In its basic sense, weather forecasting involves predicting how the current state of the atmosphere will have changed after a certain period of time. This could be done on a local city level, or even trended out to a year or more. It could be done for the individual, or for a shipping lane or flight path. Ironically, one of the most difficult aspects of forecasting is understanding exactly what is going on "right now". This is much more than looking out the window and seeing if the sun is shining. All forecasts depend on an accurate assessment of the current conditions. This is why painstaking efforts are taken to get as much weather data as possible.
    There are more than 10,000 land-based weather recording stations that report weather information each hour, all over the world. This is a ton of information, but often its just not enough to see exactly what's going on because there are areas between the data points that aren't reported. For example, here in West Virginia, there are hourly observations for Morgantown, Clarksburg, and Elkins, but nothing between Elkins and Cumberland, MD. This means many weather forecasts will not take into account the weather differences from the high mountains of Tucker and Garrett counties, and therefore more than likely it'll be wrong for that area. Meteorologist need to know more than the temperature and the sky conditions to get an accurate picture of the weather. Each hour, weather forecasters feast upon brand new reports of the following surface weather data: Temperature, Dewpoint, Barometric Pressure, Wind Speed and Direction, Cloud Cover, Visibility, and the current weather conditions. This is quite a lot of information to handle, certainly when you need to track it for hundreds of locations each hour. Obviously a system was developed to handle all the incoming data visually so that many reporting stations can be analyzed simultaneously.

Surface Observation

    Each hour, meteorologists obtain data from airports, weather services, sailing ships, and the like to attempt to get the most accurate picture of the current weather conditions. Each meteorologist would then have access to all these reports for their own analysis. If all is done correctly, each weather person would have the same picture of what's going on (minus the localized variations known only to the home-based meteorologist). The following key helps the meteorologist graphically display the weather conditions for all to use:

The following is the surface recording key from JetStream, a National Weather Service Educational Site.

	In the upper left, the temperature is plotted in Fahrenheit. In this example, the temperature is 77°F.
	Along the center, the cloud types are indicated. These cloud types use the same cloud codes as found in the cloud chart section. The top symbol is the high-level cloud type followed by the mid-level cloud type. The lowest symbol represents low-level cloud over a number which tells the height of the base of that cloud (in hundreds of feet) In this example, the high level cloud is Cirrus, the mid-level cloud is Altocumulus and the low-level clouds is a cumulonimbus with a base height of 2000 feet. [more on these symbols]
	At the upper right is the atmospheric pressure reduced to mean sea level in millibars (mb) to the nearest tenth with the leading 9 or 10 omitted. In this case the pressure would be 999.8 mb. If the pressure was plotted as 024 it would be 1002.4 mb. When trying to determine whether to add a 9 or 10 use the number that will give you a value closest to 1000 mb.
	On the second row, the far left number is the visibility in miles. In this example, the visibility is 5 miles.
	Next to the visibility is the present weather symbol. There 95 symbols which represent the weather that is either presently occurring or has ended within the previous hour. In this example, a light rain shower was occurring at the time of the observation. [See all the symbols]
	The circle symbol in the center represents the amount of total cloud cover reported in eighths. This cloud cover includes all low, middle, and high level clouds. In this example, 7/8th of the sky was covered with clouds. [see the complete list of symbols]
	This number and symbol tell how much the pressure has changed (in tenths of millibars) in the past three hours and the trend in the change of the pressure during that same period. In this example, the pressure was steady then fell (lowered) becoming 0.3 millibars LOWER than it was three hours ago. [see all pressure tendency symbols]
	These lines indicate wind direction and speed rounded to the nearest 5 knots. The longest line, extending from the sky cover plot, points in the direction that the wind is blowing FROM. Thus, in this case, the wind is blowing FROM the southwest. The shorter lines, called barbs, indicate the wind speed in knots (kt). The speed of the wind is determined by the barbs. Each long barb represents 10 kt with short barbs representing 5 kt. In this example, the station plot contains two long barbs so the wind speed is 20 kt, or about 24 mph. [more about wind barbs]
	The 71 at the lower left is the dewpoint temperature. The dewpoint temperature is the temperature the air would have to cool to become saturated, or in other words reach a relative humidity of 100%.
	The lower right area is reserved for the past weather, which is the most significant weather that has occurred within the past six hours excluding the most recent hour. [see the complete past weather symbols]

    As you can probably imagine, maps with this sort of data on them would get a little cluttered. The image on the left is the current surface observation chart, with its hundreds of observation points. The trained meteorologist can pretty much dissect this map in minutes, but another set will come out in an hour, so often a weather forecaster is behind the clock in being able to not only analyze the surface conditions, but build a forecast. That is why computers have been so helpful in recent times to not only plot the data, but analyze it for the meteorologist. The picture at left is contouring surface pressure. It is important for this to be calibrated to sea level, because since many observation sites are located at various altitudes above sea level, and since air pressure drops quickly with height, not putting all the observations on the same plane would prevent an equal plane for comparison.

    Computers now do the work not only of recording and collecting the data, but also plotting it and contouring it for the meteorologist. A forecaster would usually need to see the following maps, either overlaid or separate, to see how different weather elements have evolved since the previous hour:
 

    There are other elements that can be contoured, pretty much anything that can be compared from one station to another. Meteorologist would analyze these to get a good feel of how the weather has been changing, so that discovered patterns can aid in a good forecast.

Upper Air Observation

    Not only is it vital to get good information on what's going on at the surface, but it's also essential to know what's going on all the way up the troposphere. Clouds are forming, the jet stream is blowing, and little twists of turbulence are moving about, advancing cold and warm air with definite effects on the surface below. It would be ridiculous to try to analyze the entire 3D atmosphere at every level with the same resolution as the surface, so there are certain height levels that have been chosen for the different things that often can be spotted at each altitude. Unlike the surface pressure observations, there is no need to calibrate them for a certain altitude, rather, the altitude itself is measured against a pressure level. Instead of assessing the barometric pressure at a surface of zero meters above sea level, instead the height at which the barometric pressure reaches a certain level is measured (this is called "Geopotential Height"). 
    How are these data points gathered? Well, many of them are done through special computer assessment today, but others are still done by what's called a Radiosonde, where a balloon will carry a measuring instrument into the atmosphere and then it would radio back the information. This would establish a vertical profile of the atmosphere with height, and show the meteorologist atmospheric properties at all levels. The information viewed here looks quite complicated, but the trained meteorologist can assess it quite easily. Each of these data points can then be contoured across the US at different height levels to ascertain the weather that is occurring there.
    At the 850mb level (altitude at which the pressure over a certain location is 850 millibars), temperatures and humidity there are good indications of weather or not snow or rain, or something in between, will be falling on a given location. The 700mb level, assessing humidity here would help indicate moisture available for strong thunderstorms. The 500mb level is a good indicator for where the energy that is driving surface storms is moving. The 300mb level would show the Jet Stream, and gives many other helpful bits of information for the meteorologist. 

Satellite Data

    When meteorologist need to look at the sky from above to determine cloud cover and movement, they enlist the use of satellites. This way it's a sampling of photographs more so than looking at an individual data plot. There is a wealth of information provided by satellites. There are two main kinds of weather satellites: Polar Orbiters and Geostationary Satellites. The polar orbiters are closer to the earth and can provide detailed information for the region that they are photographing, but doesn't stay over the same location for long and can only render about 14 pictures over a given location per day. Geostationary satellites, though much farther away (the orbit must be about 22,000 miles from Earth in order for their orbit to be in phase with the same location on the ground). Geostationary satellites, from their name, are able to stay over the same location on Earth and provide the user with continuous pictures of the evolution of clouds and other data. They can also retrieve and resend data from the radiosondes to users around the world. The US operates two Geostationary satellites, one over the East Coast and one over the West Coast. These satellites would take a picture of the atmosphere, but not necessarily as in a photograph. Rather, they would analyze a certain property of the atmosphere that could be measured empirically, and then send that back to the user transformed into picture using an Algorithm, a mathematical formula designed to transform numerous numbers into a tangible display. There are several types of algorithms currently used to display data retrieved by satellites.

	Visible Satellite Imagery - The satellite would in this case be measuring reflectivity of sunlight from clouds. This gives a good indication of the thick clouds (those that reflect much sunlight) and the thin clouds (ones that sunlight can travel right through). Since the satellites often view the Earth from an angle, vertical development can also be seen. The visible satellite imagery is poor at night, when lack of sunlight equals lack of reflectivity.
	Infrared Satellite Imagery - The satellite here would be analyzing the temperature returns from the Earth. Clouds would reflect back much cooler temperatures than the surface, and can obviously be quantified with an algorithm. The high clouds are usually the coldest, and the low clouds are the warmest. Sometimes in a fog situation, the temperature of the ground and that of fog blends in, and thus cannot be observed accurately with an infrared picture. The infrared satellite image (IR) can be colored or enhanced to contour the different temperatures, and often giving the user a good picture of vertical development of a storm system below. The IR satellite also has a poor time distinguishing between thick and thin clouds, as cold cirrus will return the same temperature as the top of a cumulonimbus cloud.
	Water Vapor Imagery - Satellites can also examine certain wavelengths of light that are only absorbed by water existing in the gas state. Since this involves absorption, it can only see the top third or so of the atmosphere. Nevertheless, it is a great way to understand how moisture is moving in the atmosphere regardless if it's already developed into clouds.
	Derived Satellite Images - Satellites can also look at the clear sky onto the ground and ascertain the vertical profile of several atmospheric quantities, and send them back to the user. This can help fill in the holes of radiosondes that are set out hundreds of miles apart. This comes in quite handy during the severe weather season, but are not usually transferred onto a television set for the untrained viewer due to the large amount of information.

Radar Data

    There are also some ground-based observational tools the meteorologist uses to analyze the clouds above. Radar can not only be used to detect aircraft and other man made objects, but can also examine precipitation as it falls. It used to be that a radar could only assess reflectivity, by sending out a pulse of energy and scoring the return using an algorithm for distance and strength of the return. New radars, called Doppler Radar, can now also analyze the speed at which measured precipitation is moving towards or away from the radar location (this is called a "Phase Shift". The radar sends out pulses travel nearly the speed of light and basically spends most of its time listening for the returns. A Doppler radar looks different from your normal disc-shaped radar; it resembles more of a soccerball. Currently in use are conventional radars that can assess precipitation, and also the Dopplers that can put many extra dimensions to a precipitation return. Though the radars can return data to the user anytime it hears a return, usually it waits about 5 minutes to get an entire picture of the area or a sweep. Precipitation with the best reflectivity scores the highest returns (measured in decibels or dBZ). This would be wet hailstones and the like, through sleet and freezing rain, and then rain, all the way down to snow which shows up faintly on a radar due to its poor reflective properties.

    Live Doppler imagery is quite useful during severe weather outbreaks, as it can constantly channel detailed information about the immediate area for the meteorologist. Not only can it track precipitation amounts as they fall from the sky, but the ability to calculate storm-relative velocities from the phase shifts of falling precipitation has become a virtual lifesaver. Severe thunderstorms capable of producing tornadoes will have a rotation in their core that can be detected by Doppler radar by analyzing the wind patterns inside the storm. This can permit a weather service to issue a Tornado Warning in advance of one forming whereas prior they'd be left to waiting until one formed to get the word out.

Computer Modeling

    Not only has the improvement in computer technology lead to better analysis of current weather date, but complex mathematical formulas can be written for the computer to extrapolate that data out over time to create a picture of the atmosphere a certain time interval later. This would represent a computer model of the forecast. These are extremely advanced mathematical formulas that are resolving many interdependent variables over time across a resolved terrain field. In fact, computers are still unable to process the information quick enough and to a resolution high enough to satisfy the current envisioned capability (but it's catching up quickly). A computer model rendered forecast is called a solution. It gives a meteorologist a good idea of how the weather would be a certain time in the future using atmospheric physics. Though tremendous strides have been made in ensuring the accuracy of these formulas, ironically the initial conditions state (the data entered into the computer at time = 0 point from which all solutions are processed can never be perfect). There is always some hill or mountain that isn't resolved properly on the surface (understand that everything has to be contoured mathematically), and there are always holes between the data points the computer ingests. With each passing advancement in computer modeling, come more capabilities to increase the resolution of the terrain and multiply the number of observations the model can begin with. A trick meteorologists have employed recently is to vary slightly the starting conditions for a computer model, and run several different solutions together on the same map. This is what is called an Ensemble Forecast. It gives the forecaster an idea of how the weather could look given slightly different starting conditions, and thus a broader look at the forecast in general.

    Regardless to the advances of the computer models, without a human perspective into the computer models' solutions the forecasts could sometimes get incredibly inaccurate. Data is sometimes ingested poorly, whereby a 30 knot wind would be scored 300 knots, throwing a huge wrench into the system. Moreover, several different models have been created with slight variations in their formulas and their terrain renderings, resulting in slightly different forecasts each time. Computer models have Biases as well, some that like to put more precipitation than necessary into a solution, and others that are consistently too warm. Without the meteorologist the forecast would at times be doomed. The future of forecasting is always to continually enhance the model, but also for the meteorologist to discover more about the atmosphere and the processes that put down the rain and snow each year.