Modeling Method for Temperature Anomaly Analysis

Kittisak Kerdprasop

, Paradee Chuaybamroong

and Nittaya Kerdprasop

Data and Knowledge Engineering Research Unit, School of Computer Engineering,

Suranaree University of Technology, Nakhon Ratchasima, Thailand

Department of Environmental Science, Thammasat University, Thailand

Keywords: Environmental Analytics, Temperature Anomaly, Reanalysis Data, Chi-Squared Automatic Interaction

Detection, CHAID Algorithm.

Abstract: This study applies intelligent analytical methods to analyze temperature anomaly events during the past

seven centuries of countries in the Southeast Asia including Thailand, Malaysia, Myanmar, and Cambodia.

The temperature reconstruction during the years 1300 to 1999 were used as data source for anomaly

analysis. In the analytical process, correlation analysis was applied to initially investigate the temperature

variability concordance among the Southeast Asian countries. The results are that temperature variability

patterns in Thailand, Myanmar, and Cambodia are moderately correlated to each other. On the contrary, the

temperature variation patterns of Malaysia do not correlate to other countries in the same region. The further

in-depth analysis focuses on the temperature anomaly of Thailand that shows high variability from the 14

to 16

centuries. Several machine learning algorithms had been applied to estimate the temperature anomaly

of Thailand based on the anomaly events among the neighbors. The learned models reveal that Myanmar

temperature anomaly most associate to the Thailand’s temperature variation. The performance of each

model had been assessed and the results reveal that the chi-squared automatic interaction detection, or

CHAID, is the best one with 0.624 correlation coefficient and relative error around 0.611.

1 INTRODUCTION

Climate change has been reported to have strong

influence over various natural dangers such as global

wildfires (Jolly et al., 2015), major volcanic

eruptions (Fujiwara et al., 2015), intense tropical

cyclones (Wing, Emanuel, and Solomon, 2015),

mega-heatwave (Sánchez-Benítez et al., 2018), and

extreme cold (Hartmann, 2015; Liu et al., 2015).

Temperature and precipitation anomalies are two

important factors to estimate climate changes. To

assess climate variation and trends, researchers

deploy several interpolating techniques, for instance,

analyzing the stratospheric temperature change

(Seidel et al., 2016), estimating the Antarctic and

Arctic surface air temperature anomalies over land

and sea ice (Comiso et al., 2017; Dodd et al., 2015;

Francis and Vavrus, 2015; Turner et al., 2016),

examining the cloud amount anomalies (Liu and

Key, 2016), and observing wind and temperature

over the ocean surface (Dong and Dai, 2015;

England et al., 2014; Randel and Wu, 2015).

These techniques require temperature record as a

major source of information for the climate variation

assessment. Temperature reading using thermometer

from the ground-based weather stations and

instrumental reading from ships and buoys are

common form of temperature data acquisition. But

the major shortcoming of this kind of data source is

that the instrumental data are available for only the

past one or two centuries.

To observe temperature trends and variations

over a long period spanning across several centuries

or a millennium, scientists have to rely on some

forms of natural proxy records such as tree rings

(Cai et al., 2018; Seim, 2016) and sediments from

lakes (Li et al., 2017; McColl, 2016). Such natural-

based reconstruction data are now complemented

with the state-of-the-art reanalysis technique that

combines instrumental record with satellite

observations to form an atmospheric data set suitable

for studying climate change (Cowtan and Way,

2014; Donat and Sillmann, 2014; Kobayashi et al.,

2015; Saha et al., 2014; Simmons et al., 2017; Xu et

al., 2018). Reanalysis data are now widely adopted

for observing temperature trends in many areas

274

Kerdprasop, K., Chuaybamroong, P. and Kerdprasop, N.

Modeling Method for Temperature Anomaly Analysis.

DOI: 10.5220/0007224902740280

In Proceedings of the 10th Inter national Joint Conference on Computational Intelligence (IJCCI 2018), pages 274-280

ISBN: 978-989-758-327-8

globally (Kern et al., 2016; Song et al., 2016; Way

and Bonnaventure, 2015).

In this work, we use reanalysis data of surface

temperature anomaly in eastern and south-central

Asia (Shi et al., 2015) to analyze the anomaly

association patterns among four countries in the

Southeast Asia. We apply correlation analysis and

machine learning techniques to capture the anomaly

association patterns. The applied machine learning

techniques include artificial neural network (ANN),

classification and regression tree (CART), and chi-

squared automatic interaction detection (CHAID).

Machine learning has recently been applied to the

climatology domain, but the technique is limited to

cluster analysis (Horton et al., 2015; Kretschmer et

al., 2018). This work introduces a classification

scheme to support the work of climatologists as well

as to expand the frontier of climate change study.

2 ANOMALY ANALYSIS

METHODOLOGY

2.1 Area of Study

We focus our anomaly analysis on the neighborhood

countries of Thailand sharing some common

characteristics based on the climatic type (Figure 1).

Thailand locates at 102.5 longitude and 17.5

latitude. Country in the northwest is Myanmar

(102.5 longitude, 2.5 latitude) with the same tropical

wet and tropical wet and dry climate zones as in the

north and the west parts of Thailand. Cambodia in

the east (107.5 longitude, 12.5 latitude) is in the

tropical wet and dry zone sharing the same climate

type as the northeastern of Thailand. Malaysia in the

south (102.5 longitude, 17.5 latitude) is in the

tropical wet zone as most southern part of Thailand.

2.2 Temperature Anomaly Analysis

Steps

To study the temperature anomaly patterns of

countries in the Southeast Asia, we perform the

following steps of data analytics:

Step 1: Data Extraction. The temperature

reconstruction data during the rainy season (June-

July-August) of the four countries are extract from

the original data set that contains surface

temperature anomaly of 126 countries in the east and

central Asia. These data had been reconstructed in

2015 by Feng Shi from China and his international

team using hundreds of proxy climate data (Shi et

al., 2015) Data are made publicly available by the

National Centers for Environmental Information

(http://ncdc.noaa.gov/ paleo/study/18635).

Step 2: Correlation Analysis. Surface

temperature anomalies of the selected four countries

during the years 1300 to 1999 are analyzed with

Pearson correlation to explore their association of

anomaly event occurrence.

 Tropical wet and dry

 Tropical wet

Figure 1: Geographical map of the study area in Southeast

Asia (shown on the above map) covering (1) Myanmar,

(2) Thailand, (3) Cambodia, and (4) Malaysia, with the

climate chart (on the bottom) showing the two weather

styles of this region: tropical wet along the coastal areas of

Myanmar, Thailand, and Malaysia and tropical wet and

dry in the mainland regions. (sources: http://www.

nationsonline.org/oneworld/map/physical_world_map_32

00.htm and http://www.asiafastfacts.com/asiaclimate.

html).

Step 3: Predictive Model Building. We apply

five learning algorithms to construct a predictive

model with Thailand’s temperature anomaly as a

Modeling Method for Temperature Anomaly Analysis

275

target of the model. These algorithms are ANN,

CART, CHAID, linear regression, and generalized

linear model.

Step 4: Model Evaluation. The five models are

assessed based on their correlation metric and

relative error on predicting the target event. The best

model with the highest correlation and the lowest

error is to be reported as the temperature anomaly

estimator.

3 ANALYSIS RESULTS

3.1 Correlation Analysis Result of

Temperature Variability

From the exploration of temperature anomalies

among the four Southeast Asian countries

(summarized in Table 1), we found that temperature

in Cambodia is the most fluctuate one with the

variance as high as 0.171. Cambodia also shows the

cold period with its minimum temperature anomaly

at -1.688

C. The country showing clearly the warm

period during the past millennium is Malaysia with

the mean temperature anomaly at 0.093

C. In the

century while Cambodia was in the cold phase,

Malaysia was in the warm phase (as shown in Figure

2).

Table 1: Temperature anomaly statistics.

Region /

Country

Temperature Anomaly (

Min Max Mean Variance

Eastern and

south-central

Asia (E&SC

Asia)

-0.766 0.089 -0.323 0.030

Malaysia

(MAL)

-0.724 0.831 0.093 0.083

Cambodia

(CAM)

-1.688 0.284 -0.613 0.171

Myanmar

(MYR)

-0.524 0.446 -0.027 0.038

Thailand

(THA)

-1.315 0.783 -0.243 0.113

Figure 2: Temperature anomaly comparison of eastern and

south central Asia against anomalies in Malaysia,

Cambodia, Myanmar, and Thailand.

The association of temperature anomaly patterns

through the correlation analysis (as displayed in

Table 2) is the result from the second step of our

analysis. The strongest association pattern through

Pearson’s correlation is the temperature anomalies

between Thailand and Myanmar. Malaysia shows

weak correlated temperature patterns to other

neighboring countries. Instead, among the four

regional countries, temperature pattern of Malaysia

is closest to the east and central Asia with

correlation coefficient 0.125, whereas Cambodia

shows opposite direction of pattern.

Table 2: Pearson correlation of temperature anomaly.

E&SC

Asia

MAL CAM MYR THA

E&SC

Asia

-- 0.125

-0.175

0.051 0.009

MAL 0.125

-- 0.034 0.107 0.044

CAM

-0.175 0.034 -- 0.070

0.313

MYR

0.051 0.107 0.070 --

0.549

THA

0.009 0.044 0.313

0.549

3.2 Temperature Estimation Model

The five machine learning algorithms that have been

used to model temperature anomaly association

among Thailand and the other three neighboring

countries in the region are assessed their

performances based on the correlation coefficient

and the relative error. Results are summarized in

Table 3.

IJCCI 2018 - 10th International Joint Conference on Computational Intelligence

276

Table 3: Performance comparison of estimation models.

Model Correlation

coefficient

Relative

error

CHAID 0.624 0.611

CART 0.611 0.627

ANN 0.575 0.673

Linear Regression 0.559 0.688

Generalized Linear Model 0.558 0.688

It can be seen from the results that CHAID is the

best machine learning algorithm to estimate

temperature anomaly of Thailand based on

anomalies of the neighbors. The CHAID model is

shown in Figure 3.

CHAID is a tree-based machine learning

algorithm that grows tree and split data set into

subsets based on the result from the chi-square test

(Kass, 1980). The tree is to be interpreted from the

root node on the left-hand-side to reach a

conclusion, which is the target node on the right-

hand-side. From Fig. 3, the interpretation of this tree

model to estimate temperature anomaly (TA) in

Thailand is as follows.

• In case of TA in Myanmar ≤ -0.326, the TA in

Thailand is around -0.643.

• In case of TA in Myanmar > -0.326 but less

than or equal to -0.212, the TA in Thailand is

around -0.486.

• In case of TA in Myanmar > -0.212 but less

than or equal to -0.058, also taking into account

TA in Cambodia:

− If the TA in Cambodia ≤ 0, then TA in

Thailand is expected to be around -0.319.

− But if the TA in Cambodia > 0, then TA in

Thailand is expected to be around 0.130.

• In case of TA in Myanmar > -0.058 but less

than or equal to 0.044, taking into account TA

in Cambodia:

− If the TA in Cambodia ≤ -0.939, then TA in

Thailand is around -0.259.

− If the TA in Cambodia > -0.939 but less

than or equal to -0.378, then TA in

Thailand is around -0.343.

Predictive Factors

Thailan

Temperature

nomal

MYR: ≤ -

0.326



-0.643

MYR:

(-0.326, -

0.212]



-0.486

MYR:

(-0.212, -

0.058]

CAM: ≤



-0.319

MYR:

(-0.212, -

0.058]

CAM: >



0.130

MYR:

(-0.058,

0.044]

CAM: ≤

-0.939



-0.259

MYR:

(-0.058,

0.044]

CAM:

(-0.939,

-0.378]



-0.343

MYR:

(-0.058,

0.044]

CAM: >

-0.378

MAL:≤ -0.192



-0.027

MYR:

(-0.058,

0.044]

CAM: >

-0.378

MAL:

(-0.192, 0.247]



-0.201

MYR:

(-0.058,

0.044]

CAM: >

-0.378

& MAL: > 0.247



0.022

MYR:

(0.044,

0.149]



-0.117

MYR: >

0.149

MAL: ≤

-0.192



0.114

MYR: >

0.149

MAL:

(-0.192,

0.181]

CAM: ≤ -

0.939



-0.337

MYR: >

0.149

MAL:

(-0.192,

0.181]

CAM:

(-0.939, -

0.003]



0.037

MYR: >

0.149

MAL:

(-0.192,

0.181]

CAM: > -

0.003



-0.090

Figure 3: CHAID model for estimating temperature

anomaly of Thailand based on the neighboring anomalies.

Modeling Method for Temperature Anomaly Analysis

277

Predictive Factors

Thailan

Temperature

nomal

MYR: >

0.149

MAL:

(0.181,

0.247]



0.135

MYR: >

0.149

MAL:

(0.247,

0.332]



-0.283

MYR: >

0.149

MAL: >

0.332



-0.007

Figure 3: CHAID model for estimating temperature

anomaly of Thailand based on the neighboring anomalies

(cont.).

− But if the TA in Cambodia > -0.378, then

also consider the TA in Malaysia:

If TA in Malaysia ≤ -0.192, then TA in

Thailand is around -0.027.

If TA in Malaysia > -0.192 but less

than or equal to 0.247, then TA in

Thailand is around -0.201.

If TA in Malaysia > 0.247, then TA in

Thailand is around 0.022.

• In case of TA in Myanmar > 0.044 but less than

or equal to 0.149, the TA in Thailand is around

-0.117.

• In case of TA in Myanmar > 0.149, also taking

into account TA in Malaysia:

− If the TA in Malaysia ≤ -0.192, then TA in

Thailand is around 0.114.

− If the TA in Malaysia > -0.192 but less than

or equal to 0.181, then also consider TA in

Cambodia:

If TA in Cambodia ≤ -0.939, then TA

in Thailand is around -0.337.

If TA in Cambodia > -0.939 but

less than or equal to -0.003,

then TA in Thailand is around

0.037.

If TA in Cambodia > -0.003, then TA

in Thailand is around -0.090.

− If the TA in Malaysia > 0.181 but less than

or equal to 0.247, then TA in Thailand is

around 0.135.

− If the TA in Malaysia > 0.247 but less than

or equal to 0.332, then TA in Thailand is

around -0.283.

− If the TA in Malaysia > 0.332, then TA in

Thailand is around -0.007.

4 CONCLUSIONS

This research presents the statistical and machine

learning approaches to learn correlated and

associated patterns from historical temperature

anomaly events among countries in the Southeast

Asia including Myanmar, Thailand, Cambodia, and

Malaysia. The temperature anomaly data used in this

work are obtained from the multi-proxy

reconstruction of east and south-central Asia during

June-July-August of the past millennium between

the years 1300-1999 C.E.

Correlation analysis results reveal that climate

variations in Myanmar and Thailand closely

resemble, but anomaly events in Malaysia are quite

different from other countries. From the temperature

anomaly record of Cambodia, the cold events during

the 18

century are noticeable and contrasting to the

warm events in Malaysia within the same timeframe.

Machine learning methodology is further applied

to study associative patterns of temperature

variations across countries. Such patterns are to be

analyzed through modeling within the classification

and regression framework. The results from

applying five algorithms to induce patterns with

numeric target, which is the temperature anomaly of

Thailand, reveal that CHAID algorithm is the best

one. The CHAID model employs temperature

anomaly in Myanmar as the first factor to estimate

temperature anomaly in Thailand. In case of

complicate estimation, the model takes temperature

anomaly of Cambodia as the second factor. This is

in accordance with the correlation analysis results

that Thailand’s temperature anomalies closely

correlate to the anomalies in Myanmar and

Cambodia. But the CHAID model provides more

information than the correlation analysis in that the

model can quantify the conditional temperature

anomalies necessary for making accurate estimation

over the target’s temperature anomalies.

ACKNOWLEDGEMENTS

This work was financially supported by grants from

the Thailand Toray Science Foundation, the National

Research Council of Thailand, and Suranaree

University of Technology through the funding of the

Data and Knowledge Engineering Research Units.

IJCCI 2018 - 10th International Joint Conference on Computational Intelligence

278

REFERENCES

Cai, Q., Liu, Y., Duan, B., Sun, C., 2018. Regional

difference of the start time of recent warming in

Eastern China: prompted by a 165-year temperature

record deduced from tree rings in the Dabie

Mountains. Climate Dynamics, vol. 50, pp. 2157–

2168.

Comiso, J.C., Gersten, R.A., Stock, L.V., Turner, J., Perez,

G.J., Cho, K. 2017. Positive trend in the Antarctic sea

ice cover and associated changes in surface

temperature. Journal of Climate, vol. 30, pp. 2251–

2267

Cowtan, K., Way, R.G., 2014. Coverage bias in the

HadCRUT4 temperature series and its impact on

recent temperature trends. Quarterly Journal of the

Royal Meteorological Society, vol. 140, pp. 1935–

1944

Dodd, E.M.A., Merchant, C.J., Rayner, N.A., Morice,

C.P., 2015. An investigation into the impact of using

various techniques to estimate Arctic surface air

temperature anomalies. Journal of Climate, vol. 28,

article 21.

Donat, M.G., Sillmann, J., 2014. Consistency of

temperature and precipitation extremes across various

global gridded in situ and reanalysis datasets. Journal

of Climate, vol. 27, article 17.

Dong, B., Dai, A., 2015. The influence of the interdecadal

Pacific oscillation on temperature and precipitation

over the globe. Climate Dynamics, vol. 45, pp. 2667–

2681.

England, M.H., McGregor, S., Spence, P., Meehl, G.A.,

Timmermann, A., Cai, W., Gupta, A.S., McPhaden,

M.J., Purich, A., Santoso, A., 2014. Recent

intensification of wind-driven circulation in the Pacific

and the ongoing warming hiatus. Nature Climate

Change, vol. 4, pp. 222–227.

Francis, J.A., Vavrus, S.J., 2015. Evidence for a wavier jet

stream in response to rapid Arctic warming.

Environmental Research Letters, vol. 10, article

014005.

Fujiwara, M., Hibino, T., Mehta, S.K., Gray, L., Mitchell,

D., Anstey, J., 2015. Global temperature response to

the major volcanic eruptions in multiple reanalysis

datasets. Atmospheric Chemistry and Physics

Discussions, vol. 15, pp. 13315–13346.

Hartmann, D.L., 2015. Pacific sea surface temperature and

the winter of 2014. Geophysical Research Letters, vol.

42, pp. 1894–1902.

Horton, D.E., Johnson, N.C., Singh, D., Swain, D.L.,

Rajaratnam, B., Diffenbaugh, N.S., 2015. Contribution

of changes in atmospheric circulation patterns to

extreme temperature trends. Nature, vol. 522, pp. 465–

469.

Jolly, W.M., Cochrane, M.A., Freeborn, P.H., Holden,

Z.A., Brown, T.J., Williamson, G.J., Bowman,

D.M.J.S., 2015. Climate-induced variations in global

wildfire danger from 1979 to 2013. Nature

Communications, vol. 6.

Kass, G.V., 1980. An exploratory technique for

investigating large quantities of categorical data.

Applied Statistics, vol. 29, no. 2, pp. 119–127.

Kern, Z., Németh, A., Horoszné Gulyás, M., Popa, I.,

Levanič, T., Hatvani, I.G., 2016. Natural proxy

records of temperature- and hydroclimate variability

with annual resolution from the Northern Balkan–

Carpathian region for the past millennium – Review &

recalibration. Quaternary International, vol. 415, pp.

109–125.

Kobayashi, S., Ota, Y., Harada, Y., Ebita, A., Moriya, M.,

Onoda, H., Onogi, K., Kamahori, H., Kobayashi, C.,

Endo, H., Miyaoka, K., Takahashi, K., 2015. The

JRA-55 reanalysis: General specifications and basic

characteristics. Journal of the Meteorological Society

of Japan, Ser. II, vol. 93, pp. 5–48.

Kretschmer, M., Coumou, D., Agel, L., Barlow, M.,

Tziperman, E., Cohen, J., 2018. More-persistent weak

stratospheric polar vortex states linked to cold

extremes. Bulletin of the American Meteorological

Society, vol. 99, pp. 49–60.

Li, J., Kong, L., Yang, H., Wang, Q., Yang, X., Shen, J.,

Zhao, C., 2017. Temperature variations since 1750 CE

inferred from an alpine lake in the southeastern margin

of the Tibetan Plateau. Quaternary International, vol.

436, pp. 37–44.

Liu, Y., Key, J.R., 2016. Assessment of Arctic cloud cover

anomalies in atmospheric reanalysis products using

satellite data. Journal of Climate, vol. 29, pp, 6065–

6083.

Liu, Z., Jian, Z., Yoshimura, K., Buenning, N.H., Poulsen,

C.J., Bowen, G.J., 2015. Recent contrasting winter

temperature changes over North America linked to

enhanced positive Pacific-North American pattern.

Geophysical Research Letters, vol. 42, pp. 7750–7757.

McColl, J.L., 2016. Climate variability of the last 1000

years in the NW Pacific: high resolution, multi-

biomarker records from Lake Toyoni. PhD Thesis,

University of Glasgow, U.K., available at

http://theses.gla.ac.uk/7793

Randel, W.J., Wu, F., 2015. Variability of zonal mean

tropical temperatures derived from a decade of GPS

radio occultation data. Journal of the Atmospheric

Sciences, vol. 72, pp. 1261–1275.

Saha, S., Moorthi, S., Wu, X., Wang, J., Nadiga, S., Tripp,

P., Behringer, D., Hou, Y.-T., Chuang, H., Iredell, M.,

Ek, M., Meng, J., Yang, R., Mendez, M.P., van den

Dool, H., Zhang, Q., Wang, W., Chen, M., Becker, E.,

2014. The NCEP climate forecast system version 2.

Journal of Climate, vol. 27, pp. 2185–2208.

Sánchez-Benítez, A., García-Herrera, R., Barriopedro, D.,

Sousa, P.M., Trigo, R.M., 2018. June 2017: The

earliest European summer mega-heatwave of

reanalysis period. Geophysical Research Letters, vol.

45, pp. 1955–1962.

Seidel, D.J., Li, J., Mears, C., Moradi, I., Nash, J., Randel,

W.J., Saunders, R., Thompson, D.W.J., Zou, C.-Z.,

2016. Stratospheric temperature changes during the

satellite era. Journal of Geophysical Research:

Atmospheres, vol. 121, pp. 664–681.

Modeling Method for Temperature Anomaly Analysis

279

Seim, A., 2016. Spatiotemporal climate and atmospheric

circulation variability in Asia inferred from tree rings,

PhD Thesis, University of Gothenburg, Sweden.

Shi, F., Ge, Q., Yang, B., Li, J., Yang, F., Ljungqvist,

F.C., Solomina, O., Nakatsuka, T., Wang, N., Zhao,

S., Xu. C., Fang, K., Sano, M., Chu, G., Fan, Z., Gaire,

N.P., Zafar, M.U., 2015. A multi-proxy reconstruction

variations in Asian summer temperatures over the last

millennium. Climate Change, vol. 131, no. 4, pp. 663–

676.

Simmons, A.J., Berrisford, P., Dee, D.P., Hersbach, H.,

Hirahara, S., Thépaut, J.-N., 2017. A reassessment of

temperature variations and trends from global

reanalyses and monthly surface climatological

datasets. Quarterly Journal of the Royal

Meteorological Society, vol. 143, pp. 101–119.

Song, C., Ke, L., Richards, K.S., Cui, Y., 2016.

Homogenization of surface temperature data in High

Mountain Asia through comparison of reanalysis data

and station observations. International Journal of

Climatology, vol. 36, pp. 1088–1101.

Turner, J., Lu, H., White, I., King, J.C., Phillips, T.,

Hosking, J.S., Bracegirdle, T.J., Marshall, G.J.,

Mulvaney, R., Deb, P., 2016. Absence of 21st century

warming on Antarctic Peninsula consistent with

natural variability. Nature, vol. 535, pp. 411–415.

Way, R.G., Bonnaventure, P.P., 2015. Testing a

reanalysis-based infilling method for areas with sparse

discontinuous air temperature data in northeastern

Canada. Atmospheric Science Letters, vol. 16, pp.

398–407.

Wing, A.A., Emanuel, K., Solomon, S., 2015. On the

factors affecting trends and variability in tropical

cyclone potential intensity. Geophysical Research

Letters, vol. 42, pp. 8669–8677.

Xu, W., Li, Q., Jones, P., Wang, X.L., Trewin, B., Yang,

S., Zhu, C., Zhai, P., Wang, J., Vincent, L., Dai, A.,

Gao, Y., Ding, Y., 2018. A new integrated and

homogenized global monthly land surface air

temperature dataset for the period since 1900. Climate

Dynamics, vol. 50, pp. 2513–2536.

IJCCI 2018 - 10th International Joint Conference on Computational Intelligence

280