Spatial Prediction of Rainfall using Universal Kriging Method: A Case Study of Mysuru District

DOI : 10.17577/IJERTCONV4IS20010

Download Full-Text PDF Cite this Publication

Text Only Version

Spatial Prediction of Rainfall using Universal Kriging Method: A Case Study of Mysuru District

M.S. Ganesh Prasad a ,

a Professor, Department of Civil Engineering, The National Institute of Engineering, Mysuru, Karnataka, India

Sushma N.b

bResearch Student, Department of Civil Engineering, The National Institute of Engineering, Mysuru, Karnataka, India

Abstract: Rainfall is a key factor for determining the sustainability for conservation of living species on the earth. Rainfall is an unpredictable and random phenomenon in nature. The change in rainfall pattern and its impact on surface water resources is an important climatic problem faced by society today. However, rainfall is the most elusive and uncertain phenomenon to predict. Geostatistical methods are now commonly used to analyze spatial variability of random phenomenon. In recent years, kriging interpolation technique is gaining more frequent use for many applications, since it gives better prediction than regression methods. The objective of this study is to predict rainfall for Mysuru district using universal kriging interpolation method. Data from a total of 21 raingauge stations of Mysuru district have been used for interpolating rainfall for the years 1994, 1999, 2004, 2009 and 2014. The results indicated that the spatial variation of rainfall in 1994, 2004 and 2009 varies from moderate to very high level whereas in 1999 and in 2014 the rainfall varies from moderate to high level. Based on the comparison of predicted rainfall and the actual observed rainfall at few of the stations, the prevalent results from universal kriging interpolation method adopted in the present study have been found to be satisfactory and encouraging.

Key words: Rainfall, Prediction, Interpolation, Geostatistics, Universal kriging.

  1. INTRODUCTION

    The need for water is increasing due to population growth, economic development, agricultural planning and urbanization. The amount of utilization of water for all these aspects depends on rainfall. Rainfall is a key factor for determining the sustainability and for conserving living species on the earth. Spatial and temporal variability of rainfall is due to natural internal process within the climate system or due to variations caused by human activities. The change in rainfall pattern and its impact on surface water resources is an important climatic problem faced by society today. The use of rainfall data is essential and fundamental for the rainfall- runoff process. Rainfall data of a particular area are analyzed using several methods. Many researchers have been using spatial statistics methods 1, 2, 3 and

    4 to predict rainfall and its variation. Universal kriging interpolation method has been used in many studies 4 ,

    5 and 6 to interpolate rainfall and has been found that it

    is most suitable when compared to other deterministic methods. The details of different methods of kriging may be found in 7, 8. The objective of the present study is to predict rainfall for Mysuru district in Karnataka state using Universal kriging.

  2. STUDY AREA AND DATA USED

    Mysuru is the third largest city in terms of population in the state of Karnataka, India. The total geographical area of Mysuru district is 6,241 Sq.km. The study area lies between the North latitudes 11° 44' N and 12° 37' N and

    East longitudes between 75° 57' E and 77° 12' E. The district is bound on the north by Mandya and part of Hassan districts and on the east by Chamarajanagar. Kodagu forms its western boundary and the southern portion is covered by Kerala and part of Chamarajanagar district. Figure 1 shows the location of study area. The climate of the district is moderate throughout the year, and the district gets rainfall during two seasons, namely, the southwest monsoon season or rainy season, which is between June to September and retreating monsoon season during October and November. The rainfall data from 25 rain gauge stations distributed across the study area has been used in the present study. Table 1 shows the details of rain gauge stations in the study area and their location.

    Figure 1: Location of study area

    Table 1: Rain gauge Stations along with their locations

    Sl.no

    Raingauge stations

    Latitude N

    (Degree)

    Longitude E

    (Degree)

    1

    H.D. kote

    12.088

    76.327

    2

    Bankavadi

    11.866

    76.300

    3

    Beervalu

    11.947

    76.462

    4

    Hampapura

    12.121

    76.475

    5

    Beechanahalli

    11.986

    76.351

    6

    Hunsur(trs)

    12.31

    76.289

    7

    Ratnapuri

    12.227

    76.320

    8

    Hirige

    12.221

    76.204

    9

    Bherya

    12.589

    76.349

    10

    Chunchanakatte

    12.502

    76.293

    11

    Saligrama

    12.560

    76.268

    12

    Hebbal

    12.359

    76.608

    13

    Mysuru city

    12.309

    76.640

    14

    Naganahalli

    12.378

    76.652

    15

    Elwala

    12.343

    76.583

    16

    Jayapura

    12.204

    76.553

    17

    Biligere

    12.149

    76.794

    18

    Kowlande

    12.001

    76.785

    19

    Bettadapura

    12.471

    76.104

    20

    Bailukuppa

    12.40

    75.981

    21

    T .Narasipura

    12.21

    76.899

    22

    Mugur

    12.12

    76.943

    23

    Talakadu

    12.19

    77.030

    24

    Piriyapatna

    12.33

    76.098

    25

    Nanjanagud

    12.11

    76.680

    Out of these, data from 21 stations have been used for interpolation purpose and the data from the remaining four stations are used for testing the accuracy of prediction. The rain gauge stations used for testing the predicted values are highlighted in Table 1.

  3. METHODOLOGY

    The present study includes interpolation of rainfall data in the form of point rain gauge readings from 25 rain gauge stations spread across the study area. The point rain gauge data are handled using MS EXCEL spreadsheets in order to calculate yearly rainfall and other statistical parameters. The analysis has been carried out using Geographic information system (GIS).

    Figure 2: Flowchart indicating methodology

    The overall methodology adopted in the present study is shown in Figure 2. The point rainfall data collected from the statistical depatment of Mysuru district has been analyzed and average annual rainfall has been computed. Base map showing the district boundary was digitized using toposheets. Layer showing the location of rain gauge stations was prepared in GIS platform.

  4. KRIGING INTERPOLATION METHOD

    In kriging interpolation, the variable to be mapped is considered as a regionalized variable and by computing a semivariogram, the spatial autocorrelation of the data is modeled. Once the semivariogram model is known for the data set, it is possible to estimate the value of the variable at any unmeasured location.

    The input data points (i.e. location data of raingauge stations and corresponding rainfall) are examined to construct a sample or experimental variogram which shows the degree of spatial autocorrelation as a function of distance between points. As the distance between points being compared is increased, the difference between them also increases. However, their differences become equal in value to the standard deviation after a certain distance and the semivariance no longer increases. The semivariogram becomes flat and the corresponding semivariance is called the sill. The distance at which this occurs is called the range of the regionalized variable. The range defines the region where the variable is spatially dependent. Another parameter of a variogram is the nugget, which is the variance at zero distance. Thus, the semi variogram model to be used plays an important role in kriging interpolation. Figure 3 shows a typical variogram for a random field with these parameters.

    (h)

    Sill

    nugget

    h

    Range (h)

    Figure 3: Typical variogram of a random field

    Figure 3: Spatial distribution of rainfall for the years 1994, 1999, 2004,

    2009 and 2014

    Universal kriging

    Universal kriging is applied when the mean is assumed to show a polynomial function of spatial coordinates 9. Generally, the mean value for spatial data cannot be assumed constant, since it depends on the location of the sample. For this purpose, the spatial universal kriging is used which aims to predict z(x) at unsampled location. It splits the random function into a linear combination of deterministic functions, the smoothly varying and nonstationary trend, or also called drift, and a random component representing the residual random function. In universal kriging more the number of sample data points used better the precision of universal kriging.

  5. RESULTS AND DISCUSSION

    Figure 3 shows the estimated spatial variation of rainfall categorized into four classes viz., Low (0-600mm), Moderate (600-800mm), High (800-1000mm) and Very high (1000-1500mm).The spatial variation of rainfall in 1994 varies from moderate to very high; In 1999 spatial variation of rainfall varies from moderate to high; in 2004 and 2009 spatial variation of rainfall varies from moderate to very high and spatial variation of rainfall of 2014 varies from moderate to high level. The prediction of year-wise spatial variations in rainfall shows that the study area has fluctuations in annual rainfall for the years 1994, 1999, 2004, 2009, 2014 which vary between slight and extreme rainfall.

    Figure 4: Estimated and observed values of rainfall for the years 1994, 1999, 2004, 2009 and 2014

    Correlation coefficients are determined using estimated and observed values. The predicted R 2 value (figure 4) for the year 1994, 2004 and 2009 is near to one which is more reliable when compared to the 1999 and 2014. Higher the R2 value, better the model fits the results.

    The spatial distribution of rainfall (figure 3) in 1994, 2004 and 2009 varies from moderate to very high values and shows correlation of 0.964, 0.889 and 0.910 (figure 4) ; whereas the spatial distribution of rainfall in 1999 and 2014 (figure 3) varies from low to moderate values and the R2 value shows correlation of 0.622 and 0.715 (figure 4). The number of data points used in this prediction is 21; the value of R2 may vary due to dependency of spatial interpolation locations and also on the correctness of sample rainfall data points.

  6. CONCLUSIONS

Rainfall is the most elusive and uncertain phenomenon to predict. Geostatistical methods are now commonly used to analyze spatial variability of rainfall which is considered as a random phenomenon. In this study an attempt has been made to predict rainfall in Mysuru district with the use of geostatistical method such as universal kriging. The prevalent results from universal kriging were found to be satisfactory and encouraging. Further more studies are required with more number of data points and the accuracy may be scrutinized. When interpolating with very few neighbourhood sample points care should be taken owing to the fact that this method can extrapolate outside the range of data values and cause a poor results. The study reported here has made aware of the need for further work in order to find the ways and means to improve the accuracy of rainfall input for interpolation techniques.

REFERENCES

  1. DongWoo Jang, HyoSeon Park and JinTak Choi Selection of Optimum Spatial Interpolation Method to Complement an Area Missing Precipitation Data of RCP Climate Change Scenario,2015.

  2. Tahersaud and AbdulmohsinAlshaik, Spatial Analysis of Rainfall in Southwest of Saudi Arabia using GIS, 1998.

  3. Goovaerts P, Geostatistical approaches for incorporating elevation into the spatial interpolation of rainfall 2000.

  4. Basistha, A., Arya, D. S., and Goel, N. K., Spatial Distribution of Rainfall in indian Himalayas A case study of Uttarakhand Region, Water Resour. Manag. 2008.

  5. Damijana Kastelec and Katarina Komelj, Spatial Interpolation of Mean Yearly Precipitation using Universal Kriging, 2002.

  6. S. Ly, C. Charles, and A. Degre ,Geostatistical interpolation of daily rainfall at catchment scale, the use of several variogram models in the Ourthe and Ambleve catchments, Belgium.,2011.

  7. Chiles and Delfner Comparison of kriging with external drift and regression- kriging, 1999

  8. Lloyd and Atkinson Geographic information systems ,2001.

  9. SarannLy, Catherine Charles, Aurore Degré , Different methods for spatial interpolation of rainfall data for operational hydrology and hydrological modeling at watershed scale 2012.

  10. L mitas and h mitasova, Spatial interpolation1996.

  11. Ishappa Muniyappa Rathod, Aruchamy.S Spatial Analysis of Rainfall Variation in Coimbatore District Tamilnadu using GIS2010.

  12. MarcG. Gentonand Reinhard Furrer, Analysis of Rainfall Data by Robust Spatial Statistics using S+Spatialstats 1998.

  13. N. Q. Hung, M. S. Babel, S. Weesakul, and N. K. Tripathi , An artificial neural network model for rainfall forecasting in Bangkok, Thailand 2009.

  14. Y. Knight. B. Yu, G. Jenkins and K. Morris, Comparing rainfall interpolation techniques for small subtropical urban catchments

  15. Alexandre Lanciani, Marta Salvati, Spatial interpolation of surface weather observations in Alpine meteorological service, 2008.

  16. Goutami Bandyopadhyay, The Prediction of Indian Monsoon Rainfall, A Regression Approach

  17. Goovaerts, P, Using elevation to aid the geostatistical mapping of rainfall erosivity 1999.

  18. Carrera-Hern´andez, J. J. and Gaskin, S. J , Spatio temporal analysis of daily precipitation and temperature in the Basin of Mexico, J. Hydrology 2007.

  19. Chow, V. T.: Handbook of applied hydrology A compendium of water-resources technology, McGraw-Hill, Inc. 1964.

  20. Lu, G. Y. and Wong, D. W., An adaptive inverse-distance weighting spatial interpolation technique, 2008.

  21. Verworn, A. and Haberlandt, U. Spatial interpolation of hourly rainfall effect of additional information, variogram inference and storm properties, Hydrol. Earth Syst. 2011.

  22. Vicent-Serrano, S. M. Saz-Sanchez, M. A., and Cuadrat, J. M. Comparative analysis of interpolation methods in the middle Ebro Valley (Spain): application to annual precipitation and temperature, 2003.

  23. Berne, A., Delrieu, G., Creutin, J.-D., and Obled, C. Temporal and spatial resolution of rainfall measurements required for urban hydrology, 2004.

  24. Boer, E. P. J., de Beurs, K. M., and Hartkamp, A. D Kriging and thin plate splines for mapping climate Variables, Int. J. Appl. Earth Obs, 2001.

  25. Borga M. and Vizzaccaro, A, On the interpolation of hydrologic variables: Formal equivalence of multiquadratic surface fitting and kriging, 1997.

Leave a Reply