Forecasting precipitation using sarima model

Mathematical Theory and Modeling www.iiste.org
ISSN 2224-5804 (Paper) ISSN 2225-0522 (Online)
Vol.4, No.11, 2014
50
Forecasting Precipitation Using SARIMA Model: A Case Study of
Mt. Kenya Region
Hellen W. Kibunja1*
, John M. Kihoro1, 2
, George O. Orwa3
, Walter O. Yodah4
1. School of Mathematical Sciences, Jomo Kenyatta University of Agriculture and Technology
P.O. Box 62000-00200, Nairobi, Kenya
2. Co-operative University College of Kenya, Computing and E-learning
P.O. Box 24814-00502, Nairobi, Kenya
*
E-MAIL: hkibunja@gmail.com
Abstract
Precipitation estimates are an important component of water resources applications, example, in designing
drainage system and irrigation. The amount of rainfall in Kenya fluctuates from year to year causing it to be very
hard to predict it through empirical observations of the atmosphere alone. Our objective was to determine the
forecasted values of precipitation in Mt. Kenya region and also to determine the accuracy of the SARIMA model
in forecasting precipitation in the same region. This research considers a univariate time series model to
forecast precipitation in Mt. Kenya region. We fitted the SARIMA model to our data and we picked the model
which exhibited the least AIC and BIC values. Finally, we forecasted our data after following the three
Box-Jenkins methodologies, that is, model identification, estimation of parameters and diagnostic check. Having
three tentative models, the best model had two highly significant variables, a constant and with p-values<
0.01 respectively.
This model passed residual normality test and the forecasting evaluation statistics shows ME=
-0.0053687, MSE=0.96794, RMSE=0.98384 and MAE= 0.75197. Indeed, SARIMA model is a good model for
forecasting precipitation in Mt. Kenya region
Keywords: SARIMA, Precipitation, Forecast, Mt. Kenya, AIC and BIC
1.0 Introduction
Time series methods determines future trend based on past values and corresponding errors. Since a time series
method only require the historical data, it is widely used to develop predictive models. A time series is simply a
set of observations measured at successive points in time or over successive periods of time. Time series analysis
is used to detect patterns of change in statistical information over regular interval of time. These patterns are
projected to arrive at an estimate for the future. Time series forecasting methods are based on analysis of
historical data. It makes the assumption that past patterns in data can be used to forecast future data points.
Several methods have been used in forecasting weather. We have Non parametric Methods like the Artificial
Neural Networks and parametric Methods. Some of the models under parametric are: Extrapolation of trend
curves, Exponential smoothing, The Holt-Winters forecasting procedure and Box Jenkins procedure.
1.2 Background Information
Precipitation estimates are an important component of water resources applications, example, in designing
drainage system and irrigation. Major sectors of economy in Kenya such as agriculture, livestock keeping,
hydro-energy generation, transport, tourism, among others are highly dependent on climate. Severe weather and
extreme climate events and other climatic fluctuations have been shown to have a high influence on the social

Vol.4, No.11, 2014
51
and economic activities of the country and the performance of the country’s economy KMD (2009). It has also
been noted that the past development projects may not have taken into consideration the potential impacts that
the climate has on their success. Due to the failures associated with lack of timely and effective forecasts, the
agricultural activities in the country have been immensely affected causing massive losses to farmers who would
have easily avoided these outcomes with prior notice; integration of technology in agriculture have brought with
it crops that are rainfall specific. Traditionally, long rains occur from March through to May and short rains from
October to December but because of climatic changes, this trend is somehow changing. These changes normally
occur on aspects of weather such as wind speed, humidity, temperature , precipitation which occurs in a variety
of forms; hail, rain, freezing rain, sleet or snow among others. Therefore, there is need more accurate forecasting
techniques to be applied in predicting climatic patterns. Precipitation estimates being an important component of
water resources applications, an accurate estimate of rainfall is needed. There are also concerns with producing
valid estimates using appropriate methods. In order to develop a comprehensive solution to the forecasting
problem, including addressing the issue of uncertainty in predictions, a statistical model must be developed.
2.0 Literature Review
Rainfall prediction is a challenging task especially in the modern world where we are facing the major
environmental problem of global warming which has rendered the previously employed methods to redundant.
Earlier forecasting methods such as simple quantitative precipitation forecasts used by Klein and Lewis
(1970),Glahn and Lowry (1972) and Pankratz, 1983 have lost their edge due to the changing patterns and
variability in rainfall that may be associated with global warming. However, the world of statistics has been
evolving over time leading to creation of more efficient and effective methods allowing researchers to make
enormous efforts in addressing the issue of accurate precipitation predictability. Borlando et al., 1996 used
ARIMA models to forecast hourly precipitation in the time of their fall and the amounts obtained were compared
with the data to measure rain. They came to the conclusion that with increasing duration of rainfall, the
predictions were more accurate, and shorter duration of rainfall, rain rate difference will be more than the actual
corresponding value. Yusof and Kane, 2012 analyzed the precipitation forecast using SARIMA model in
Golastan province and found the seasonality measure in SARIMA to be highly useful in measuring precipitation.
2.1 SARIMA Models theory
Box Jenkins (1970) generalize ARIMA model to deal with seasonality. Autoregressive Integrated Moving
Average (ARIMA) models are generalizations of a simple AR model that uses three tools for modeling serial
correlation in disturbance. The first tool is an autoregressive, or AR term. Each AR term corresponds to the use
of lagged value of the residual in forecasting equation for the unconditional residual. The AR model of order p,
AR (p) has the following form:
= + + ⋯ + + ……………………………………………………… (1)
With the use of a lag operator B, the equation becomes:
1 − − − ⋯ − = = ………………………………………………..(2)
Where for B holds =
Next tool is integration of order term. Each integration order corresponds to the differentiation of the series
being forecast. The first order differentiation component means that the forecasting model is designed for the
first difference of the original series .The second order component corresponds to the second difference and so
on. The third tool is a Moving Average, MA term. The MA forecasting model uses lagged values of a forecast
error to improve the current forecast. The first order MA term uses the most recent forecast error. The second

Vol.4, No.11, 2014
52
term uses the forecast error from two most recent periods and so on. MA process of order q, MA (q) has the form:
It is written as
= − − ⋯ − …………………………………………………………………… (3)
Using lag operator,
= 1 − − ⋯ − = ……………………………………………………….. (4)
When modeling time series with systematic seasonal movements, Box-Jenkins recommended the use of seasonal
autoregressive (SAR) and seasonal moving average (SMA) terms. The seasonal autoregressive process of order P
can be written as:
= Φ + Φ + ⋯ + Φ + …………………………………………………….. (5)
Or
Φ = ……………………………………………………………………………………… (6)
The seasonal MA of order Q can be written as
= − Θ − ⋯ − Θ ………………………………………………………………… (7)
Or equivalently,
= Θ …………………………………………………………………………………….. (8)
In all the four components above, s denotes the length of seasonality. Finally, we can write the general
SARIMA , , × !, ", # $with constant model as
Φ 1 − %
1 − &
= ' + Θ ……………………………… (9)
Where the constant equals
' = ([ 1 − − − ⋯ − 1 − Φ −Φ − ⋯ − Φ ] …………………………………… (10)
3.0 Materials and Methods
3.1 Study Area
The study concentrated on statistical modeling of precipitation in Mt. Kenya region in central Kenya. This region
is predominantly agricultural dependent; its profitability would significantly increase if there was access to
reliable and timely forecast of rainfall data. This region would benefit from the success of this study. The region
also has other sectors that depend on reliable forecasts of climatic conditions such as tourism, some service
industry such as electricity and water supply. Mount Kenya region is the source of major rivers in Kenya and the
climatic conditions in this area are highly unpredictable.
3.2 Study Data
The data employed in this research comprises precipitation and wind monthly data collected from Kenya
meteorological department covering a period of 1995 to 2010 for wind data and 1970 to 2011 for precipitation
data but will be limited to the available wind data. This data is highly reliable as it is collected on a daily basis in
the stations and therefore future data needs may be easily met from the station.
4.0 Results
4.1 Data Analysis Process
Data was analysed using Gretl which has inbuilt functions like MLE to deal with ARIMA models. Preliminary
data analysis was performed on hourly daily precipitation from 1995-2010 using Box-Jenkins modeling
methodology. Time series plot was done using raw data to assess the stability of the data and the following time
series plot was obtained.

Vol.4, No.11, 2014
53
4.2 Precipitation Time Series Plot
Figure 1 plot show that our data is stationary. A non stationary series is the one in whose values do not vary with
time over a constant mean and variance.
4.3 ACF and PACF plots of precipitation
Figure 2 show ACF and PACF plots of precipitation. The auto-correlation indicates that there is no seasonality.
Seasonality normally causes the data to be non-stationary the average values because the average values at some
particular times are different than the average values at other time
4.4 SARIMA Forecasting Results
SARIMA model was fitted after following Box-Jenkins four major steps in modeling time series and the
appropriate model was obtained by choosing the model which yielded minimum AIC and BIC, Akaike (1979).
After a series of model tests, the following models were obtained.
4.4.1 Tentative seasonal ARIMA models
There were three tentative models as shown in table 1.
SARIMA (1, 0, 1) × (1, 0, 0)12 turns out to be the best model since it has the least values of the information
criterions. The details of this model are shown in table 2. This model has two significant variables. The
correlation matrix of this model was examined. The correlation between the parameters of the model was a
weaker one. This implies that all the parameters are important in fitting the model. The fitted model is given by:
+ +Φ + Φ = + + + …………………………………………………. (11)
Upon replacing the coefficients of the model with real values, we get the follow:
+ 0.204 + 0.464 + 0.095 = 3.84 + + + 0.117+ ……………………………………….. (12)
4.4.2ACF and PACF plots of residuals
Figure 3 show that the residuals are white noise as there are no significant spikes.
4.4.2 Normality test of residuals
Figure 4 show a histogram which has a bell shaped distribution with a p-value of 0.007 which is a good indicator
of normality in the distribution.
4.4.3 Residual Q-Q Plot
The QQ plot in figure 5 approximately follows the QQ line visible on the plot. This is a good indicator of
normality within the residuals
5.0 Conclusion
The main objective of this study was to forecast precipitation using SARIMA model and also to determine the
accuracy of the SARIMA model in forecasting precipitation in Mt. Kenya region To avoid fitting over
parametized model, AIC and BIC were employed in selecting the best model. The model with a minimum value
of these information criterions is considered as the best (Akaike (1979); Akaike (1974)). In addition, ME, MSE,
RMSE, MAE, MPE, MAPE were also employed. The ACF plots of the residuals two models were examined to
see whether the residuals of the model were white noise. SARIMA model turns to be a good model for
forecasting precipitation in Mt. Kenya region.

Vol.4, No.11, 2014
54
References
[1] Akaike Hirotugu (1974),’A New Look at the Statistical Model Identification, IEEE, Transction Automatic
Control 19(6), 716.
[2] Akaike Hirotugu (1979),’Bayesian Extension of Minimum AIC Procedure of Autoregressive Model Fitting’,
Biometrika 66(2), 237-242.
[3] Anderson Oliver D. (1977), ‘Time Series Analysis and Forecasting: Another Look at the Box-Jenkins
Approach’,Journal of Royal Statistical Society (The Statistician) 26(4), 285-353
[4] Borlando P. ,Montana R. and Raze (1996), ‘Forecasting Hourly Precipitation in time of fall using ARIMA
Models’ Journal of Atmospheric Research 42(1), 199-216.
[5] Box George Edward Pelham and Gwilyn M. Jenkins (1976), ‘Time Series Analysis; Forecasting and Control’,
Holden-Day, San Fransisco.
[6] Box George Edward Pelham, Gwilyn M. Jenkins and Reinsel G. C. (1976), ‘Time Series Analysis;
Forecasting and Control’, Holden-Day, San Fransisco (3).
[7] Chatfield Chris (2004),’The Analysis of Time Series: An Introduction’, John Wiley & Sons, NewYork, U.S.
3(1),69-71
[8] Glahn Harry R. and Dale A. Lowry (1972), ‘The Use of Model Output Statistics MOS in Objective Weather
Forecasting’, Journal of Applied Meteorology 11, 1203-121.
[9]Klein William H. and Frank Lewis (1970),’Computer Forecasts of Maximum and Minimum Temperature’,
Journal of Applied Meteorology 9,350-359.
[10]Kenya Meteorological Department, KMD (2009), Kenya Outlook for the March-May 2011”long rains”
Season’, Ministry of Environment and Mineral Resources.
[11]Pankratz Allan(1983),’Forecasting with Univariate Box-Jenkins Concept and Cases’, John Wiley & Sons,
Inc. New York 78(1), 684-709.
[12] Stock J. H. and Watson M. W. (1998), ‘Forecasting in Dynamic Factors Models Subject to Structural
Instability’, National Bureau of Economic Research 6(2), 98-102.
[13] George C. Tiao and Box G. E. P. (1975),’Intervention Analysis with Applications to Economic and
Enviromental Problems’, Journal of the American Statistical Association, 70(349), 70-79.
[14]Fadhilah Yusof and Ibrahim Lawal Kane (2012), ‘Modeling Monthly Rainfall Time Series Using ETS and
SARIMA Models’, International Journal of Current Research 4(1), 195-200.
APPENDIX
AIC BIC
ARIMA (1,0,1)×(0,0,0)12 586.16 599.17
ARIMA (1,0,1)×(0,0,1)12 563.66 579.93
ARIMA (1,0,1)×(1,0,0)12 547.46 563.72
Table1: Seasonal ARIMA models
Coeff. Std. error z p-value
Const. 3.843 0.1753 21.91 1.99e-106***
0.2039 0.1828 1.116 0.2645
Φ 0.4641 0.0670 6.924 4.38e-012***
0.1171 0.1806 0.6486 0.5166
Note: p-value <0.05 considered statistically significant

Vol.4, No.11, 2014
55
Table2: SARIMA model
Performance Statistics
ME -0.0053687
MSE 0.96794
RMSE 0.98384
MAE 0.75197
AIC 549.2842
BIC 565.5717
Table 3: performance Statistics
Figure 1: Precipitation Time Series Plot
0
50
100
150
200
250
300
350
400
450
1996 1998 2000 2002 2004 2006 2008 2010
y
Precipitation Time Series Plot

Vol.4, No.11, 2014
56
Figure 2: ACF and PACF plots of Precipitation
Figure 3: ACF and PACF plots of residuals
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0 5 10 15 20
lag
ACF for y
+- 1.96/T^0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0 5 10 15 20
lag
PACF for y
+- 1.96/T^0.5
-0.25
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0.2
0.25
0 5 10 15 20
lag
Residual ACF
+- 1.96/T^0.5
-0.25
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0.2
0.25
0 5 10 15 20
lag
Residual PACF
+- 1.96/T^0.5

Vol.4, No.11, 2014
57
Figure 4: Normality test of residuals
Figure 5: Residual Q-Q Plot
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
-4 -3 -2 -1 0 1 2 3
Density
uhat1
Normality test of Residuals
uhat1
N(-0.0053687,0.99423)
Test statistic for normality:
Chi-square(2) = 9.761 [0.0076]
-4
-3
-2
-1
0
1
2
3
-3 -2 -1 0 1 2 3
Normal quantiles
Q-Q plot for residual
y = x

Vol.4, No.11, 2014
58
Figure 6: Graph of Forecasts
Nomenclature
AIC: Akaike Information Criterion
BIC: Bayesian Information Criterion
SARIMA: Seasonal Autoregressive Integrated Moving Average
ME: Mean Error
MSE: Mean Squared Error
RMSE: Root Mean Squared Error
MAE: Mean Absolute Error

Business, Economics, Finance and Management Journals PAPER SUBMISSION EMAIL
European Journal of Business and Management EJBM@iiste.org
Research Journal of Finance and Accounting RJFA@iiste.org
Journal of Economics and Sustainable Development JESD@iiste.org
Information and Knowledge Management IKM@iiste.org
Journal of Developing Country Studies DCS@iiste.org
Industrial Engineering Letters IEL@iiste.org
Physical Sciences, Mathematics and Chemistry Journals PAPER SUBMISSION EMAIL
Journal of Natural Sciences Research JNSR@iiste.org
Journal of Chemistry and Materials Research CMR@iiste.org
Journal of Mathematical Theory and Modeling MTM@iiste.org
Advances in Physics Theories and Applications APTA@iiste.org
Chemical and Process Engineering Research CPER@iiste.org
Engineering, Technology and Systems Journals PAPER SUBMISSION EMAIL
Computer Engineering and Intelligent Systems CEIS@iiste.org
Innovative Systems Design and Engineering ISDE@iiste.org
Journal of Energy Technologies and Policy JETP@iiste.org
Information and Knowledge Management IKM@iiste.org
Journal of Control Theory and Informatics CTI@iiste.org
Journal of Information Engineering and Applications JIEA@iiste.org
Industrial Engineering Letters IEL@iiste.org
Journal of Network and Complex Systems NCS@iiste.org
Environment, Civil, Materials Sciences Journals PAPER SUBMISSION EMAIL
Journal of Environment and Earth Science JEES@iiste.org
Journal of Civil and Environmental Research CER@iiste.org
Life Science, Food and Medical Sciences PAPER SUBMISSION EMAIL
Advances in Life Science and Technology ALST@iiste.org
Journal of Biology, Agriculture and Healthcare JBAH@iiste.org
Journal of Food Science and Quality Management FSQM@iiste.org
Journal of Chemistry and Materials Research CMR@iiste.org
Education, and other Social Sciences PAPER SUBMISSION EMAIL
Journal of Education and Practice JEP@iiste.org
Journal of Law, Policy and Globalization JLPG@iiste.org
Journal of New Media and Mass Communication NMMC@iiste.org
Journal of Energy Technologies and Policy JETP@iiste.org
Historical Research Letter HRL@iiste.org
Public Policy and Administration Research PPAR@iiste.org
International Affairs and Global Strategy IAGS@iiste.org
Research on Humanities and Social Sciences RHSS@iiste.org
Journal of Developing Country Studies DCS@iiste.org
Journal of Arts and Design Studies ADS@iiste.org

The IISTE is a pioneer in the Open-Access hosting service and academic event management.
The aim of the firm is Accelerating Global Knowledge Sharing.
More information about the firm can be found on the homepage:
http://www.iiste.org
CALL FOR JOURNAL PAPERS
There are more than 30 peer-reviewed academic journals hosted under the hosting platform.
Prospective authors of journals can find the submission instruction on the following
page: http://www.iiste.org/journals/ All the journals articles are available online to the
readers all over the world without financial, legal, or technical barriers other than those
inseparable from gaining access to the internet itself. Paper version of the journals is also
available upon request of readers and authors.
MORE RESOURCES
Book publication information: http://www.iiste.org/book/
IISTE Knowledge Sharing Partners
EBSCO, Index Copernicus, Ulrich's Periodicals Directory, JournalTOCS, PKP Open
Archives Harvester, Bielefeld Academic Search Engine, Elektronische Zeitschriftenbibliothek
EZB, Open J-Gate, OCLC WorldCat, Universe Digtial Library , NewJour, Google Scholar

Forecasting precipitation using sarima model

More Related Content

What's hot (19)

Similar to Forecasting precipitation using sarima model (20)

More from Alexander Decker (20)

Forecasting precipitation using sarima model