Abstract

Weather forecasting is one of the most important and challenging field for scientists and engineers. The advent of technology has enabled us to obtain forecasts using complex mathematical models. For the last three decades, artificial intelligent based learning models like neural networks, genetic algorithms and neuro-fuzzy logic have shown much better results as compared to Box-Cox modeling approaches. Further accuracy is expectable by constructing a consortium of statistical and artificial intelligent methods. For weather forecasting, our trend is also towards the hybrid models. The accuracy of forecasting models can be made using different measures of assessments. In this paper, some hybrid methods are discussed with their merits and demerits. In this research we are going to use java for the front end designs and MySQL for the database.

CHAPTER ONE

INTRODUCTION

1.1 BACKGROUND OF THE STUDY

The problem of how to assess the performance of weather forecasts has been a long-standing topic in meteorology. Since 1950, progress in numerical weather prediction (NWP) has prompted meteorologists to develop and apply many different objective verification and evaluation techniques to forecast fields. A comprehensive review and discussion of standard methods was compiled recently by Jolliffe and Stephenson (2003). Over the past two decades, the use of commercial weather forecast products by industrial decision makers has become much more widespread. In parallel the range of competing weather forecast services to choose from has widened. As a result there is a pressing need to include the user s perspective in the verification methodology (referred to as user oriented verificationin the recent MO guidance report by Ebert et al., 2005).

Because the mainstream verification methods were primarily devised to answer the needs of model developers, it is not surprising that a common concern in the literature on the subject is to monitor and improve a forecast system following requirements posed by atmospheric scientists, rather than the needs of specific end-users who are more interested in the performance of a set of forecasts relative to their own specific demands (see e.g. Glahn, 2004).

Another reason for the lack of research in assessing the usability of weather forecasts is the belief that users are only concerned with economic value (Katz and Murphy, 1997). In theory, this would be the case if all users were able to formulate exactly their own utility functions. However, utility functions are often very difficult to determine in practice, and real-world decision-making processes are often a lot more complicated than the idealized cost-loss models used in the meteorological literature. For these reasons, many users often find it simpler to look at the forecast fitness for purpose, which in this work will be referred to as forecast quality. It goes without saying that good meteorological performance constitutes an essential must-have characteristic of weather forecasts, so that a proper quality assessment strategy cannot be envisaged without due consideration of relevant meteorological parameters. Of course, principles that constitute best practice for meteorological quality control can also be extended to a wider range of verification techniques, including the assessment of economic value.

Nowadays forecast users are able to receive weather forecasts from many different sources. The spread of quality between all available products is considerable, and in some cases the origin of the forecasts is obscure. Even when performance statistics are available from competing providers, in general those measures of quality are not presented in forms that allow immediate comparison, and they do not relate directly to the practical applications for which the forecasts are supplied. Typically, it is the suppliers rather than the users who assess the quality of weather forecasts. This situation has the potential to lead to conflicts of interests that may affect the credibility of the whole industry. There is therefore an obvious need to develop independent standards of practice that are scientifically sound, adhered to by forecast suppliers, and trusted by their customers.

It’s good to note that although research in forecast verification is continually developing new methodology for new products, the complexity of this problem goes well beyond what has been previously addressed by the classical forecaster-oriented verification methodology.

1.2 OBJECTIVE OF THE STUDY

This study primarily aims at initiating the development of a system which beats the current method of weather forecast by accurately predicting weather conditions. This project examines fundamental and tries to curb issues raised by current verification practice and by the lack of generally agreed standards. Its aim is to present practical solutions and make feasible proposals that will allow the industry to tackle these problems in a way that is beneficial to both forecast users and providers.

The project objectives can be summarized as follows:

1. To collect real time Rainfall data of Ekwulobia region.

2. To pre-process the data using Data Cleaning technique for predict the accurate result in MATLAB 2010b.

3. To use Intelligent Approach for develop the Fuzzy Inference System with neural Network for rainfall forecasting.

4. To calculate the parameters such as PSNR and Means Squared Error Ratio to identify the accuracy.

5. To Implement the Min-Max Normalization on the Real Data.

6. Generate Results and Compare with actual results.

1.3 STATEMENT OF THE PROBLEM

Presently, accurate weather forecasting has been a major problem in Nigeria. The old methods have met many unresolved challenges which has become a major concern for Meteorology department in Nigeria. The task is complicated in the field of meteorology because all decisions are made within a visage of uncertainty associated with the weather system (Hasan et al., 2008). Chaotic features associated with atmospheric phenomenal has also attracted the attention of modern scientist. The challenges are mainly related to:

· Poor observing network

· Poor model performance

· Gap between modelling and model use

· Lack of training to catch up with new tools (e.g. GPS, EPS) and to update knowledge (interaction research-operation).

· Lack of documentation (e.g. Forecaster’s handbook) and systematic verification.

1.4 SCOPE OF THE STUDY

The scope of this research covers the development of a system software which will be implemented in meteorological institution on Nigeria. The system is going to have two input variables and one output variable. The inputs variables are the wind speed and the temperature at a particular time and the output will be the amount of the rainfall expected. Temperature and the wind speed are chosen because it was proved that they are the factors that influence the occurrence of rainfall. The input variables values (wind speed and temperature) are grouped into five using the linguistic variable terms which are:

· Negative large: Very Low

· Negative small: Low

· Zero: Normal

· Positive small: High

· Positive large: Very High

1.5 RESEARCH QUESTIONS

This research work will be guided by the following research questions:

i. Why proposing a modern weather forecasting method for the organization. Is it really important?

ii. How will the proposed computer-based system affect the existing conventional system?

1.6 LIMITATIONS OF THE STUDY

There are lots of problems I encountered during this research process which limited but didn’t stop me from doing a thorough research. These problems can be formulated as:

· Collection of the historical data, facts and figures about the rainfall is a difficult process.

· To make the estimate about the rainfall regions, i.e. some having low rainfall, medium rainfall, heavy rainfall. Estimation of these regions is also a very difficult process.

· Choosing the prediction technique is also a matter of concern

· Generating technique of (adaptive neuro-fuzzy inference system) ANFIS is also challenging.

· For cleaning the data, pre-processing of the data is also seems like very tough process.

· Performance analysis of different rainfall regions is also difficult

1.7 ASSUMPTION OF THE STUDY

During the process of data collection, information relating to weather forecasting services was obtained from Nigerian Meteorological Services. The information was collected from the admin staff during the course of my research. Hence, it is assumed that all the data collected are correct and contains no false information.

1.8 DEFINITION OF TERMS

· Weather forecasting: is the application of science and technology to predict the conditions of the atmosphere for a given location and time.

· Rainfall intensity: Rainfall is classified according to the rate of precipitation: Light rain — when the precipitation rate is < 2.5 mm (0.098 in) per hour. Moderate rain — when the precipitation rate is between 2.5 mm (0.098 in) - 7.6 mm (0.30 in) or 10 mm (0.39 in) per hour

· Fuzzy Logic:  is an approach to computing based on "degrees of truth" rather than the usual "true or false" (1 or 0) Boolean logic on which the modern computer is based.

· Back-propagation : is an algorithm for supervised learning of artificial neural networks using gradient descent

· Biclustering: is a data mining technique which allows simultaneous clustering of the rows and columns of a matrix.

· numeric weather prediction

---

---