PROJECT—2 9 Topic : Weather Prediction J OBJECTIVE To study how weather is predicated (prediction of monsoon from past data) and understand the use op mathematics in weather prediction. INTRODUCTION Weather simply refers to the condition of air on the earth at a given place and time. It is @ continuous, ese properties make weather forecasting data-intensive, multidimensional, dynamic and chaotic process. Th a formidable challenge. Forecasting is the process of estimation in unknown situations from historical data Weather forecasting was one of the most scientifically and technologically challenging problems around the world in the last century. of the atmosphere and Weather forecasts are often made by collecting quantitative data about the current stat of atmospheric processes to project how the atmosphere will evolve in future using scientific understanding nn of the state of the The task of weather forecast is to provide the best description of the short-time evolutior atmosphere ~ wind, temperature, pressure, humidity, and precipitation — as represented on a 3D lattice. Weather forecasting is done to describe and predict the instantaneous values of the main atmospheric variables. Weather prediction could be one day/one week or a few months ahead. The accuracy of weather forecass however, falls significantly beyond a week. IMPORTANCE OF WEATHER FORECASTING There is a vast variety of end uses to weather forecasts. Weather wamings are important forecasts because they are used to protect life and property. Forecasts based on temperature and precipitation are important to agriculture, and therefore to traders within commodity markets. Temperature forecasts are used by utility companies to estimate demand over coming days. On an everyday basis, many use weather forecasts to determine what to wear on a given day. Since outdoor activities are severely curtailed by heavy rain, snow and wind chill, forecasts can be used to plan activities around these events, and to plan ahead and survive them. jects) Mathematics - 12 (Practicals & Prol (2) ACE Scanner 46ae wa Vi Ce Warmer ocean esure — ee et ‘agummer monsoon B Winter monsoon information about the uncertainty of weather forecasts can be sunmarmosscon portant for decision-makers. (eg, public transportation ruthorties, airlines, shipping, and the military), as well as the pubic. Weather forecasting is also used in many situations like Mere weather alerts and advisories, predicting the behaviour ofthe cloud for air transport, prediction of waterways in a sea, sgriultural development and avoiding forest fire. Improvements in weather information enable forecasters to support public safety officials in better anticipating and preparing for the impacts of weather events, Meteorologists play an integral role in predicting weather. Their skills are critical to the process of generating forecasts and warming the public of weather threats. HISTORY AND PRIMITIVE METHODS For millennia, people have tried to forecast the weather. In 650 BCE, the Babylonians predicted the weather from cloud patterns as well as astrology. In about 350 BCE, Aristotle described weather patterns in Meterologica. Later, The Ophrastus compiled a book on weather forecasting, called the Book of Signs. Chinese weather prediction lore extends at Jeast as far back as 300 BCE, which was also around the same time ancient Indian astronomers developed weather-prediction methods. In 904 CE, Ibn Washiyya’s Nabatean Agriculture, translated into Arabic from an earlier Aramaic work, discussed the weather forecasting of atmospheric changes and signs from the planetary astral alterations, signs of rain based on observation of the Iunar phases, and weather forecasts based on the movement of winds. Ancient weather forecasting methods usually relied on observed patterns of events, also termed pattern recognition. For example, it might be observed that if the sunset was particularly red, the following day often brought fair weather. This experience accumulated over the generations to produce weather lore. However, not all of these predictions prove reliable, and many of them have since been found not to stand \P to rigorous statistical testing. NUMERICAL WEATHER PREDICTION cee weather prediction (NWP) uses mathematical models of the atmosphere and ‘oceans to predict the OF coy based on current weather conditions. Though first attempted in the 1920s, it was not until the advent mputer simulation in the 1950s that numerical weather predictions produced realistic results. Tn the dai : the daily operation of weather forecasts, powerful supercomputers are used to predict the weather by Projects a7 fe ACE Scannersolving mathematical equations that model the atmosphere and oceans, In numerical weather prediction (NWP), the supercomputers simplify the mathematics used in meteorology. Instead of equations, the supercomputers look at small Parametric changes with respect to small physical motions and plug these observations into a series of mathematical algorithms that represent the Physical properties of the atmosphere. Using these algorithms, the supercomputers predict what will happen globally in the future. These computers can manipulate vast datasets collected from observations and perform extremely complex calculations to search for optimal solutions with a dimension as high as 108. The idea of NWP was formulated as early as 1904, long before the invention of the modem computers that are needed to complete the vast ji number of calculations in the problem. In the 1920s, Lewis Fry Richardson used procedures originally developed by Vilhelm Bjerknes to produce, by hand, a six-hour forecast for the state of the atmosphere ove, ‘two points in central Europe, taking at least six weeks to do so. ; Several areas of mathematics play fundamental roles in NWP, including mathematical models and their associated numerical algorithms, computational nonlinear optimization in very high dimension, huge datase ‘manipulation, and parallel computation. Even after decades of active research with the increasing power of supercomputers, the forecast skill of numerical weather models extends to about only six days. We rely on the models extensively because they have incorporated things like chaos theory and can give ensemble forecasts, which allow for changing probabilities and small statistical variations giving very different ) results. MATHEMATICAL CONCEPTS USED IN WEATHER PREDICTION In order to predict the weather, we always need to first establish an understanding of the system. Understanding and explaining the weather system requires ground understanding of Calculus and Geometry (Cartesian system) The displacement of features in this system drags in Linear Algebra and Vector Calculus, Numerical, Rel, Functional and Complex analysis and dynamical systems. Now in order to deal with errors in the pre a better understanding of statistics is needed. Probability is also used in monsoon prediction. CARTESIAN COORDINATE SYSTEM In meteorology and other atmospheric sciences, we mostly use the standard x, called the local Cartesian coordinate system, and the spherical coordinate system. Local Cartesian Coordinate System The local Cartesian coordinate system applies to three dimensions. The convention The zero point, x = y = z= 0 or (0,0,0), is arbitrary. x increases to the east; x decreases to the west. ¥y increases to the nort 8 , and z coordinate system, is simple: decreases to the south, ‘Mathematics - 12 (Practicals & Projects) dl (2) ACE Scannerjnoreases goin UPS = decreases going down. z . ‘a istance vector extending ffom the origin to (x,y, 2) is given by r=ix tjy + kz init vectors (length I along standard coordinates) are # (east); (north; (up) : we consider motion in two dimensions as being separate from motions in t Gr orizontal with @ subscript H; for example, rH = i x+y, ‘mis coordinate system works well over relatively small scales sate, where Eart’s curvature is not important. This is why the ervdinate system. This coordinate system does not work so wel ns in the vertical. We usuall s . ly denote where rH is a horizontal distance vector. on Earth, perhaps the size of an individual qualifier "local" is used in the name of the Il for large-scale motion on Earth, which is zup ynorth xeast axis ans ‘The local Cartesian system when placed on the globe SPHERICAL COORDINATE SYSTEM We use the following terms: = distance from the center of the Earth © = latitude (90° to +90°, or —1 /2 to +n/2) d= longitude (-180° to 180°, or ~x to +n) PA is at © = 40.8° and A=-77.9° Note that 1o of latitude is always 111 km or 60 nautical miles, but 1° of longitude is 111 km only at the equator. It is smaller in general and equal to 111 km x cos (@). Note that 1 nm = 1.15 miles. To find the horizontal distance between any two points on Earth's surface, we first need to find the angle of the arc between them and then we can multiply this angle by Earth's radius to get the distance. To find the angle of the arc, Ac, we can use the Spherical Law of Cosines: Ac = arccos(sin IRsinb2 + cost facos2!2cosA2) Where, the latitude and longitude of the two points are 1, 41 and 2, 42, respectively and AX=[), — 24] is the absolute difference between the longitudes of the two points. Note that the angle of the arc must be in radians, where 2m radians = 360°. To find the distance, simply multiply this arc angle by the radius of the Earth, 6371 km. DOMAIN The horizontal domain of a model is either global, covering the entire Earth, or regional, covering only part of the Earth. Regional models (also known as limited-area models, or LAMs) allow for the use of finer a ‘icing than global models because the available computational resources are focused on a specific are instead of being spread over the globe. This allows regional models to resolve explicitly smaller-scale meteorological Projects 3 (2) ACE ScannerYF phenomena that cannot be represented on the coarser grid of a global mel Regional mel we sity Todel to specify conditions atthe edge oftheir domain (boundary condvens) 1 ei Slams outside the regional model domain to move ino its are. Uncerainy ane ree cra introduced by the global model used for the boundary conditions as errors attributable to the regional model itself. 8 wel STATISTICS Some people feel weather predictions are no more than extensive statistical analysis. Forecast ensembles are (YPE for statistical post-processing. Meteorologists rely on stath ists also rely licti to a predictions that extend from three are another hallmark of the use ist development of statistical ee xd statistical science. The n sia fer 7 he oop mes oe reas predictions has contributed significantly to this increase iy enhancing the scope ethods forecasting accuracy. : js 1-1800s, with the invention of the telegraph, Dai ame cc tein ee i ares om da ar i aan fof weather-related data began to sor meal shinee Sy weather was predicted, Today, weather-related data and the scientific discipline of statistics are much more sophisticated The data, for example, is now collected through satellites, barometers, radars, weather balloons and many other instruments that are located on land, in water and the atmosphere. The data these instruments are collecting includes atmospheric pressure, temperature, speed, rain, humidity, and many other components of current weather conditions. ‘The Weather Company's prediction process begins with the aggregation of data from over 195,000 personal ‘weather stations across the globe. They then combine the data with forecasting and statistical models to provide the hyper-local weather forecasts that we turn to on our computers, tablets and phones each day. Because forecast models based upon the equations for atmospheric dynamics do not perfectly determine weather conditions, statistical methods have been developed to attempt to correct the forecasts. Statistical models were created based upon the three-dimensional fields produced by numerical weather models, surface ) observations and the climatological conditions for specific locations. These statistical models are collectively referred to as model output statistics (MOS), and were developed by the National Weather Service for theit Suite of weather forecasting models in the late 1960s. educated guesses, in reality, they are the resul, of Ty biased and under dispersive, and there is a neg to apply rigorous scientific techniques to extra ‘on statistical techniques to provide weather Model output statistics differ from the perfect prog technique, which assumes that the output of numericel weather prediction guidance is perfect. MOS can correct for local effects that cannot be resolved by the model due to insufficient grid resolution, as well as model biases. Because MOS is run after its respective Blobal or regional model, its production is known as post-processing, Forecast parameters within MOS include maximum and minimum temperatures, percenta ine ‘nil ge chance of rain within od, precipitation amount expected, chance that the preci a several hour period, precipi and surface winds, pitation will be frozen in nature, chance for thunderstorms, cloudiness: ly where large scal 3 a wilt range of land areas. ‘ange scale processes dominate the weather over 50 eProett) ‘Mathematics - 12 (Practicals & Proie* (2) ACE ScannerChare2 ct {he quantum of rain India gets Assupercomputing bday has been slowly declining, Capacity expanded, 70.5, 70 ticularly post 1970s the quantity of 60 ‘Annual precipitation (in mm) weather data India 50 funnels into models a has also gone up < / 20 Zero — 0 600 97 r 7 "950 1960 1970 1980 1990 2009 2010 2996 bas pas PARTIAL DERIVATIVES ecause meteorology is a three-dimensional science, four if you include time, the mathematics used in meteorology can require extensive use of partial derivatives. Partial derivatives allow you to look at how omething such as wind speed changes when you move in one direction only, say, straight north. This could be important to pilots. They also let us determine the gradient of a field. That is, identify what direction to move in order to see the greatest temperature increase, for instance. And even how much it increases after you go a certain distance in that direction, We use the concept of partial derivatives in weather prediction. To forecast the weather, we construct a formula for the temperature as a function of several environmental variables, each of which is not entirely predictable. Weather forecast would change as one particular environmental factor changes, holding all the other factors constant. Let the temperature T depend on variables x and y, T = f(x, y). The rate of change of f with respect to x, holding y constant, is called the partial derivative of f with respect to x and is denoted by f(x, y). Similarly, the rate of change of f with respect to y is called the partial derivative of f with respect to y and is denoted by fs 9) We define f+ hy) ~ fe.) i ie 1, Y) ~ I fe = jig BRE 1 +) ~ fe, les y) = Him SEY a Se») Let flr, y) = 39? — ye Then, f(z, ») = tim cae a tig, thw? timg ; = gy? = lim 20 + AP = 9? hey > fg = tim ao? + xh? + Dhyx — 997 tim gh + Aye = Ip i ; in iy Hi (2) ACE Scanner¥ Wve use our knowledge of single-variable calculus to compute partial derivatives, To cajey,. ‘and differentiate /lx, y) with respect t0 : d(x] eS =I In practice, Fy) you view y as a constant firs y) =? as expected sinc ap] Similarly, fy(ss y) = Day sine “= 2" Geometrical Meaning ; ; ; Suppose the graph of z= (s,) isthe surface shown. Consider the partial derivative of f with respect to x at a point (¥p» Yo)- -_ ' Holding y constant and varying x, we trace out a curve that is the intersection of the surface with the vertical plane y = Yo- : . The partial derivative fey Yo) measures the change in z per unit incre a along this curve. That is, fry Yo) is just the slope of the curve at (¥oy ‘¥o)- The ‘geometrical interpretation of f,(%p, Yo) is analogous. MULTIPLE LINEAR REGRESSION ather conditions and predict climate behavior. One family of Meteorologists use various models to forecast we: e ersistence). CLIPER is based on Multiple Linear Regression such models is called CLIPER (climatology and p af methods and it’s used as the baseline model to compare other forecasting methods like the Hurricane Weather Research and Forecasting (HWRF) model and the Weather Research and Forecasting Model (WRF). One thing to note about multiple linear regression is that the independent variables must not be collinear, i, they do not have to have correlation coefficient between each other, else there will be difficulty in assessing the relationship between the dependent and independent variables. Before multiple linear regression is performed on a range of data values, we need to ensure that a linear relationship exists between each independent variable and the dependent variable. The amount of residual error must be almost constant at each point in the model. This model has multiple independent variables, unlike that of simple linear regression which has just the one. This can be represented by : Ym E myx + myx, & myx, tO +e y is the dependent variable. Xj, %y% are the independent variable. ‘my, my, My, are the values of the slopes. c is the intercepts ¢ is the residual error. GRAPHS Graphical rainfall forecast show data i oa im quantitative amounts which makes the process of weather prediction 2 Mathematics - 12 (Practicals & Proje! -» (2) ACE ScannerRAINFALL % OF LPA ase8 te 832288338 ©n@aor PEPESEESEES ERSTE 1988 1989 1990 1991 1992 1993 1994 aan Fig. 1: Performance of operational forecast (empirical model) for all india rainfall. (Source: Dr. D. S. Pai, IMD, Pune) 20E 160E | 160W 4 02 02 0A Fig. 2: Correlation coefficients between the observed CMAP (1979-1999) and the simulated June-August precipitation anomalies made by five-model multi-ensemble mean (adopted from Wang et al., 2005). Model Simulated Rain T3 = ‘Observed Rain —_— Mpg 04-03-0.2-0.100 0.1 0.20304 05 Projets 33 (2) ACE Scannermer Monsoon Rainfall (a) Epochal Patterns of All-India Sumi 1 eo. 3s ‘Standard Deviation Anomaly (%) 960 1970 1980 1990 2008" ter0 "7930 1940 1950 1 1880 1890 1900 1010 TOD yeorelatye Window soon rainfall simulated from uncoupled GFDLR3O (GOGA) model, ) model. Scatter plots of standardized anomalies of obsereg XGA and MLM are shown in (c) and (d), respecte, Fig. 3: Correlation maps of (a) observed SSTs and mor simulated SSTs and monsoon rainfall from the coupled (MLM from GO and simulated monsoon rainfall for each year during 1950-1999 from (adopted from Kumar et al., 2005) PROBABILISTIC FORECASTING Often the news weather reports determine the chance of raining in percentage. For example they say that tee is a 60% chance of rain, This percentage of chance of raining is calculated using the concepts of probabil M T WTH F S_ S hance of) 70% | BO% | 90% | 80% | 60% 20% | 0% fa faa? li li? | es 44? (64 4g Fone on. To calculate probability we divide the number of favourable outcomes by the total number of possible outcomes. If we want to know the probability that it will rain, this will be the number of days in our databes that it rained divided by the total number of similar days in our database. If we have data for 100 days wit similar weather conditions (the sample space and therefore the denominator of our fraction), and on 60 of these days it rained (a favourable outcome), the probability of rain on the next similar day is 60/100 or 6? The probabilistic technique is a complex way of making a fore ici rememb a previous weather event which is expected to be mimicked es Seen crear cae ae method of observing rainfall in places such as oceans, as well as the forecasting of precipitation amo" and distribution in the future. aia) Probability forecasts can be produced by several di weather event can be a forecaster’s judgment of tie Galion tar can also be produced directly from NWP models or from statistical For example, a probability frees the event will occur. Probability oe analyses of the output from these ™ . I ticals & proje* (2) ACE Scanner ‘Mathematics - 12 (Pract‘Aprobability Soe (Px P), also referred to as chance of precipitation or chance of rain, is a measure of the probability that at least some minimum quantity (0.01 inches or 0.25 mm) of precipitation will fall in any particular spot and occur within a specified forecast period. Mathematically, PoP is defined as follows: probability of Precipitation (P x P)= C x A, where, C = the confidence that precipitation will occur somewhere in the forecast area, and where A = the percent of the area that will receive measureable precipitation, if it occurs at all precipitation chance ranges from 0% to 100%. The precipitation chance is often rounded to 10% increments. The closer the chance is to 100%, then the more likely precipitation is. The closer the chance is to 0%, then the less likely precipitation is. One aspect of precipitation chance is a timeframe. Generally this timefiame is divided into 12 hour time frames which are generally a daytime timeframe and a night-time timeframe. Thus, when assessing precipitation chance, it is important to note the day of the week and timeframe the forecast is being made for. ‘Another aspect of precipitation chance is the location, Generally the location is for a specific location such as at a measuring station. If a person lives close to the measuring station then the precipitation chance forecast is fairly reliable for that person. However, if a person lives a significant distance from the measuring station, there can be a number of influencing factors including elevation changes, proximity to a body of water, closer or further to an uplift mechanism, and land surface changes. These factors can either increase or decrease the precipitation chance for a person that does not live relatively close to the measuring station. UNCERTAINTY AND ERRORS IN PREDICTION The inaccuracy of forecasting is due to the chaotic nature of the atmosphere, the massive computational power required to solve the equations that describe the atmosphere, the error involved in measuring the initial conditions, and an incomplete understanding of atmospheric processes. Hence, forecasts become less accurate as the difference between current time and the time for which the forecast is being made (the range of the forecast) increases. The use of ensembles and model consensus help narrow the error and pick the most likely outcome. Although no single model is accurate 100 percent of the time and observations are imperfect, access to a wide range of models with different abilities, strengths, and characteristics allow scientists to run ensemble forecasts with perturbed initial conditions or physics, helping increase certainty in the prediction. The equations governing weather are highly nonlinear. This means that the ‘cause and effect” relationships between the basic variables can become ferociously complex. To deal with the potential loss of predictability, forecasters study not one, but many forecasts, each one started from slightly different initial states, which reflect our ignorance of exactly how a weather system such as Hurricane Sandy formed. If the forecasts predict similar outcomes, we can be reasonably confident, but if they produce very different scenarios, then the situation is more problematic. CONCLUSION Weather forecasting is the application of science and technology to predict the conditions of the atmosphere for @ given location and time. There is a vast variety of end uses to weather forecasts. They are important to Protect life and property, agriculture and everyday basis for the common man to plan out his day. Weather forecasting has come a very long way since the Babylonians and the Greeks started observing the skies, and it was the pioneering work of Vilhelm Bjerknes and Lewis Fry Richardson at the beginning of the ‘Oth century that kicked off the development of modern weather forecasting. But without the invention and Subsequent improvement of computers, numerical weather prediction would still be in its infancy. Understanding chaos and developing new and better weather observation methods also contributed to improving forecast Sccuracy. A six-day forecast nowadays are very accurate and current research suggests that these percentages Will continue to increase in future. While much has been accomplished in improving weather forecasts, there tem ins much room for improvement. Projects 55 Re