Meteorology is of course a deeper science than that (for the actual scientists in the field,) but saying “we don’t know how weather works” is a gross mis-statement of what we actually do know.
Prediction was awful before Lorenz developed Chaos theory to model changes. It’s now built into computer models based on past system evolution. We feed in the current state data for the area and out comes a ‘model’ of evolving projected states. Due to the number of interactions involved, so increasing uncertainties, the accuracy of the predictions decreases exponentially with time.
Weather forecasting systems work by simulating/generating an image of the atmosphere at a priori state and then make a time integration into the future.
Weather forecast is based on rigorous, multidisciplinary sciences of the atmosphere, ocean, and land, and the interactions among them. The weather forecast that you see on TV or on your mobile app actually starts with collecting real-time data using satellites, weather radars, and ground and upper-air in situ measurements around the world. There are now over 20 weather satellites worldwide and over 150 weather radars in the US alone. For each round of data forecast, many billions of data points are integrated into a numerical weather prediction model using a process called data assimilation.
Weather forecast is an initial value problem. To produce a forecast, a model processes today’s observed weather parameters (hence the name initial value) based on millions of highly nonlinear differential equations that evolve the dynamics, thermodynamics, and radiative processes for the entire earth surface, allowing the atmosphere, ocean, land surface, and snow and ice to interact with each other. These equations describe the earth’s atmosphere at tens of layers in the vertical, the upper and lower layers of the oceans; and the entire earth at horizontal intervals of kilometers and at time steps of minutes. These models are very computation intensive and run on supercomputers.
All atmospheric calculations are based on a single formula, known since the says of Newton: (P∗V)/T=constant
where P
is pressure, V
is volume and T
is absolute temperature (temperature over absolute zero, -273 C). This formula is sometimes referred to as Boyle's law.
The accuracy of weather forecast still has much to be desired. This is because within the description above, there are several ways (not in a particular order) that errors could be introduced. (1) Weather observations are not perfect. Depending of the country, observations over land are tens to hundreds of kilometers apart. Over the ocean, they are few and far between. Satellites have improved weather observations, but they have accuracy problems of their own. (2) The equations that describe the atmosphere, ocean, and land at regular intervals over the earth also have gaps between them. (3) Over time, the highly nonlinear equations will magnify any computational errors or errors introduced into the initial value. (4) Imperfect understanding—and hence imperfect mathematical representations—of the physical processes of the atmosphere, ocean, and land in the model (e.g., cloud physics and radiation transfer).
More accurate weather forecast requires better understanding of the physical processes in the atmosphere, ocean, and land; more accurate observations (especially from satellites), and more powerful computers.
In simple terms they familiarize themselves with the past atmospheric state and because they are taught by humans how most processes behave in the atmosphere, they are able to “grow” and imitate the next state of the atmosphere. Hence weather prediction.
Weather forecast is a complex process. It starts with weather observations from ground-based, balloon-borne, and satellite sensors. Countries around the world all share their observations to provide a continuous depiction of the global weather. Weather forecast is based on weather models. Each model consists on millions of highly nonlinear equations that evolve the dynamics, thermodynamics, and radiative processes for the entire earth’s atmosphere, ocean surface, and land surface and their interactions.
To produce a forecast, a weather model evolves today’s observed weather parameters over time; hence weather modeling is often called an initial value problem. These equations describe the earth’s atmosphere (from the earth’s surface to a altitude of over 50 km), ocean surface, and land surface at spatial intervals between a few to tens of kilometers apart while the model computation marches forward in time steps of a few minutes each. A weather forecast represents these cumulative time-based changes over forecast periods of hours to days. Post-processing (often based on statistical models) then transforms model forecast into weather conditions that we want (e.g., temperature, rainfall probability, wind speeds and directions, and cloudiness). Local weather forecasters, who are familiar with weather features of their respective locations (e.g., a mountain region, a coastal zone, or a desert), will then interpret the model output to produce a forecast in plain language.
Weather is forecasted by taking measurements of various atmospheric conditions and formulating an interpretation of how these atmospheric conditions will react with each other. These measurements include how much moisture is in the air, the air temperature and the air pressure, plus others from numerous ground locations as well as vertically above those locations In addition satellite data is used to understand cloud cover and direction of air mass movements. Finally, simulations of what could happen are developed on computers that integrate data from the past and look what weather was generated by similar conditions in the past such that it can be used as a guide to what can happen currently. From this mass of data collection and simulation by computers, a forecast is developed. This forecast is an integration of the measured physical conditions of the atmosphere using the remotely electronically measured data as well as ground station data to create a physical model that predicts airmass movement and interactions. In general a forecast improves dramatically when predicting soon to happen weather events. Generally a forecast of the weather that is greater than 3 or 4 days away has a probability of being accurate of less than 50 percent. That in a nutshell is how weather forecasts are created.
One of the best things, from a forecasters' point of view, about weather forecasts is the calibration - that is, the forecast probability. But people usually just ignore that information. If they say 20% chance of snow, most people just think "it probably won't snow." Of course that actually means "if we get these exact conditions ten times this year, it will snow two of those times."
Disclaimer: I am not a meteorologist and have little knowledge of the field. I'm just a guy who likes data and has 12 weather apps on my iPhone.
0 Comments