This project titled "An approach to study seasonal variations and quantify the height of sea level acceleration for global mean sea level and the Bay of Bengal using time series analysis" presented by Anika Tabassum (ID Saad Bin Omar and Mustafizur Rahman (ID in the Department of Computer Science and Engineering. Daffodil International University has been accepted as satisfactory in partial fulfillment of the requirements for the Bachelor of Science degree in Computer Science and Engineering and has been approved for its style and content. We declare that this project has been done by us under the supervision of Masud Rabbani, Lecturer, Department of Daffodil International University CSE.
We also declare that neither this project nor any part of this project has been submitted elsewhere for the award of any degree or diploma. We are grateful and express our sincere thanks to the supervisor Mr. "Masud Rabbani", Lecturer, Department of CSE Daffodil International University, Dhaka. Syed Akhter Hossain, Head of CSE Department for his kind help in completing our research and other faculty members and staff of CSE Department of Daffodil International University.
In our work, the forecast was made according to seasonal effects, so the result can be accurate according to seasonal changes.
Introduction 1
Motivation 1
Rational of the study 1-2
It should be clarified that our frequency was monthly, and it can be argued that we have listed all predictions at high frequency. So we can argue that we have established a dependency in climate models and forecasts, which is why our technique is quite different from them and obviously transparent. What will be the approaches to establish the model if the right model is found.
If machine learning has its own categories to measure or define the problems, how can it be impossible to find a good model. Now if we want to discover some strange events and analyze a pattern, the anomaly detection algorithm will be the assistant to define the answer. So if we want to find out a numerical value, we have to do it by doing regression analysis.
Time series analysis has four components according to which the data will be constructed, or we can say that the data that we want to analyze with time series analysis must be components of time series analysis. We then extracted the time series components from this data set to train our data. These components were numerical values and to reveal them, we performed regression analysis as mentioned in the previous section.
After that, we calculated each of them with special laws, which were later described step by step when necessary. For this reason, the components of the time series were constructed using regression analysis. So we can argue that the result satisfies the causes of sea level rise.
Expected output 3
Report layout 3
For research on sea level rise, spatial and seasonal variations were edited in a study and this was for three different locations in the Red Sea. Tide gauges make the most authentic measurements of sea level counting the height of the sea relative to a neighboring geodetic standard. Apart from instrument platform effects, softening or appearance of the land at any particular location changes the rate of relative sea-level rise in that certain area.
One of the main obstacles to progress seems to be the fact that sea level rise is highly subject to the dynamics of ice sheets and glaciers which are not yet sufficiently understood [9]. To analyze the use of dynamic models for predicting sea level rise [10] or statistical models based on climate-related predictors [11] for sea level prediction, some research studies have been conducted. All the study was performed for specific local sea level or global sea level.
This is an approach to study seasonal variations and to estimate the height of sea level acceleration for global mean sea level and the Bay of Bengal using time series analysis. In table 1, we have shown the start and end time to continue forecasting, firstly for Local sea level, which is for the Bay of Bengal and secondly for global mean sea level. Data Analysis in Excel File: After getting the data of sea level anomaly, we smoothed the data of SL (sea level) by the moving average (MA) which is an important algorithm of TSA.
Completing the calculation in the excel file: After evaluating the training part, we tested it with a year for which we already had data and thus could predict the acceleration of sea level rise for the next year. So, SLRP using TSA has the main objective to predict sea level of GMSL and BOB. On the blue line, the dots show the rise and fall of sea level and that is for real data.
Prediction of rising sea levels using time series analysis has been developed in the environment of an Intel Core i GHZ processor with 8.0 GB RAM on Windows 10 [16]. Due to the thermal propagation of water in the oceans and the melting of ice caps and glaciers on land, today's sea level rise is usually caused by global climate change. It is not possible to predict sea level rise using physics alone, but if the overall result is calculated based on more than 15 years, then each of the causes of the rise must have occurred.
An improved sea-level forecasting plan for cargo handling in the US-affiliated Pacific Islands.
BACKGROUND
Related Work 4
Research summary 5
Our data has been collected from the sources mentioned above and it was not so easy to collect data on PDO. In weka, data has been processed to show rise and fall of real data, training data and prediction data using regression analysis. In Excel, prediction has been made for 2015 and 2008, which are for GSML and BIR respectively.
4”, a screenshot of regression analysis for real data of GMSL, which has been performed for 1993-2014 and generated by weka, is shown. 5”, a screenshot of regression analysis for real data of PDO, which has been performed for 1993-2014 and generated by weka, is shown. 6”, A screenshot of regression analysis after prediction for GMSL has been shown for 2015 which was our known data and we predicted it to check our prediction with the known data.
8” forecast for GSML was calculated in excel and further converted to wekas which are calculated for 2015 when the data was known. The 9” forecast for GMSL was shown in weeks calculated for 2018 when the data was unknown. The 10” forecast for BOB was calculated in excel and further converted to wekas, which are calculated for 2008 when the data was known.
11” prediction for BIR has been shown in weka which is calculated for the year 2018 where they were unknown data. In Table III, the mean value has been calculated for real data by writing down an average of the SL(sea level) values. In Table IV, mean value has been calculated for training data by averaging the values of all data in the training data set.
In Table V, the mean is calculated for the training data by averaging all data values of the prediction data set. 22 In Table VI, the mean is calculated for the real data by subtracting the mean of the SL (sea level) values. In Table VII, the mean is calculated for the training data by subtracting the average of all data values of the training data set.
In Table VIII, the mean is calculated for the training data by subtracting the mean of all data values of the prediction data set.
Scope of the problem 5
Challenges 5
The training part has therefore been extended to 22 years for GMSL and 15 years for BOB. Initially, the Excel file in which we implemented all the calculations, which we explained in the data study in the Excel file section, was converted to CSV for further implementation in weka. After predicting the value, a figure was also plotted for the prediction value with real data, with the x-axis showing years divided into monthly segments and the y-axis showing gains.
7", A screenshot of the post-prediction regression analysis for 2008 BOB, which was our known data, was shown and we predicted it to check our prediction with this known data. The SLRP using TSA was calculated in Microsoft Excel and developed in Weka (version 3.8) [17] to prepare the model.Here, the data were trained from 1993 to 2014 and the rightmost values show the amount of GMSL in 2015 as a forecast.
Here, the data is trained from 1993 to 2007 and the best fit values show the amount of BOB in 2015 as a forecast. Using this CMA, a regression analysis was performed to detect all TSA components, and the irregularities and seasonality of our data were calculated by dividing our real data by the central moving average. To extract the component, regression analysis is done to get the numerical value to perform further process.
While the data is collected monthly, the standard deviation error can be reduced if the calculation continues according to daily data.