AN ANN BASED WEATHER FORECASTING SYSTEM

INTRODUCTION

Weather is the specific condition of the atmosphere at a particular place and time. It is
measured in terms of such things as wind, temperature, humidity, atmospheric pressure,
cloudiness, and precipitation. In most places, weather can change from hour-to-hour, day-to-day,
and season-to-season
Weather phenomena, usually of a complex nature, have a direct impact on the safety
and/or economic stability of human beings. The impact of extreme weather phenomena on
society is growing more and more costly, causing infrastructure damage, injury and the loss of life
due to which future state of weather has always been a subject of curiosity for general people that
intended the need of weather forecasting.
Weather forecasting is the application of science and technology to predict the state of
the atmosphere for a future time and a given location. . Weather prediction could be short
range(one day/one week) or long range(few months/year ahead) depending on time ( about how
much prior the weather has to be predicted ) . Weather forecasting remains a complex business,
due to its chaotic and unpredictable nature.
In past days, forecasting was done by simple observation of the sky but then with the
change of time new methods evolved, from relatively simple methods to highly complex
computerized mathematical models. Presently, the most general method used for prediction is
Numerical Weather Prediction (NWP) which uses the power of computers to make a forecast.
The shortcoming of the NWP method lies in the use of very complex equations used in the
processing of models resulting in error in the predictions. An alternative approach is artificial
intelligence techniques used for modeling chaotic behavior of weather which is based on
technique of extracting patterns from observed past events and extrapolating them into the future.
Artificial Neural Network (ANN) can model complex nonlinear relationships and is a data-driven
method.

WEATHER FORECASTING
A large number of methods for forecasting have been existing from years. First of these
methods is the Persistence Method; the simplest way of producing a forecast. The persistence
method assumes that the conditions at the time of the forecast will not change. The Climatology
Method is another simple way of producing a forecast. This method involves averaging weather
statistics accumulated over many years to make the forecast. Another method called Analog
Method is a slightly more complicated method of producing a forecast. It involves examining
today's forecast scenario and remembering a day in the past when the weather scenario looked
very similar (an analog). The forecaster would predict that the weather in this forecast will behave
the same as it did in the past. Next is the Numerical Weather Prediction (NWP) which uses the
power of computers to make a forecast.
 Numerical Weather Prediction (NWP) uses the power of computers to make a forecast.
Complex computer programs, also known as forecast models, run on supercomputers and
provide predictions on many atmospheric variables such as temperature, pressure, wind, and
rainfall. The basic idea of NWP is to sample the state of the atmosphere as a fluid at a given time
and use the equations of fluid dynamics and thermodynamics to estimate the state of the fluid at
some time in the future.
The NWP method is flawed in that the equations used by the models to simulate the
atmosphere are not precise. This leads to some error in the predictions. In addition, there are
many gaps in the initial data since we do not receive many weather observations from areas in
the mountains or over the ocean. If the initial state is not completely known, the computer's
prediction of how that initial state will evolve will not be entirely accurate.
Several artificial intelligence techniques have been used in the past for modeling chaotic
behavior of weather. In recent years, more and more intelligent weather forecast based on
artificial neural networks (ANNs ) has been developed and its performance evaluated. Due to
their potential to represent complex nonlinear behavior such as the relationship between future
temperatures and available data that contribute to certain weather conditions, ANNs are
becoming increasingly prominent in many areas of weather forecasting. Also, the considerable
success of ANNs in short term load forecasting and agriculture risk management provides
encouragement to apply these methods to other forecasting problems. Due to these reasons,
much research works are being carried out on the application of artificial intelligence techniques
to the weather forecasting problem.

ANN BASED APPROACH

In ANN based approach, Neural network models are developed for forecasting the
weather .The developed ANN model is based on one of the neural network architecture, it may be
among Multi-Layered Perceptron (MLP), Elman Recurrent Neural Network (ERNN), Radial Basis
Function Network (RBFN) or any appropriate architecture. The network can be trained for
function approximation (nonlinear regression), pattern association, or pattern classification. The
training process requires a set of examples of proper network behavior - network inputs and
target outputs and training algorithm.

The ANN is used to learn the relationship among past, current and future weather

parameters like temperature. Forecasting is based on extracting patterns from observed past
events and extrapolating them into the future and one should expect artificial neural
networks (ANN) to be good candidates for this task for at least two reasons. First, ANN
are able to approximate numerically any continuous function to the desired accuracy. In
this sense, they should be expected to model complex nonlinear relationships much better
than the traditional linear models that still form the core of the forecaster’s methodology.
Secondly, ANN are data-driven methods, in the sense that it is not necessary for the
researcher to postulate tentative models and then estimate their parameters.

EXPERIMENTAL WORK
We have worked on designing & implementing a Neural Network based model for
forecasting weather parameters.
There are generally four steps in developing the ANN model:
• Assemble the training data

• Create the network object

• Train the network

• Simulate the network response to new inputs

we have tried to apply ANN based forecasting for making Time series rainfall prediction
for Varanasi region of U.P. state of India. We have obtained most of the input data from
the Ozone Unit sub observatory at BHU in Varanasi and from the MET observatory at
Babatpur, Varanasi
A time series is a sequence of data points, measured typically at successive times, spaced
at (often uniform) intervals. Time series forecasting is the use of a model to forecast
future events based on known past events: to forecast future data points before they are
measured.

We collected data (rainfall) of 18years’s three months(monsoon season)for the prediction

of time series (next year) rainfall Prediction. We divided our every year’s three
month(monsoon season-JJA) data into six sets each comprising of 15 days ,and then
calculated mean of each set, in this way for every year there are six data set ,resulting in
six neural networks for training which helped in simplifying the training of ANN. We
used 17 years data as training patterns and one year data for the simulation purpose.
In our work, we created the network using Multilayer Feedforward (MLFF) architeture
and trained with back propagation algorithm. The trained ANN can be simulated and used
to predict the future weather conditions. Standard statistical measures like mean, standard
deviation, root mean square error etc. has also been worked out for verification purpose.

CONCLUSION
Use of Artificial Neural Networks to forecast different Weather parameters, is
becoming a popular field of study since last few years. We tried to use this technique for
doing few experiments on Time Series Rainfall Prediction. The results obtained has been
quite accurate as evident from the graphs and the calculated Statistical indicators like
mean, standard deviation, root mean square error etc (as shown in fig 1). We also wanted
to work with ANN based forecasting techniques in areas other than weather such as
Economic and Social parameter trends. It is a very interesting area worth working in and
we would take this opportunity to recommend to the readers of this report (particularly
students) to take it as an area worth exploring.