Estimating suspended sediment concentrations in surface water of Tri An lake (Vietnam) using landsat multispectral images

Vietnam is located downstream of some major rivers, so Vietnam has a vast river network. Although the quality of upstream river water is generally good, the downstream sections of major rivers reveal poor water quality, and most of the lakes and canals in urban areas become sewage sinks. The weakness in industrial wastewater management is the main cause of water pollution in Vietnam. Many industrial enterprises use freshwater to carry waste from their factories into canals, rivers, and lakes. Most of the enterprises do not have any wastewater treatment system and many industrial zones do not have a wastewater treatment centre for their factories. Industrial wastewater is directly discharged into canals, lakes, ponds, and rivers, causing serious pollution of surface water [10].

Concentration of suspended sediment is one of the major parameters for monitoring water turbidity. Furthermore, suspended sediment has a substantial influence on temperature, stratification, biology and carries chemical substances and pathogenic agents. The traditional methods for measuring water quality variables are time-consuming and do not give a synoptic view of a water body or, more significantly, a synoptic view of different water bodies across the landscape. It requires excessive traveling, sampling, and expensive laboratory analysis, especially for a large area; thus it is very difficult to report and predict the water pollution situation in time. Remote sensing technology with advantages such as wide area coverage and short revisit interval have been effectively used for environmental pollution applications, such as for monitoring water quality [7].

MATERIAL AND METHODS

LANDSAT imagery. To estimate SSC we used LANDSAT ETM+ multispectral image on 12 February 2010 in Tri An lake (Vietnam) (figure 1).

Figure 1 - The LANDSAT multispectral image (Tri An lake, 12-02-2010) with color composite RGB=432

The Enhanced Thematic Mapper ( ETM+) on board LANDSAT-7 is a multi-spectral radiometric sensor that records eight bands of data with varying spectral and spatial resolutions (30m spatial resolution for red, green, blue, near infrared, and two bands of medium infrared; 60m for thermal infrared; and a 15m panchromatic band). With average spatial resolution, multispectral image LANDSAT ETM+ performance applications in the region study (table 1).

Table 1 - Characteristic of LANDSAT ETM+ multispectral image

No.	Bands	Wavelength (µm)	Spatial resolution (m)
1	Blue	0.459 – 0.515	30
2	Green	0.525 – 0.605	30
3	Red	0.630 – 0.690	30
4	Near infrared (NIR)	0.775 – 0.900	30
5	Mid-infrared (MIR)	1.550 – 1.750	30
6	Thermal infrared (TIR)	10.40 – 12.50	60
7	Mid-infrared (MIR)	2.090 – 2.350	30
8	Panchromatic	0.520 – 0.900	15

In situ measurements. All in situ data were collected during research cruises in February 2010 in a set of stations distributed along the Tri An lake [8]. Table 2 shows the coordinates of all water sampling stations and SSC (mg/l).

Table 2 - In situ measurements of SSC (mg/l) in February 2010

No.	Water sampling stations	Coordinates		TSS (mg/l)
		X	Y
1	Hieu Liem – Vinh Cuu	0416720	1228426	2
2	Vinh An	0419014	1230046	2
3	Chien khu D bridge	0422343	1229191	3
4	Ma Da – Vinh Cuu	0432420	1232295	2
5	Phu Cuong – Dinh Quan	0447830	1239000	3
6	The Lake of Phu Cuong – Dinh Quan	0447491	1232700	2
7	La Nga – Dinh Quan	0444048	1239451	6
8	Thanh Son – Dinh Quan	0427157	1230796	13
9	Village 3, La Nga – Dinh Quan	0436964	1240483	13
10	La Nga bridge, 1km upstream	0445939	1235825	23

Methodology. For atmospheric correction, we use the dark object (DOS). This model works as follows step:

In the first step, the digital number is converted to spectral radiance using the sensor calibration parameters by the following equation [9]:

L max - L min

L =                 ( DN - DN min) + L min (1)^,

λ DN max- DN min where Lλ – spectral radiance at the sensor’s aperture in watts/m2.ster.µm; DNmax and DNmin – the maximum and minimum quantized calibrated pixel value in digital number (DNmax = 255, DNmin = 1); DN = the quantized calibrated pixel value in digital number; The Lmax and Lmin are the spectral radiances for each band of LANDSAT image at digital numbers 1 and 255 respectively in watts/m2.ster.µm (table 3).

Table 3 - The values of Lmax and Lmin of LANDSAT multispectral image

Band	Name	Wavelength	Lmax	Lmin
1	Blue	0.45 – 0.515 µm	191.600	-6.200
2	Green	0.525 – 0.605 µm	196.500	-6.400
3	Red	0.63 – 0.69 µm	152.900	-5.000
4	NIR	0.75 – 0.90 µm	241.100	-5.100
5	MIR	1.55 – 1.75 µm	31.060	-1.000
7	MIR	2.09 – 2.35 µm	10.800	-0.350

In the second step, the theoretical radiance of a “dark object” is computed with the theoretical reflectance of 1% as proposed by Chavez (1996) and Moran et al. (1992) [1, 4]:

_L      0.01. cos( θ _s ). d ²

(2),

^{λ 1% =}    π . ESUN .

λ where: ESUN – the mean exoatmospheric solar spectral irradiance (watt/ m2.ster.µm) and show in table 4; θs – the solar zenith angle in degrees; d – Earth – Sun distance in astronomical units, which can be computed using the following formula: d = (1.0 -0.013729.cos(D-4)), where D – the day number of the year.

Table 4 - LANDSAT ETM+ solar spectral irradiances (ESUN)

Band	Name	Wavelength	ESUN (watts/m2.ster. µm)
1	Blue	0.45 – 0.515 µm	1997
2	Green	0.525 – 0.605 µm	1812
3	Red	0.63 – 0.69 µm	1533
4	NIR	0.75 – 0.90 µm	1039
5	MIR	1.55 – 1.75 µm	230.8
7	MIR	2.09 – 2.35 µm	84.90
8	PAN	0.52 – 0.90 µm	1362

Step 3 computes a value for the haze that may be present in the image:

^L Xha ze = ^L X    ^L X 1% (3).

reflectance values. Surface

The last step converts spectral radiance to surface reflectance can be calculated using following equation:

= П . d ².( L _x - L _X haze )

Figure 2 - The algorithm of extraction of suspended sediment from LANDSAT image

In surface water, calibration relationship improving can be established between reflectance ratios and in situ measurements data [6]. In turbid sediment-dominated water, the reflectance ratios between two near infrared bands or between a near infrared and a visible band are highly correlated to in situ measurements data [1-4, 6]. In this study, we use the reflectance ratios of near infrared band (band4) and green band (band2) LANDSAT TM, ETM+ multispectral image, and in situ data to build the regression model. In this study, for extracting water – land coastline, we use automated techniques, which given by Alesheikh A. (1996) [5]. The algorithm for estimating of SSC form LANDSAT image showed in figure 2.

RESULTS AND DISCUSSIONS

Study area. The Tri An lake is a hydroelectric dam on the Dong Nai river in Vinh Cuu district, Dong Nai province, Vietnam. The dam was built from 1984 to 1987 with assistance of the Soviet Union and knows as Tri An lake. This is one of the largest lakes in Vietnam with a full capacity of 2765 km³ and surface area of 323 km² [8].

According to the results of monitoring surface water quality, the Tri An upstream location, the location of Vinh An, Phu Cuong, La Nga towns and Dong Nai river mouth, surface water is seriously polluted. In this area, concentration of suspended sediment (SS), chemical oxygen demand (COD), biochemical oxygen demand (BOD 5), nitrate, iron and microbiology organisms in water many times in excess of standards [8].

Figure 3 - The ratio image of surface reflectance band4 and band2

Water – land coastline extraction. The reflectance ratios of near infrared and green bands LANDSAT multispectral image in study area showed in figure 3. The results of extraction water – land coastline using Alesheikh’s automated technique showed in figure 4. In this figure, white color represents water and black color represents land.

Development of SSC – water reflectance relations. The relationship between reflectance ratios band4 and band2 LANDSAT TM, ETM+ multispectral image and in situ measurements data was determined with linear, exponential and power regression. Good relationships for suspended sediment and reflectance ratios were found in power regression with 2-second and 3-rd order (the square correlation coefficient R² = 0.9826) (figure 5).

In order to find the best model, model for estimating SSCs have been selected with R2 = 0.9826 following equation:

55 ( mg- ) = 131.55.( ^P band ⁴) 2 - 94.335( ^P band ⁴ ) + 18.838

(5),

¹                P hand 2                P hand 2

where p _{hand 4} and p _hand 2 are surface reflectance of near infrared and green bands LANDSAT multispectral image.

Figure 4 - Coastline extraction from LANDSAT multispectral image

SS(mg/l)

a)

SS(mg/l)

c)

SS(mg/l)

d)

Figure 5 - The relationship between SSC (mg/l) and ratio of surface reflectance band 4 and band 2 LANDSAT image (a – linear, b – exponential, c – polynomial with order 2, d – polynomial with order 3 regression)

Retrieval of SSC spatial distribution. To convert the digital numbers into SSC we use the calculating module of ERDAS Imagine 9.3. To show how the empirical relation (equation 5) behaves when applied to the surface water, we extract the SSC distribution map (fig. 6) for regions in southwest Dong Nai province using LANDSAT ETM+ image. Figure 6 shows SSC variations along the Tri An lake on 12 February 2010. In this example, the estimation of SSC ranges from 1.5 to 88.0 (mg/l). The dark blue area has the highest SSC value, which indicates the highest turbidity. In the contrary, the light blue area corresponds low turbidity on the water surface. The highest SSC is located in coastal region and the lowest SSC is located in centre of Tri An lake.

1.5(mg/l)                                                         88.0(mg/l)

Figure 6 - SSC generated using developed algorithm

CONCLUSIONS

Nowadays, water pollution has become a global issue, directly caused by human populations grow, industrial and agricultural activities and climate change. A simple and operational method is presented to estimate the suspended sediment concentration in turbid estuarine water using satellite images.

The model to estimate suspended sediment concentration from reflectance ratios band4/band2 of LANDSAT multispectral image is used in the study and proved useful for the surface water in Tri An lake. Water quality mapping using remote sensing images shows a general pattern of decreasing SSC outward from the coastline. This can be a powerful information for water management authorities for water quality assessment, and the design of monitoring program to target the most dynamic and problematic area.