THE ROLE OF OCEAN STATE INDICES IN SEASONAL AND INTER-ANNUAL CLIMATE VARIABILITY OF THAILAND Manfred Koch and Werapol Bejranonda Department of Geohydraulics and Engineering Hydrology, University of Kassel, Germany Email: kochm@uni-kassel.de Keywords: Ocean state indices, climate variability, Thailand, correlations, teleconnections, prediction
Abstract Thailand has a long coastline with the Pacific Ocean as well as with the Indian Ocean. Because this peculiar location, Thailand s local climate and its surface water resources are strongly influenced by the mix of tropical dry and tropical wet seasons which, in turn, depend themselves upon the thermal states of the Pacific and Indian Oceans. Because of this teleconnective oceanic impact on local climate variations in different regions of Thailand, the somewhat extreme climate pattern Thailand has seen in recent years, namely,long periods of drought in some provinces must be seen in this light, i.e. may not be attributed alone to long-term predictions of climate change from global climate models for the region. With a better understanding of these ocean-thai seasonal-weather-pattern teleconnections, water authorities in Thailand would get a tool at hand to forecast short-term extreme seasonal climate pattern in a particular region allowing them to better manage the adequate supply of surface water resources to the users. In this study the spatial and temporal relationships, i.e. teleconnections between the oceanic climates and the regional weather in Thailand are assessed by various techniques of stochastic time series analysis. Time series of the sea surface temperature (SST) and various ocean indices of the Pacific and the Indian Oceans as well as the time series of 121 meteorological stations from 5 regions across Thailand that include humidity, evaporation, temperature and rainfall during 1971-2007 are examined using cross-correlation, linear-regression and wavelet transform methods. The results of the analyses show that the El-Niño 1+2 SST anomaly index of the Pacific Ocean correlates the strongest with the Thai local climate. The most sensitive parameters to the ocean indices are the minimum temperature at stations in the northern and northeastern inland regions of Thailand, and the number of rainy days in the eastern, central and southern, and coastal regions. In the south the rainfall varies positively with El-Niño along the Gulf of Thailand, but negatively along the Andaman Sea, with maximal correlations at lag-times of 2-4 months. Using the ocean indices as external regressors, the classical ARIMA-forecast (no external regressors) of the local climate variables temperature and rainfall could be improved significantly for the 2000-2007 time period, with the NS-coefficient increasing from originally 0.51 to 60 and 0.32 to 0.44 for minimum and maximum temperatures, respectively. However, the rainfall is not better predicted when the El Niño 1+2 SST ARIMA-EX model is employed.
Overview 1. INTRODUCTION 1.1 Study area 1.2 Objectives 2. CLIMATE ANALYSIS 2.1 Wavelet analysis 2.2 Cross-correlation 2.3 Regression analysis 3. TELECONNECTIONS 4. APPLICATION FOR CLIMATE PREDICTION 3.1 Pilot stations for application of teleconnection 3.1 Teleconnections in long-range prediction 5. CONCLUSIONS
1. INTRODUCTION 1.1 Study area (1) NN NE CC EE Indian Ocean 121 meteorological stations Pacific Ocean ± SS 0 40 80 160 240 320 400 Kilometers meteorological station province border Thailand borderline distributed into 5 regions North (NN) Northeast (NE) Central (CC) East (EE) South (SS)
1. INTRODUCTION 1.1 Study area (2) Ocean state indexes / oscillation Indian Pacific
1. INTRODUCTION 1.2 Objectives Identify teleconnection existence relationship local weather - regional index by employing statistical framework Predict local weather towards the impact of climate change on water/climate by using regional indices Methods Wavelet analysis; cross-correlation, regression analysis, ARIMA models Goal To investigate the climate change impact on water resources prediction improvement by using regional indices
2. CLIMATE ANALYSIS / TELECONNECTIONS 2.1 Wavelet analysis min temp max temp
2. CLIMATE ANALYSIS / TELECONNECTIONS 2.2 Cross-correlation (1) Average cross-correlation coefficients between ocean indices and the basic meteorological time series in EE region showing Niño relates highly with the climate variables in Thailand Indices avg.evap avg.rh extr.max extr.min max.rain 24hr mean.max temp mean.min temp mean.temp Rain mm/day Indian-SETIO 0.02 0.00 0.14 0.09-0.02 0.17 0.09 0.13 0.00 0.05-0.01 Indian-SWIO -0.14-0.02 0.07 0.04 0.03 0.08 0.06 0.05 0.00 0.07 0.02 Indian-WTIO -0.07 0.00 0.20 0.13-0.02 0.23 0.14 0.19-0.04-0.02-0.01 Pacific-EP -0.20 0.44-0.20 0.47-0.07-0.43 0.50 0.35 0.22-0.09-0.42 Pacific-Nino1.2 0.34-0.61 0.43-0.15-0.34 0.64 0.24 0.58-0.21-0.76-0.42 Pacific-Nino3 0.35-0.40 0.45-0.19-0.30 0.60-0.04 0.60-0.17-0.63-0.36 Pacific-Nino3.4 0.25-0.35 0.29 0.13-0.22 0.46 0.25 0.48-0.13-0.42-0.31 Pacific-Nino4 0.13-0.16 0.20 0.32-0.15 0.23 0.31 0.33-0.03-0.31-0.25 Pacific-NOI -0.04 0.00-0.14-0.04 0.06-0.16-0.09-0.18 0.01 0.10-0.01 Pacific-PDO 0.12-0.06 0.18 0.18 0.02 0.20 0.17 0.20 0.05-0.16-0.06 Pacific-PNA -0.02-0.06 0.06 0.06-0.06 0.08 0.12 0.17-0.07-0.01 0.01 Pacific-SOI -0.09 0.08-0.12-0.11 0.09-0.20-0.13-0.19 0.01 0.13 0.10 Pacific-WP 0.06-0.13 0.07 0.02-0.09 0.03 0.06 0.01-0.09-0.05-0.07 rainy.day total.rain
2. CLIMATE ANALYSIS / TELECONNECTIONS 2.2 Cross-correlation (2) Cross-correlation of four Niño SST indices with local monthly minimum, maximum temperature and rainfall station 478201 in EE region during 1971-2006 Minimum temperature Maximum temperature Rainfall Niño 4 Niño 3.4 Niño 3 Niño 1+2 Lag time (month)
2. CLIMATE ANALYSIS / TELECONNECTIONS Average cross-correlation coefficient between El Nino 1+2 index with the basic meteorological time series of 121 stations in 5 regions showing El-Nino related highly with the evaporation and temperature in Thailand 2.2 Cross-correlation (2)! 5! 5 NN! 5! 5! 5 5! NE! 5! 5 5! 5! 5 CC EE SS ± 160 240! 5 320 400 Kilometers meteorological station province border Thailand borderline region cc ee ne nn ss avg. evap 0.61 0.34 0.52 0.71 0.58 avg. rh -0.54-0.61-0.49-0.75-0.14 extr. max extr. min max. rain. 24hr 0.72 0.43 0.77 0.75 0.52-0.45-0.15-0.49-0.79-0.36-0.43-0.34-0.37-0.53-0.06 mean mean mean.max..min..temp temp temp rain. mm. day. 0.74-0.28 0.56-0.22 0.64 0.24 0.58-0.21 0.72-0.60 0.67-0.49 0.72-0.74 0.21-0.40 ENSO 1+2 (Nino 0.65-0.07 0.671.2)0.00 rainy. total. day rain -0.75-0.76-0.61-0.73-0.53-0.52-0.42-0.21-0.61-0.11
2. CLIMATE ANALYSIS / TELECONNECTIONS 2.3 Regression analysis Regression of El Niño 1+2 and observed minimum temperature at station 478201 during 1971-2005 showing coefficients of determination (R 2 ) by linear-regression of annual and seasonal values (a = annual, s1=dry season [Oct-Dec] in grey dots, s2= pre-monsoon [Jan-Mar] in blue dots, s3= monsoon1 [Apr-Jun] in red dots and s4= monsoon2 [Jul-Sep] in green dots)
3. TELECONNECTIONS (1) Cross-correlation of minimum temperature versus Nino1+2 optimal crosscorrelation coef. lag (month) at optimal correlation -0.8-3 ± 0 50 100 200 300 400 Kilometers +0.7 Max Cross-correlation coeff. Nino 1+2 vs Min temp. -1.0-0.9 +0.1 +0.2-0.8 +0.3 ± -0.7 +0.4-0.6 +0.5 0.0-0.5 +0.6-0.4 +0.7-0.3 +0.8-0.2 +0.9-0.1 +1.0 Thailand borderline 0 50 100 200 300 400 Kilometers Nino 1+2 took 3 months before temp -3 Time-lag (month) at best cross-correlation Nino 1+2 vs Min Temp. -11-10 -9-8 -7-6 -5-4 -3-2 -1 0 +1 +2 +3 +4 +5 +6 +7 +8 +9 +10 +11 Thailand borderline
3. TELECONNECTIONS (2) -0.5 Cross-correlation of monthly rainfall versus Nino1+2 Correlation coef. +6 [NINA took 3 months after rain] Time-lag (month) at optimal correlation ± 0 37.5 75 150 225 300 375 Kilometers -0.3 +0.5 Autocorrelation coeff (within -/+11 months) Monthly rainfall vs NINA 1.2-1.0-0.9-0.8-0.7-0.6-0.5-0.4-0.3-0.2-0.1 0.0 Thailand borderline +0.1 +0.2 +0.3 +0.4 +0.5 +0.6 +0.7 +0.8 +0.9 +1.0 ± 0 37.5 75 150 225 300 375 Kilometers -3 Time-lag (month) at best autocorrelative Monthly rainfall vs NINA 1.2-11 -10-9 -8-7 -6-5 -4-3 -2-1 0 +1 +2 +3 +4 +5 +6 +7 +8 +9 +10 +11 Thailand borderline
3. TELECONNECTIONS (3) Cross-correlation of monthly rainfall vs Nino1+2 +0.5 Correlation coef. EP / NP Autocorrelation coeff (within -/+11 months) Mean temp. vs EP -1.0 +0.1 EP / NP : Eastern Pacific Northern Pacific Oscillation -0.9 +0.2-0.8 +0.3 ± -0.3-0.7-0.6-0.5-0.4-0.3-0.2 0.0 +0.4 +0.5 +0.6 +0.7 +0.8 +0.9 Indian ocean side : inverse variation Pacific ocean side : direct variation -0.1 +1.0 0 40 80 160 240 320 400 Kilometers Thailand borderline
4. APPLICATION FOR CLIMATE PREDICTION 4.1 Pilot stations for application of teleconnections Rayong province : Seaboard Autocorrelation coeff (within -/+11 months) Mean min. temp. vs NINA 1.2-1.0-0.9-0.8 +0.1 +0.2 +0.3 Time lag -1 to -3 months from ENSO 0 El Nino 1 El Nino 34 El Nino 4-0.7 +0.4 ± 0 37.5 75 150 225 300 375 Kilometers -0.6 +0.5 0.0-0.5 +0.6-0.4 +0.7-0.3 +0.8-0.2 +0.9-0.1 +1.0 Thailand borderline Timg lag (month) -1-2 -3 Min temp Max temp
4. APPLICATION FOR CLIMATE PREDICTION 4.2 Teleconnections in long-range prediction (1) Verification of minimum temperature prediction up to year 2006 using various ARIMA models, based on the calibration period 1974-2000 ARIMA (1,1,2)(2,0,2)[12] ARIMA model NS coef.= 0.773 Temperature ( C) ARIMA (1,1,2)(2,0,2)[12] ARIMA (1,1,2)(2,0,1)[12] ARIMA model with Niño 1+2 (predict 2 months advance) NS coef.= 0.837 ARIMA model with CGCM3 NS coef.= 0.822 Time (year)
4. APPLICATION FOR CLIMATE PREDICTION 4.2 Teleconnections in long-range prediction (2) Model accuracy in forward prediction by AR-models the Nash Sutcliffe coefficient 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 calibration verification prediction (1 yr) No Index No Index With Niño 0.1 0.0 MxT MnT Rn MxT MnT Rn MxT MnT Rn AR ARIMA ARIMA with external regression Model
4. APPLICATION FOR CLIMATE PREDICTION 4.2 Teleconnections in long-range prediction (3) 1.0 Nash-Sutcliffe model efficiency coefficient of monthly maximum temperature prediction at station 48459 in Khongyai Basin 1.0 Nash-Sutcliffe model efficiency coefficient of monthly minimum temperature prediction at station 48459 in Khongyai Basin 0.8 0.8 Nash Sutcliffe efficiency 0.6 0.4 0.2 0.0 trend analysis trend analysis Nash Sutcliffe efficiency 0.6 0.4 0.2 0.0 trend analysis trend analysis using nino data Calibration Verification (scenario A1B) Verification (scenario A2) Calibration Verification (scenario A1B) Verification (scenario A2) -0.2-0.2 Nash Sutcliffe efficiency 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 Nash-Sutcliffe model efficiency coefficient of monthly rainfall prediction in Khongyai Basin Calibration Verification (scenario A1B) Verification (scenario A2) trend analysis trend analysis
4. APPLICATION FOR CLIMATE PREDICTION 4.2 Teleconnections in long-range prediction (4) Nash-Sutcliff coefficient
5. CONCLUSIONS The Niño indices mostly relate to local Thai climate variables and local geography The strongest influence of Niño 1+2 SST on the local climate is in the premonsoon season (Jan-Mar) Teleconnection of SST and regional climate is able to improve the prediction of climate in Thailand, particularly the temperature Auto- and co-regressive models allow to create a long-range time series of local climate along with the change of SST Using a combination of selected predictors in global-scale and regional/local climate indicators provides for better long-term forecasting