Rainfall Difference between Urban and Non-urban Areas in Bangladesh

by M.A. Hoque


Introduction

Bangladesh lies in the tropical region, in the southern part of Asia. It has an area of 144,000 sq. km extending between latitudes 20° and 27° North and longitudes 88° and 93° East. 23˝° latitude and 90° longitude pass through the middle of the country. It is surrounded by parts of the Himalayas in the North and the East, and the Bay of Bengal in the South. The presence of the bay and the hilly regions causes heavy rainfall in the northeastern and eastern hilly region. The country experiences most of its rainfall during the rainy season consists of May, June and July months. Other than during the rainy season, the country is struck by cyclones in March-April and October-November often. The rainfall patterns tells that the country is having plenty of rains all over its territory round the year except during dry season from December to February.


Problem Statement

The three main rivers of the country, the Padma, the Meghna and the Jamuna, deposit a great deal of alluvial soil each year making it a mainly agricultural country. Bangladesh, a rural country, started experiencing urbanization not before 100 years. The proportion of urban population was remained only 2.43% during the period 1901-1921. After that it increased and reached 8.8% in 1974, 15.2% in 1981 and 20.2% in 1991.

The isolines in Figure 2 shows that there is a concentric pattern of rainfall over the study area having constant increasing from north-east to south-west in 1990. On the other hand, in 1970 shows a different pattern, not concentric, increasing reversibly from south-west to north-east with a rapid rainfall change at the north-east corner of the study area. From the maps, it can also be assumed that the rainfall changes in July'90 was not influenced by the urban areas, while in July'70 it seems that the urban area could have some kind of influence on the rainfall pattern.

Therefore, the purpose of this paper is to examine whether there is any statistically significant difference between rainfall in urban and non-urban areas. Two-sample difference of means Z test and Wilcoxon Rank Sum test were used to examine the phenomenon.


Study Area


Bangladesh is mainly a rural country. Though a huge number of population live in urban areas, only a few areas are urbanized. The Capital city, Dhaka and the port city, Chittagong have been experiencing the major urbanization for last decades. For this study, a rectangular shaped study area has been chosen around Dhaka which experiencing rapid urbanization over the decades. The area is delineated from satellite image that was prepared by EGIS, Dhaka, Bangladesh. The study are covers both urban and non-urban areas (Figure 1).



Methodology

To identify the difference of the rainfall in urban and non-urban areas, both parametric and non-parametric tests were used. To check the differences over the period, the tests were conducted for 1970 and 1990. The month July was chosen, as it is the mid point of the rainy season. ArcView GIS and SPSS software packages were used for spatial and statistical calculation. The steps are as follows:

Step 01: Generating rainfall surface for the entire country

There are only 3 rainfall stations in the study area (Figure 2), and a few around the area. Since it is hard to determine which stations have the influence on that area, rainfall surfaces of 900m cell size were generated using ArcView for both July’70 and July’90. Then the surfaces were clipped for the study area. Table 1 shows a sample of the data that used to generate the rainfall surface. 

Table 1: Monthly total rainfall (in cm)

Station ID	Station Name	July'90		July'70

R209		Ruhea		 558.90		 749.70

R221		Thakurgaon	 584.70		 440.80

R206		Rangpur		 307.00		 476.20

R006		Bogra		 549.10		 532.80

R025		Pabna		 367.80		 387.90

R369		Noakhali	 842.30		 747.60

R311		Fatikchari	1057.10		1202.00

R328		Rangamati	 865.40		1491.90

R260		Bhola		 451.20		 630.00

Source: Bangladesh Flood Forecasting and Warning Center and SWMC
Step 02: Generating sample points

200 sample points were generated randomly within the study area using ArcView. Among them 162 are fallen in non-urban area and 38 in urban areas (Figure 2). Since in the both areas, the number of sample points are more than 30, t test was used instead of Z test.

Step 03: Calculating rainfall for sample points

ArcView’s spatial analysis module was used to compute the rainfall for each point. Data for both July’70 and July’90 were estimated for each sample points from the surfaces. Table 2 shows a sample of the data for each sample points. 

Table 2: Rainfall collected for sample points in cm

ID	URBAN Area	RAIN July'90	RAIN July'70

1	No	 	400.42		 484.93

2	Yes	 	413.87		 663.53

3	No	 	405.02		 449.36

4	No	 	506.36		 437.26

5	No	 	375.74		 508.37

6	No	 	390.71		 728.88

7	No	 	427.67		 507.06

8	Yes	 	420.26		 637.46
Step 04: Computing the p values

SPSS software was used to run the t test and Wilcoxon Rank Sum test. The tests were conducted for both of the years. 1 sample ANOVA was conducted to test whether the variances between the samples are equal. This test was used to choose pooled variance estimate (PVE) over separate variance estimage (SVE).

Two rainfall isoline maps (Figure 2) were prepared to visualize the rainfall pattern over the study area during the study periods. It also helps to visually detect the rainfall pattern changes over the urban and non-urban areas in July'70 and July'90.


Results

Descriptive statistics of the samples say that the sample in urban areas have a mean of 411.67 with standard deviation of 10.46 for July'70 and have 618.18 mean with 76.9 standard deviation for July'90. Samples in non-urban areas have 404.35 mean value with standard deviation 35.65 for July'70 and have 626.43 mean value with standard value of 132.32 for July'90. 

Table 3a: Independent 2 Samples Test Results		Table 3b: Wilcoxon Rank Sum Test Result

		  t	p value						  Z	p value

July'70		-2.236	 0.027				July'70		-3.542	0.001

Jult'90		 0.496	 0.621				July'90		-0.084	0.933
In t test "Equal variance assumed" option was chosen as, in ANOVA (Analysis of Variance) test, F values are around 1 and p values are high (for July'90: F = .13 and p = .719, for July'70 F = 1.56 and p = .212) which strongly infer not to reject null hypothesis, that is the variance are equal.

From table 3a and 3b, in both parametric and non-parametric tests small p values (0.027 and 0.001) suggest that in 1970, there is less chance of making Type I error by rejecting the null hypothesis that there is no significant difference between rainfall in urban and non-urban areas. So, the null hypothesis is rejected for 1970 with above 97% confidence. So, It can be inferred that in 1970 the urban and non-urban rainfall varied significantly.

On the other hand, in 1990 p values in both tests are higher suggesting a higher chance (62% and 93%) of making Type I error by rejecting the null hypothesis. So, the null hypothesis is not rejected here and accepting that there is no significant difference in rainfall between urban areas and non-urban areas in July'90.


Conclusion

Although there is a hypothesis that the rainfall varies between urban and non-urban areas, in this study shows two different scenarios – one matches with the hypothesis, the other does not. Rainfall in both years not only shows different statistical results, it also shows different, almost the reverse rainfall pattern when the isolines were drawn. In 1990, Bangladesh had a severe flood, 1 in 100 year flood. Because of this or other climatic factors (like late or early monsoon or extreme rainfall) may influence to come up with this result. This study should have looked at into further details to examine the scenarios. Finally, it is to be mentioned that the urban extent was assumed to be same from 1970-90, though the urban area has expanded over the period. Again, the results could be effected by the fact that rainfall stations' data in urban areas were not available for this study.


Reference

Bangladesh Bureau of Statistics, 1991 http://home.bangla.net/ndb/ana_vol1/density.htm
http://www.emulateme.com/content/bangladesh.htm
McGrew, J and Monroe, C. B. 2000 An Introduction to Statistical Problem Solving in Geography Dubuque, Wm. C. Brown Publishers