Optical remote sensing of the Gulf of Gabès – relation between turbidity , Secchi depth and total suspended matter

Abstract. Optical remote sensing is used here in the Gulf of Gabes to provide scientific information to support environmental management. The Gulf of Gabes is located in the southern east coast of Tunisia. It is a shallow continental shelf with semi-diurnal tides with average amplitude of 2 m. Industrial activities in this area since the early 1970s may have contributed to the degradation of the biodiversity of the ecosystem with eutrophication problems, and disappearance of benthic and planktonic species. To assess the long-term effect of anthropogenic and natural discharges on the Gulf of Gabes, turbidity and light penetration in the coastal waters is assessed from in situ measurements of total suspended matter concentration (TSM), Secchi depth and turbidity (TU) and from remote sensing data using a semi empirical algorithm to retrieve turbidity and total suspended matter products from moderate resolution imaging spectrometer MODIS Level 2 (L2) data.


Introduction
The objective of this study is to use the optical remote sensing technique and in situ measurement to estimate water quality in the Gulf of Gab ès and impact of industrial discharge.The industrial zone is located in the central Gulf of Gab ès in the port of Gannouche (Fig. 1).Industrial activities began in the seventies involving a very notable environmental deterioration; this has been facilitated by the monotone morphology of the bottom and the effect of the tide.
The ocean colors data becomes the most useful utility for the detection of optical properties and the components of water.
In fact, the mapping of the suspended matter in coastal waters which is resulting from satellite and in situ data, gives information about the quality of water, the assessment and monitoring of suspended sediment distribution.Introduction

Conclusions References
Tables Figures

Back Close
Full In order to monitor water quality, several studies have been previously published on how to extract TSM from ocean color and mapping data.Nechad et al. (2010) have used and calibrated MODIS, Sea-viewing Wide Fieldof-view Sensor (SeaWiFS) and Medium-Spectral Resolution, Imaging Spectrometer (MERIS) red bands (600-700 µm) to retrieve TSM map of the North Sea.Also Miller et al. (2004) has used MODIS terra 250 m band 1 (620-670 µm) to map TSM in the coastal water of the North Gulf of Mexico.Ouillon et al. (2008) propose a global algorithm for tropical coastal waters based on one or three bands: turbidity is first calculated from remote sensing reflectance Rrs681 and then if turbidity <1 NTU it is recalculated using Rrs620 × Rrs681/Rrs412.These above studies have shown satisfactory results between in situ measurements and marine reflectance extracted from satellite ocean colors.
In this study, in situ measurements of turbidity (TU) and MODIS-derived TSM and TU are compared and correlated, to adapt a semi-empirical model established for the turbid waters, to the Gulf of Gab ès.The performance of these models is assessed and satellite maps of TSM in this region are presented.

Regional situations
The Gulf of Gab ès, part of the Pelagian Sea as described in Burollet et al. (1979), is located in the southern part of Eastern Tunisia.
It is a vast trough extending from Djeffara Island to the south and the Gulf of Hammamet to the north.The Gulf is characterized by two basins separated by a bathymetric level joining the Kerkennah and Jerba Islands (Fig. 1).Generally the climate of the region of Gab ès is influenced by the temperate, humid and hot Mediterranean air coming from the east, but also by the subtropical, dry, hot and sandy Saharan air coming from the south-west.The Level 2 (L2) satellite data products contain the geophysical value for each pixel, derived from the Level-1B (L1B) radiance after radiometric calibration, geometric correction, atmospheric correction and bio-optical algorithms.
The L1B are used to understand some features observed in L2, to check the quality of the atmospheric correction established by the NASA, through the information recorded at the top of the atmosphere.
TSM concentration and TU maps were obtained using respectively the algorithms (Nechad et al., 2009(Nechad et al., , 2010)), applied to MODIS remote sensing reflectance at band 667 nm (Rrs667).These bio-optical algorithms (Eqs. 1 and 2) perform well in the Belgian coastal waters.Here, they are tested for the Gulf of Gab ès.The algorithms for TSM and TU are, respectively: Aerosols Optical Thickness at 869 nm (AOT 869), Epsilon of aerosols correction at 748 nm and 896 nm (EPS78).Third, application of algorithm (Eqs. 1 and 2) to map TSM and TU.Then, application of the L2 flags, especially cloud/ice, atmospheric correction warning and atmospheric correction failure.Finally, mapping of TSM and TU for interpretation.
TU and TSM are mapped for all daily images of year 2009 and their seasonal and annual means are calculated and mapped.

In situ measurements
In coastal waters, Secchi depths and TU are important optical measurements of water quality.In the coastal waters of the Gulf of Gab ès light is considerably attenuated because of the presence of suspended matter, phytoplankton, yellow substance and phosphogypse pollution.
Turbidity is measured using the portable Hach, 2100P ISO turbidimeter, water transparency is measured via the Secchi depth and TSM was measured gravimetrically: water was sampled near the surface and filtered on-board over pre-weighed pre-ashed GF/F filters and the filters were rinsed with milli-Q water.The filters were then dried and weighed in the laboratory of GREEN LAB (Tunisia).
At each station measurements were made of TU, TSM, chlorophyll-a concentration (chl-a), sea surface temperature (SST) and transparency (Fig. 1).A summary of data is given in Table 1.

In situ measurements
Results for in situ TSM and TU show a strong correlation with a linear function expressed by: TSM (mg/l) = 1000 TU(NTU) + 0.807 and a regression of R 2 = 0.952 (Fig. 2).Introduction

Conclusions References
Tables Figures

Back Close
Full On the region of Gannouche at the release points, the turbidity reaches high values, as impossible to measure.The origin of such high turbidity comes from the recent phosphogypsum wastes on that region, forming an opaque film on the water surface causing such a strong TU and TSM needing, amount of time, to undergo either decantation or dissolution (Bjaoui, 2004).
The exten of phosphogypsum depends on current strength and bathymetry.The pollution spreads further and progressively if the water is deep and the currents in the receiving area are weak.On the other hand, if the currents are strong and the receiving area shallow, the phosphogypsum falls quickly to the bottom and is trapped by cohesive binding.
The inverse relationship between the transparency and turbidity are shown around Kneiss islands, and in the coast of Skhira and Sfax due to the large tidal range and the distinctive features of the sea bottom where there are wadis and channels under the sea with shallows which causes an abrupt variation in bathymetry from 7 m to 1.5 m.

Satellite data
TU and TSM products were obtained using the algorithm for case II waters that was applied to Belgian waters in the North Sea (Nechad et al., 2009(Nechad et al., , 2010)).The comparison of these MODIS satellite data with the measurements taken in the Gulf of Gab ès indicates a reasonable correlation, but suggests that a recalibration of the algorithm may be necessary for this region.neural network (NN), which was tried with a large set of simulated TOA reflectances for 12 of 15 MERIS bands.If this algorithm becomes operational for MODIS or SeaWIFS, we will have more satellite data for the study area.

Images of 5 July 2009
The image of 5 July 2009 was acquired during excellent conditions, clear sky and low wind.
Figure 3 shows maps of the L1B radiances taken at this date: -At 841 nm (Fig. 3a): grey area indicates the dispersion of the aerosols from land to sea following direction of the wind.The black area indicates the surface of the water seen by satellite and having low concentration of aerosols.
-The 915 nm-centred radiance band (Fig. 3b) which contains information about atmospheric water vapor.
-L1B RGB image (Fig. 3d) shows 2 types of waters: clear waters in blue and green waters indicate turbid area near the coast mainly around Kerkennah island, Skhira town and port of Gannouch along the coast.
The aerosols optical thickness at 869 nm (Fig. 3e), AOT, shows a spatial distribution which does not exceed 0.142 values.The maximum AOT is reached near the port of Gannouch.Moreover, the direction of the aerosols agrees with the direction of the wind recorded during the satellite overpass.
The remote sensing reflectance at 667 nm (Rrs667) map (Fig. 3f) shows patterns that are uncoupled from the atmospheric patterns listed above.These are related to the signal coming from the sub-surface water layer.The MODIS level2 flags for cloud/ice (cldice) and land pixels are used to set up the cloud and land masks, respectively.

Conclusions References
Tables Figures

Back Close
Full There were no pixels with bad atmospheric correction.Those affected by a warning flag (atmwarn) cover only sand banks over land.
The derived TU distribution map on 5 July 2009 (Fig. 3g) shows some areas with higher turbidity near the industrial discharge (port of Gannouch) and around the Kneiss and Kerkennah islands.The centre of the Gulf shows higher turbidity.This has been shown in most of the images processed in 2009.
Figure 4 shows a scatter plot of MODIS TU product using the algorithm of (Nechad et al., 2009) versus in situ TU at 12 matchup locations in the Gulf of Gab ès.Only the pixels where in situ measurements were taken at the time of satellite overpass (±30 min) are considered, to avoid uncertainty from the tidal effects (bottom sediment resuspension).
The regression illustrates the positive relationship found between in situ and satellite measurement with a fairly good correlation (68.9%) between MODIS and in situ TU covering the range [0.5-4 NTU], (Fig. 4) despite the fact that the algorithm is tailored for more turbid waters [0.5-85 NTU].

2009 turbidity map
The 2009 average turbidity map (Fig. 5) was computed from 58 images, with 20 in autumn (September, October and November), 10 in winter (December, January and February), 14 in spring (March, April and May) and 14 in summer (June, July and August).It shows 4 areas of higher turbidity concentration in the study area: around Kerkennah, Kneiss and Jerba Island and especially in the center of the Gulf.This distribution of maxima was observed for all TU maps along the year 2009 (Fig. 6a-f).
The standard deviation turbidity TU map showed the same average turbidity distribu-

Summary and recommendation
In order to map TU distribution in the Gulf of Gab ès, a TU algorithm (Eq.2) was adapted for this region.MODIS derived TU and in situ matchups taken in July and October 2009 were collected and compared.Analysis of all images and the mean of turbidity in 2009 show the highest TU located around the islands (Kerkenah, Kneiss and Jerba) and at the industrial port of Gannouche, extending towards the center of the Gulf out to a distance of 70 km.
In situ data and the relation established between the measurements parameters were useful to interpretations of satellite data.
In fact, the surestimation of TU and TSM values observed around the islands of the Gulf are mainly due to shallow waters and bottom reflection.However, the values of TSM and TU measured in the port of Gannouche and in the center of the Gulf were validated by in situ measurement.
This study demonstrated that the characteristics of the elaborated and calibrated algorithm for the North Sea, which was applied for the Gulf of Gab ès, showed satisfying results in the distribution of TU and TSM for this area.
In order to estimate these concentrations, a large amount of in situ data, especially turbidity and chlorophyll, from different dates would be needed in order to further validate and possibly recalibrate the local algorithms.Introduction

Conclusions References
Tables Figures

Back Close
Full  Full Discussion Paper | Discussion Paper | Discussion Paper |

TSM
= 62.86 × r/(0.1736− r) (1) TU = 50.46× r/(0.1736− r) (2) where r = π × Rrs667.The processing of MODIS images L1B and L2 was established using the ENVI (IDL) software.The elaborated program is composed of successive procedures.First, the georeferencing of all bands.Second, the extraction of standard products especially the Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | The daily images covering the area of interest are frequently affected by sun glint.Also, because of the sun glint more frequent during the summer period, the images taken on 6 July 2009 at 12:45 UTC and on 7 July 2009 at 12:15 UTC do not contain any data.The image of 19 October 2009 at 11:55 UTC was affected by cloud.A majority of pixels are flaged.Based on the work of Doerffer (2010) it may become possible to make atmospheric correction and use images affected by moderate sun glint.The procedure is an artificial Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | tion as in 2009.The value of standard deviation indicated a small variability of turbidity for 2009Discussion Paper | Discussion Paper | Discussion Paper | According to the MODIS satellite data, the Sfax region and the area surrounding the Kerkennah, Jerba and Kneiss islands show very high TU (10 NTU), values compared to neighboring regions especially the center of the Gulf of Gab ès and the Port of Gannouhe [5-7 NTU].The more or less high TU values recorded in the industrial waste which indicate that values measurement around islands are surestimated due to shallow waters and bottom reflection.
Discussion Paper | Discussion Paper | Discussion Paper |

Fig. 1 .Fig. 2 .Fig. 3 .Fig. 4 .
Fig. 1.Seaborne cruises in July 2009 in the left.Seaborne cruises in October 2009 in the right.Green circle indicate the position of in situ measurement and purple triangle indicate the match-up stations.

Table 1 .
Summary of measurement ranges (see locations in Fig.1).