Weather Forecasting

I

Introduction

Weather Forecasting, science of determining in advance changes in the circulation of the atmosphere, and the weather these bring to particular areas. Vast regions of the world do not have a variable pattern of rain, sunshine, wind, and showers. Over deserts there is rarely any rain, and over most tropical oceans the so-called trade winds change little day by day. Nevertheless the weather in all parts of the globe is related through the general circulation of the atmosphere, and weather forecasting beyond the next day or two requires a global perspective.

Weather influences almost every human activity. It dictates the clothes we wear, the houses we build, the routes aircraft fly, and when best to sow seeds, and spray or harvest crops; it influences demand for energy. It is life-threatening; few sailors or mountaineers venture forth without the latest weather forecast. Increased accuracy over recent years has undoubtedly saved many lives.

II

What Causes Weather

As the Earth orbits the Sun, the tilt of its axis causes the summer hemisphere to receive much more solar energy than the winter hemisphere. Also equatorial regions, with the Sun always almost overhead, receive more heat than elsewhere. The general circulation of the atmosphere redistributes this uneven heat by transporting warmth towards the poles and returning colder air currents. Without this circulation equatorial regions would become so hot that water would boil, whereas Arctic areas would be far colder than they are today. Over the world as a whole, solar heating is balanced by radiation from the Earth back into space; this re-radiation is much more evenly spread around the world. Its effect is felt most keenly when, on calm, clear nights, temperatures plummet.

While the main driving force of weather is solar radiation, a second is energy from rotation. The Earth rotates once every 24 hours on its axis through the poles, although this goes unnoticed except for the apparent progression of the Sun, Moon, and stars across the sky. This rotation means that the surface of the Earth speeds eastward, much faster near the equator than at high latitudes. As warm air rises in equatorial regions and flows polewards, its momentum gives rise to strong, largely westerly winds high in the atmosphere. These often narrow jet streams loop in varying patterns around each hemisphere, causing the development and decay of large weather systems such as depressions and anticyclones. It follows that any forecasting system extending beyond one or two days must include the energy transfers that generate these winds high in the atmosphere.

Weather forecasting hinges upon knowing how weather systems will develop and move. The first essential requirement is to find out what is happening now. To forecast for an hour or two, information from the local area may often be sufficient. Attempts to forecast beyond two days or so require global observations. Weather observations are made on land and at sea, at the surface and in the upper atmosphere. Many are direct measurements from conventional instruments, but increasing numbers are made remotely by radar and satellite. Information is collected from all nations of the world, checked, plotted on charts, and stored in computers.

III

Traditional Weather Forecasting

Until the 1960s and the advent of computers, forecasts were produced manually. Forecasters would analyse the positions and intensities of weather systems from observations plotted on charts. Their movement and development was predicted in light of the average upper flow across them, taking account of likely changes in the upper winds themselves. Intensities would be modified by reference to whether the pattern of high-level winds caused a net removal of air—which leads to a fall in surface pressure—or vice versa. Procedures were largely graphical and qualitative; it was particularly difficult to judge where new systems would form. Forecasts up to about 24 hours were usefully accurate; beyond that quality deteriorated rapidly. Forecasts for one day ahead in the 1960s were less accurate than three-day forecasts are now.

IV

Modern Weather Forecasting

It has long been recognized that the only reliable method of producing useful weather forecasts for more than a day ahead is numerical weather prediction, or NWP. The basis of NWP is the set of mathematical equations that govern the behaviour of the atmosphere. These are combined in a complex mathematical model, and this is applied to observations of the real atmosphere. The first attempt at NWP was carried out by Lewis Fry Richardson in 1922. He was unsuccessful because of insufficient data and computing power, but showed it to be possible. The first experimental forecast to be completed was at Princeton University in 1950, involving a simplified set of equations over a model atmosphere with only one level. That 24-hour forecast took one day to compute. Subsequent improvements in the mathematical formulation of the equations and vast increases in computer power have established NWP as the foundation of weather forecasting worldwide.

The laws of physics and the mathematical equations governing the motion of fluids have been well known for more than a century. They incorporate the principles of conservation of momentum, mass, energy, and water, and include laws of motion applied to a fluid on a rotating sphere as well as the laws of thermodynamics, radiation, and gases. The Earth's size, rotation rate, geography, and topography are known, as are daily and seasonal variations of incoming solar radiation. Other factors include surface reflectivity (albedo), melting, evaporation, cloud, rain, friction, and sea temperatures. Many of these factors vary through the period of a forecast and must be updated accordingly.

The complex set of equations cannot be solved directly over the whole atmosphere. They are adapted to operate on the atmosphere at individual points, each representing an area of the Earth's surface. The model is applied to a large array of points, laid out as a grid in the model atmosphere. Each point includes several levels up through the atmosphere, and can be regarded as a “stack” of “parcels” of air, each of which represents a particular level over the area of a grid square.

The British Meteorological Office's Global Model is one of the most powerful current NWP models. Its grid comprises 288 points on each of 217 circles of latitude, with a stack of 19 levels at each. Thus the set of equations must be solved for well over a million “parcels” of air to advance the model a step in time. Every forecast starts with a “first guess” of the initial state of the atmosphere. This is based on a short-period forecast from a previous model run, adjusted by thousands of observations from around the world. Advancing the model in time can proceed only in short steps of ten minutes or so, because changes at each “parcel” influence its neighbours. The “time step” is repeated until the required forecast period is covered. A 24-hour forecast involves more than a trillion calculations, and currently takes about 5 minutes. Major NWP systems are continually being refined as understanding of the atmosphere improves, computing power increases, and mathematical techniques advance.

The grid spacing or horizontal resolution of the British model averages about 100 km (62 mi). This is important because it dictates the minimum size of atmospheric disturbance the model can be expected to forecast. Even the highest resolution model cannot be expected to predict a shower or thunderstorm with complete accuracy, but it should give a good indication of areas in which they might develop. Vertical model resolution is also important, because there are often important variations in wind and humidity over depths less than 1 km (.62 mi), especially near the Earth's surface and high in the atmosphere. For this reason model levels are unevenly spaced, being clustered at the top and bottom of the troposphere.

For greater detail over a smaller area of interest, it is possible to nest a higher resolution model within a Global Model. This avoids the extra computing needed with thousands of extra points over the whole globe. The British Model has a system rather like a Russian doll: the Global Model contains another with 50-km (30-mi) spacing which spans Europe and the North Atlantic, and that has within it a model with 15-km (10-mi) resolution over the British Isles.

There remains an important role for the forecasters. They must allow for weaknesses in the model, take account of later information, and use experience to add detail and value.