Introduction

This article will explain everything Australian fisherman need to know about weather including how forecasts are made, why some forecasts are better than others and important weather concepts. Along the way we will bust common myths associated with weather and allow Australian fishermen to be fully informed.

When people talk about ‘the weather’ they are generally talking about two things – Observations and Forecasts. Observations are measurements of weather taken by weather stations. In Australia there are currently 870 weather stations throughout the country. These stations are fitted with instruments that enable them to measure wind, rain, temperature and humidity. They are managed by the Bureau of Meteorology (BoM) and provide a network of current conditions around the country. Some stations record all weather variables while others record only some. Similarly, some stations record values every 10 minutes while others 30 minutes and some only once or twice a day. BoM make data from these stations readily available so when weather sites report on current conditions or when the news reports maximum and minimum temperatures recorded for the day they are using data from these stations. To see where these weather stations are located throughout the country you can look at the following links for each state.

VIC - http://www.bom.gov.au/vic/observations/map.shtml

NSW - http://www.bom.gov.au/nsw/observations/map.shtml

WA - http://www.bom.gov.au/wa/observations/map.shtml

TAS - http://www.bom.gov.au/tas/observations/map.shtml

QLD - http://www.bom.gov.au/qld/observations/map.shtml

SA - http://www.bom.gov.au/sa/observations/map.shtml

NT - http://www.bom.gov.au/nt/observations/map.shtml

Forecasts are predictions of weather in future. They are made by computers that run mathematical models which simulate changing conditions over time. Data from these models is used to drive tables, charts and weather maps, or can be summarised into text forecasts which provide an overview of expected weather conditions for a region.

In order for a forecast model to predict weather ahead it first needs to know what the weather is now. Weather data from observation stations and other sources is collected and fed into the model as ‘initial conditions’. These effectively provide a snapshot of what the atmosphere looks like at the current point in time. The weather model then works out how weather will evolve over time.

Imagine a series of dots laid out in a grid across Australia. Forecast models use laws of thermodynamics and other laws to calculate how weather at one point will interact with weather at the next point and by repeating this process over and over they can work out what the weather will be at each point in the grid.

Once they get to the end they go back to the start and do it again for the next time step and so on until there is a forecast for each point in the grid and for every time step. The sum of time steps is the duration for which the forecast is valid, ie a 7 day forecast. Time steps are usually made in 1 hour or 3 hour intervals with the smaller interval being better as it provides more data. At each iteration, the model uses expected values it just calculated as initial conditions for the next time step so it’s easy to se e how small deviations in expected values at the start of the models run manifest into larger deviations at the end of the forecast period. For this reason, forecasts are most accurate close to the present time and less accurate the further they move away from the present time.

Forecast data files are often called ‘grid files’ because they contain a weather forecast for each point in the grid for the defined region.

This is an important one so take note. The ‘Resolution’ of a forecast refers to the spacing between dots in the forecast grid. The smaller the spacing, the closer the points are together and the higher the forecast resolution. Higher resolution forecasts are better guesses at what the weather will be than lower resolution models. These days high resolution generally means grids with less than 10km spacing between points.

You can think about this like a television set. Old box style televisions had fewer dots per area than modern high definition televisions which pack many more dots into the same space. Just like more dots make for a clearer picture in television, more dots also make for a better forecast in weather prediction.

In the below images, the image on the left shows a forecast grid overlaid on Australia. The image on the right shows a forecast grid overlaid on Australia where the dots are twice as close together, resulting in a forecast with twice the resolution.

In general, the best forecast model is one that has the highest resolution and the smallest time step because it contains the most amount of data for a given area. Even modern super computers are challenged to process the quantities of data required for weather forecasting hence the need for trade-offs between resolution, size of area being forecast and time step.

Since weather is a global phenomenon you can’t isolate a region like Australia and try to forecast weather there without first understanding what weather is doing in the rest of the world.

All forecast models start by predicting weather across the entire globe first. Global models are run at relatively low resolutions and these then provide a starting point to run a regional model. Once the global forecast is complete, meteorologists tell the computer to focus on a region such as Australia and increase the resolution. This then allows the computer to refine the global forecast to a high resolution and more precise forecast for the target region. If computers were much more powerful then we could run global forecast models at 10 meter resolution and that would be the end of it, we would have accurate weather predictions for every spot on earth in one go. The reality is that computers are not yet that powerful so we have to compromise.

Despite the best forecast models and powerful computers that run them, what comes out is essentially a computer-guess of what the weather will be. Refining this to a better guess is only achieved when meteorologists with experience of local weather systems review and make adjustments to the data which results in a ‘quality controlled’ forecast. You can look at this process as follows; the global forecast is a first guess, the regional forecast a second guess and the manual adjustment a third and best guess. The Australian Bureau of Meteorology do this for Australian forecast grids, a meteorologist in each state looks at the data then adjusts it based on their experience with local weather systems.

Data that come from agencies like BOM contains a unique forecast for each point in the grid, ie an hourly forecast out to 7 days. Imagine you had a wide open space with 4 posts arranged in a square. If you want a forecast exactly where one of the posts is then it’s a simple case of reading the data straight out of the file, no calculations are required. If however you want a forecast for another point in the square that isn’t right on a post then what do you do? You have two options, you can either select the forecast for the closest post and say that’s close enough, or you can try and calculate what it would be at the exact location in question. This is where the ‘processing’ aspect of websites comes in to play.

The red star represents the location where we want a forecast. We can either take the forecast for the nearest grid point or try and work out exactly what it would be where the star is.

If you are using a high resolution forecast grid like ADFD 3km (more on this later) then it doesn’t matter so much as there is unlikely to be much perceivable difference in weather within a 3km square. If a site was using lower resolution forecast models with 27, 55 or 70km grids, which were common just a few years ago, then it is more important. The process of coming up with a forecast within the square is called ‘interpolation’. The most common method to handle this situation is simply to look at surrounding grid points and half the difference. Ie if you are in the middle of two grid points and one says 10 knots and the other says 12 knots then you would say its going to be 11 knots in the middle. If you are closer to the first than the second then it may work out to be 10.5.

Interpolation does get a bit more interesting when a low resolution forecast model is used and you want a forecast right on the coastline. Soon we will look at sea breezes and see how different methods can be used to interpolate winds on the coast.

Text forecasts are verbal summaries of conditions that can be expected for a region. For example:

WINDS: Northwest to northeasterly 10 to 15 knots tending southeast to southwesterly in the early afternoon then tending north to northwesterly in the evening.

SWELL: Southwesterly 1.5 metres, increasing to 1.5 to 2 metres during the morning.

These forecasts are made by meteorologists who look at forecast grids and try to summarise expected outcomes for a given area for a given day. They provide a convenient way for people to understand a forecast verbally, for example over the radio or from one person to another.

The text forecast is for a weather region as defined by the Bureau of Meteorology. One drawback with these forecasts is that the regions are often large and although there is variation within the region the text forecast provides an average overall. Significant variation can often be seen in coastal areas since marine regions extend 60nm (~110km) offshore whereas most recreational activities are close in to shore. Wind, wave and swell are often stronger away from land and off the continental shelf so text forecasts will often say wind, wave and swell are greater than recreational anglers may experience close in to land.

Sea Breezes are caused by the uneven heating of sea and land masses by solar radiation. As the sun rises in the sky, temperatures on land get hotter but temperatures over water don’t change much. This causes air to rise over land which sucks in air from over the ocean and creates an air current. This air current forms a cyclic motion that picks up cool air from the ocean and brings it inland to replace the rising hot air. Seabreezes are strongest right on the coast line as this is where the cyclic air currents have maximum flow.

Given the difference in temperatures increases as the day goes on, seabreezes normally pick up in the afternoon and continue into the evening until the temperature on land has cooled.

When a low resolution forecast model is used in weather prediction much of the seabreeze air current is not detectable as the forecast points are too far apart. To compensate for this, calculations are made to try and guess what the wind would be at the waters edge. This is often done for users like kite surfers who want to know wind speeds where they surf.

To understand this better, think back to earlier when we described how forecasts are made, now imagine we are using a forecast model with 27km resolution that has a grid point approx. 13km out to sea and another 13km inland. While the model calculates wind speed at these forecast points with some accuracy, since they are far away from the coast they do not pick up all of the cyclic air flows occurring above the waters edge. Some sites therefor apply a calculation which attempts to predict what wind strength would be at that point. Sites that do this typically have few locations registered as they each require a specific set of calculations exactly tuned to a point spot on the coast. Overall, this process is a method of interpolating a forecast where forecast grid points straddle a land / water boundary.

Seabreeze compensation calculations are largely unnecessary these days as high resolution forecast models accurately calculate seabreezes natively since grid points are much closer together. It’s only sites that use low resolution forecast models who continue to use this method. What’s more, compensation calculations that calculate seabreezes are only intended for a particular spot and away from that spot they are not valid.

Wind blows faster over the ocean than over land because there is less surface friction. Land has mountains, coastal barriers, trees, human made structures, and sediments that cause a resistance to the wind flow. Oceans do not have these impediments therefor wind can blow at a greater velocity. When wind blows parallel to the shoreline this effect is most noticeable as the wind has had a longer time to pick up speed (over water) or slow down (over land).

For this reason it’s important to check wind forecasts over open water where you intend to travel as models take into consideration surface resistance when estimating wind speed. Even being a small distance out from shore can experience significantly different wind speeds.

Wind travels faster over water than over land because there is reduced surface friction. When wind blows parallel to the shoreline there can be significant differences in wind speed only a short distance from the shoreline.

Over the years fishermen have become accustomed to checking weather forecasts at the town in which they launch their boat rather than over the area they intend to fish. Due to the surface friction effect this is not a good approach. The further offshore a fishermen is intending to travel the greater the potential wind speed difference becomes and this is the reason many fishermen have been caught out despite thinking they did the right thing by checking the weather before they left. A weather site that provides forecasts over open water should always be used over a weather site that provides land based forecasts if the user is intending to venture out from shore.

As the angle of wind changes and wind starts blowing more perpendicular to the shoreline the difference between velocity of wind over water and land begins to equalise. For winds that are directly on or offshore there is not much difference between land speeds and water speeds.

Myth – Overseas agencies produce forecast data for Australia as well as the BoM.

Verdict - True. There are essentially 3 agencies that produce forecast data that covers Australia. These are:

1. Bureau of Meteorology (BoM) in Australia
2. National Oceanic & Atmospheric Administration (NOAA) in USA
3. European Center for Medium Range Weather Forecasts (ECMWF) in UK.

While there are some other agencies that produce data that covers Australia it is generally not accessible and not commonly used by weather sites.

Verdict - False. BoM forecast data is relatively expensive so most sites don't use it. While all weather sites use BoM’s network of weather stations to get current conditions(observations) they do not all get their forecast data from BoM. Most weather sites do not use BoM forecast data because it only covers Australia, instead most sites and apps use forecast data from the NOAA in America. The NOAA produce a global forecast model called GFS which is available on the internet. This data is free and provides global coverage hence it offers an attractive data source to weather sites since they can operate a single forecast model at low cost and service the entire globe which broadens their market.

Fish Ranger provides users with both BOM and GFS wind forecasts so users can get a consensus rather than relying on only 1.

Verdict - False. It is not possible to improve a weather forecast by simply applying an equation, algorithm or formula. If it was then that equation would have been applied at the time the model was created. Models are nothing more than the output of a series of complex calculations so adding another calculation at the end to ‘make it more accurate’ would be easy to do if it was possible.

Although there are different forecast models available, each have a slightly different purpose, are open source, and have been optimised by meteorologists over many years. Differences observed in weather site forecasts are because different sites use different forecast models as inputs, they display data differently or, most commonly, because the forecast is not at the same point as the location shown. This can be because the site utilises a ‘nearest neighbour’ selection method to select a forecast from a grid or a pre-defined list rather than having an independent forecast at every location.

This essentially means when people are comparing forecasts on different sites they are not really comparing the same location, hence there can be large differences. BoM’s MetEye tool for example picks the closest point from a 6km grid. Some other sites that use the GFS27 model pick the closest point a 27km grid, so the actual forecast provided by websites is often not where users think it is. This has most impact along the coast where wind speed drops off significantly over land. Fish Ranger uses a high resolution grid and interpolates forecasts from surrounding grid points to prevent this problem occurring.

Verdict – False. Global forecast models are only produced to provide a starting point to run regional forecast models and to fill gaps between continents. Each of the countries that produce global models do not use them for weather forecasting in their own country! Did you get that? I’ll repeat it because it’s important each of the countries that produce global models do not use them for weather forecasting in their own country.

They instead use higher resolution regional models which are derived from the global. NOAA who produce the global GFS model at 9km then refine it to get the NAM (North American Mesoscale) model which covers just north America at 4km resolution. This is what the USA use for weather forecasting in their own country. Similarly, BoM produce a global model called ACCESS-G at 12km then refine it to get the ADFD which only covers Australia but has a resolution of 3km or 6km (state dependent) and is more accurate than ACCESS-G.

Its not to say that global models are inaccurate, certainly over the years their accuracy has improved significantly with increased computing power, but in general a locally refined regional model is a better forecast.

Verdict – False. As a government agency all data produced by BoM is available to registered users; they do not keep the best for themselves. For a fee users can access the data and use it any way required. In actual fact, users of Fish Ranger in Victoria and Tasmania get better forecast data than BoM provide on their own site. This is because Fish Ranger uses individual state grid files rather than the national grid which is used to drive MetEye.

The state files for VIC and TAS have 3km resolution whereas the national file has a uniform 6km across all states. Fish Ranger’s VIC and TAS users will therefor get more detailed forecasts than by using MetEye (In other states the data sets are equal). Th e difference between a 3km grid and 6km grid is a 4 fold increase in data due to their being 4 x as many grid points covering the same area. To see the forecast as the meteorologist who created it intended, you need to be using Fish Ranger and not MetEye.

At Fish Ranger’s insistence, at Jan 2017, BoM updated documentation to confirm this fact since it was previously only known to some behind the scenes people and therefor doubted by the public. You can confirm this here:

http://www.bom.gov.au/weather-services/about/forecasts/australian-digital-forecast-database.shtml

For anyone curious, a list of data services offered by BoM can be found here:

http://reg.bom.gov.au/other/charges.shtml

Verdict - False. The BoM’s ADFD dataset provides hourly wind data out to 7 days. BoM’s MetEye tool which displays ADFD data only shows forecasts at 3 hour time steps despite the underlying data containing 1 hour time steps. Fish Ranger shows every hour as the model predicts allowing users to better see approaching changes.

In addition Fish Ranger use another BOM forecast model to provide forecasts between 8 and 10 days, providing a full 0 to 10 day forecast range.

Verdict - False. The BOM website is a mish mash of different tools and sections that each display their own thing. It is not consolidated on a single page, it doesn’t show variables of interest to fishermen like air pressure, nor dies it do things like overlay live readings on the forecast so you can see how it is tracking. It’s also not available as an app, trying to use its tools on a mobile is difficult. The BOM can’t be in competition with private companies so while they provide quality data, they take a back seat with showcasing that data on their own site and leave it to weather apps that do a better job of consolidating and displaying it.

Verdict - False. Averaging weather sites that use low quality data just results in a lower confidence forecast. More weight should be given to sites that use better forecast models. Imagine you wanted to know how to catch a swordfish and so wanted some adv ice on doing so. Now imagine you bumped into a group of 5 people at the pub one of who happens to be a champion game boat skipper and the other 4 weekend warriors on the snapper. Would you take advice from all 5 people and average it out or would you listen to the game boat skipper and weight his advice above others? You would listen to the game boat skipper as he is the best source of information. It’s a similar situation with weather sites.

Verdict – False. While observations can only be made where there is a weather station forecasts can be made anywhere. Forecast grids cover the entire country and a set distance out to sea. By looking at data from the 4 surrounding grid points forecasts can be made for any location required, on land or sea.

Verdict - False. Every weather model is going to get the weather wrong it’s a question of how much and how often, you need to pick the one that’s right more often and stick to it. Think of this like an amateur poker player coming up against a professional poker player. Over the course of a couple of hands the amateur player may get lucky and come out on top. If however they play for many hours the professional player will end up winning because they are a better player. Similarly, the best forecast models will get it wrong occasionally but they will get it right more often than poorer performing models so back the better model and stick to it.

Verdict - False. Satellites are used to measure current conditions such as sea surface temperature but they are not used to forecast the weather. The temperatures measured by satellite are fed into forecast models as initial conditions but they play no further role.

There are 6 commonly used forecast models that cover Australia. Four of these are for general weather and two are for wave and swell.

Produced by BoM

ADFD – ADFD stands for ‘Australian Digital Forecast Database’ and is BoM’s official and best forecast data for Australia. It provides forecasts for weather, wave and swell. The ADFD has a resolution of 3km in VIC and TAS and 6km in other states making i t the highest resolution forecast data available for Australia anywhere in the world. ADFD is quality controlled which means meteorologists in each state review the computer output and make manual adjustments based on experience with their local weather systems. The wave and swell forecasts are provided out to 60nm (approx 110km) from shore. ADFD is the BEST forecast data for Australia.

http://www.bom.gov.au/catalogue/adfdUserGuide.pdf

AUSWAVE – This is a model that provides wave and swell data. It has a resolution of 12km. It is not edited by meteorologists before being released.

http://reg.bom.gov.au/nwp/doc/auswave/data.shtml

ACCESS – This is a general weather model that provides wind, temperature, rain etc data. It has a resolution of 12km out to 10 days.

http://reg.bom.gov.au/other/access.pdf

There are three common global models that cover Australia, these are GFS, IFS and Wavewatch III.

GFS - Global Forecast System (GFS) is a forecast model produced by the National Oceanic and Atmospheric Administration (NOAA) of America. It is a global model which covers every part of the Earth and is the most commonly used weather model by weather sit es all over the world. It is available for free on the internet and is not edited by meteorologists before being released. As computing power has increased, upgrades have enabled NOAA to increase resolution of the GFS model. There are currently three versions in operation with resolutions of 27km, 13km and 9km. The 9km version is better than the others but many weather sites still use the 27km or 13km version simply because they have not upgraded their systems yet.

https://en.wikipedia.org/wiki/Global_Forecast_System

Here is the download link for GFS data if anyone is interested:
http://www.nco.ncep.noaa.gov/pmb/products/gfs/

IFS - The Integrated Forecast System (IFS) is a global model produced by the European Center for Medium-Range Weather Forecasting (ECMWF) in England. IFS is not quality controlled and exists in 3 versions, a 55km, 18km version and a 9km version. Some variables from the data set are free and others need to be paid for hence it is not used as much as the GFS model.

https://en.wikipedia.org/wiki/Integrated_Forecast_System

Wavewatch III – This model is produced by the NOAA in America and provides wave and swell data. It covers the entire globe at a resolution of 50km. It is freely available on the internet.

https://en.wikipedia.org/wiki/Wind_wave_model#WAVEWATCH