a large network of climate predictions around the world in order to generate results aimed at different applications.

While all of these models are based on the same physics, some models may prioritize certain types of settings or data – such as wind speed, temperature, and humidity – over others, generating predictions slightly differently than another model.

This is why two models can produce slightly different results, even with exactly the same initial observations.

We have seen this in practice when we have accessed different free forecasting sites that use different global models and we have noticed some differences between the forecasts. And, as a rule, we are left with the question: which will be more accurate?

How does forecast resolution work?

Climate models divide a region, be it a state, a country or even the entire globe, into a set of cells – as if they were the pixels of a photo. The size of these cells reflects the resolution of the model and influences prediction usage and accuracy.

Large cells mean low resolution or an inability to tell what is happening in small areas, but a comprehensive view of broader climate trends. This forecast is useful to know if there is a storm approaching the coast of Brazil, for example, but it is insufficient if you need to know what will happen on a specific stretch of coast.

On the other hand, the smaller the cells, the higher the resolution. Higher resolution models can predict phenomena in more detail at specific locations.

Depending on the territory in which the model is implemented, it can be classified as global, from 8 km to 40 km, regional, with cells from 3 km to 6 km, local, with cells between 500 m and 3 km, and hyperlocal, reaching just a few tens of meters. It's these ultra-high resolution predictions that we're going to talk more about now.

What are hyperlocal forecasts?

As we have seen, global forecasts are not adequate to represent specific stretches of coastal areas. Similar to a photo with very large pixels, image detail is lost.

Global forecasts have a resolution of around 25 km, while hyperlocal forecasts can reach resolutions of a few meters, offering much more detailed and accurate information.

To generate hyperlocal forecasts it is necessary to refine the global forecast to smaller and smaller scales, through new specific numerical models and data of local conditions. This means that the information of a storm approaching the coast of Brazil provided by a global forecast can be refined to know how the waves of that storm will impact your port.

This is the magic of the technology we develop at i4sea!

Why are i4cast® predictions better?

At i4sea, we analyze all publicly available global climate models and compare them to obtain the most likely to come true result of all these predictions.

We then apply our own high-resolution model to tailor the forecast to our customers' needs. In this process, we increase forecasting assertiveness and are able to predict the impact and risk of sea and weather conditions for your business.

In the i4cast® hyperlocal model in operation in the Porto do Açu region, for example, the distance between the forecast points can reach 10 m, while the global forecast points closest to the ship maneuvering area are around 40 km away – corresponding to 400 football fields.

The large number of points in the hyperlocal model allows an adequate representation of the Port features, navigation channels and small bathymetric variations – which the global model is not able to represent.

The consequence of this higher resolution is a significant improvement in forecasts. In real results for Porto do Açu, we achieved:

i. Wind forecasts 3 times more assertive

ii. 5x more assertive wave forecasts

iii. 100% increase in assertiveness of port operations.

You can learn more about how Porto do Açu anticipates maneuver restrictions by clicking here.

Another interesting example of the power of hyperlocal forecasts is the wave forecast for the Port of Sines in Portugal.

The i4cast® wave models are capable of predicting, with details, the impact of waves at a berth level. Therefore it has the ability to support decision making on the efficiency and safety of Port Terminals’ operations.

Stopping here would already be a sure win for i4cast® for those who need high quality forecasts. But there's still more:

As the last step of our processing, we use artificial intelligence in our favor.

Machine learning algorithms further increase the accuracy of our predictions up to 24 hours into the future. Which generates an observed increase in assertiveness by up to 40%. You can learn more about how machine learning can improve (a lot) sea and weather forecasts by clicking here.

Finally, to close with a flourish and value the knowledge of the ancients, remember: “Clean the gutters while the weather is right” (Chinese proverb)

What are you waiting for to make proactive decisions?