Mapping and Monitoring the Sea Surface Temperature in Weda Bay Using Terra and Aqua- Modis Satellites

Temperature is one of the principal controls on all physical, chemical and biological processes in the environment. Therefore, temperature data play an important role in earth resources management activities, including managing the effects of climate change. In this study, the sea surface temperatures (SSTs) of Weda Bay, Halmahera Island, Indonesia were mapped and monitored, from January to November 2007, using thermal infrared (TIR) band 30 and 31 of Terraand Aqua MODIS satellites. The empirical prediction SST model developed, using TIR band and in-situ measurement SST, showed that the model was sufficient to predict and to map the SST within the bias ranges of ± 0.5°C. Daily SST, averaged 10 days SST, and monthly SST maps were made using 109 available Terraand AquaMODIS images. The ranges of daily and 10-day average SSTs in Weda Bay were narrow, about 2°C (28-30°C) throughout the year, while the range for monthly SSTs was only 1°C (28.75-29.75°C). Accordingly, no indication of upwelling phenomena occurred in this bay during the observation (2007), but it is possible that upwelling could have happened in the past or may in the future. Long-term monitoring from space should continue in order to get a clearer understanding of the water characteristics in Weda Bay, not only using TIR, but also using ocean color bands of MODIS.


Introduction
Temperature is one of the principal controls on all physical, chemical and biological processes in the environment. Consequently, temperature data play an important role in a wide range of earth resources management activities [1]. For over 30 years far-infrared, or Thermal Infra-Red (TIR), images of the earth's surface have been gathered by optical-mechanical scanner systems. First these were from airborne, and later from earth-orbiting, satellite platforms [2]. These images give us an opportunity to measure the radiant temperature of the earth's surface, including the sea, from a distance, by sensing in the wave-length range of 3-14 µm [1]. High-quality sea surface temperature (SST) data have been archived since the late 19th century and this makes SST one of the more robust indicators for understanding the earth's climate or for studying climate change.
Numerous investigations of the sea surface temperature have been carried out, firstly using TIR data from the "Advance Very High-Resolution Radiometer (AVHRR) sensor on the board of Tiros-n/ NOAA series satellites, and following that, by sensors on board various other satellites. Most the SST estimation using those TIR sensors have been done by employing a well-established model called the "Split-Window Model", except for those satellites that have only one band of TIR (Nimbus-7, Landsat-4, -5 and-7). This model has been used world-wide and published in many journal related to the field of remote sensing.
The earth surface energy sensed or measured in TIR region is the emitted energy of a target, not reflected energy such as invisible (0.4-0.7 µm) nor reflective infrared (near and mid infrared; 0.7-2.4 µm) regions. Thus, because it does not depend on reflected sunlight, thermal data (SST) can even be collected at night [3]. When the infrared radiation emitted from the ocean is discussed, wavelengths in the 3.5-4.0 µm and 10-13 µm range are customarily considered. The reasons are: i). Radiant power peaks in the range of SST from 9.3-10.7 µm, while ii). The atmospheric absorption minima are in the 3.5, 9.0 and 11.0 µm range. Thus, as a practical matter, 3-15 µm represents a wavelength region where sufficient radiant power is available for suitable detector response, and the optical properties of the atmosphere permit remote sensing of the surface temperature [4,5]. Furthermore, since the water is the only object whose emissivity is constant over a large area and is around one (emissivity of the water ≈ 1) in the spectral channel of interest, the SST can be reliably estimated from its radiance [3,6,7].

Study sites and the collection of sea truth data
Weda Bay is one of the three main bays on Halmahera Island, North Maluku Province (Figure 1). In-situ SST data were collected from 2 series of permanent stations, namely OP series stations (5 stations, station OP1 to OP5) and UP series station (13 stations, station UP 1 to UP 13). SST were measured using a CTD at those stations ( Figure 1) from February to August 2007, except in July when no measurements were taken due to the rough seas. SST measurements were conducted at morning, noon, afternoon between AM to AM.

Acquisition and processing of Terra and Aqua-MODIS satellite data
The Terra and Aqua-MODIS Level-1 B (calibrated radiances) data were acquired from NASA Goddard Earth Sciences Distributed Active Archive Center (GES DAAC) http://ladsweb.nascom.nasa.gov/data/ search.html. These data contain calibrated earth view observations from MODIS bands 1-7 (for land application) that are aggregated to appear at 1 Km from the original resolution of 250 m (band 1-2) and 500 m (band 3-7); band 8-16 (for ocean color study) and band 17-36 (for atmospheric study) at 1 Km resolution [9]. Thus, for this study, the MODIS data at 1 Km resolution in the format MOD021KM (for Terra satellite images) and MYOD1KM (for Aqua satellite images) with a size of approximately 160 mega bytes (Mb) were downloaded. The Terra satellite passes over Weda Bay at ± 10:30 AM local time (1:30 GMT time) in descending node (from north to south), while Aqua at ± 1:30 PM (4:30 GMT) is in ascending node (from south to north).
From 36 available bands of MODIS, there are 5 bands of TIR, bands 20 (3.7 µm), 21 (3.9 µm), 23 (4.0 µm), 30 (11 µm) and 31 (12 µm). The first three bands were used to estimate the night-time SST, while the other two bands were used for daytime SST estimation. Therefore, in this study, we used only bands 30 and 31, because the available in-situ data were collected during daytime. MODIS data are in HDFEOS format, an evolving format developed by NASA for recording all information (data and meta-data) of a certain satellite data (Terra-, Aqua-MODIS, Aster, Landsat-7). Those data in HDF were converted to multi-band Geotiff format using a software program called HEG (HDFEOS to Geotiff). Converted data were ready to analyze using various images processing package software. In this study we used IDRISI ANDES (ver. 15) software for processing the Terra-and Aqua-MODIS images. Since this data covered an area of 2330 Km viewing swath width (± 2330 × 2330 pixels), then a window area covering all of Weda Bay and small parts of the surrounding areas (± 165 × 210 pixels) were cropped from the entire images.

SST empirical prediction model development and mapping
There are two approaches to convert the relative spectral response  [18][19][20]. In this study we used the second approach.
For developing SST prediction model, we used only the in-situ SST measurements made on a clear (relatively free cloud) day by Terra-or Aqua-MODIS satellites that coincidently passed over Weda Bay on the same day. The TIR DN of MODIS band 30 and 31 values at the same coordinates as the SST measurements were extracted. These TIR DN were then correlated to the SST measurements using multiple linear regression, as follows, to estimate the SST.
The data used in the development of the SST prediction model were the data collected from February to August 2007 (see Table 1), while other data outside of those dates were used for validating the model. To establish the statistical significance of the regression models, the correlation coefficient (r), standard error of the mean estimate (SE(Y)) and F values were used. In an ideal case, r should approach 100%, SE(Y) should be zero (0), and the F value should be at least 4-5 times greater then F criterion (set at 99 or 95% of confidence level) which is an indication of negligible bias [21][22][23]. If those criteria were achieved, then we used the model for mapping the multi-temporal of the SST for the entire Weda Bay. The products of the map consist of single daily SST, 10-day average SST and average monthly SST maps.

Model development and SST mapping
A total of 51 data sets of in-situ SST measurements from February to August 2007 were collected on dates the Terra-and Aqua-MODIS satellites both passed over Weda Bay. Since in one day it was possible to get images from Terra-satellites in the morning and Aqua-satellites in the afternoon, the available data sets actually numbered more than 51 ( This model was developed using only 50 data sets out of the total available data sets (69). Because the accuracy or the quality of the insitu SST measurements was unknown, 19 datasets were excluded due to suspected bias, especially for low in-situ SST measurements (27.59~28.16°C) in February and parts of March. This model introduced a high bias that tended to underestimate -1.45 to -2.26°C for low in-situ SSTs in February and March. However, statistically, the above model is good enough to estimate the SSTs in Weda Bay after excluding those biased SST data. This was indicated by high correlation coefficient (r=0.81), low Root Mean Square (RMS) error (0.29°C), and the ratio between F value and F criterion (set at 95% of confidence level) was 38.94: 3.17 or F value >10 times greater than F criterion, which an indication of negligible bias [21]. Figure 2 displayed the plot between in-situ SST measurements versus predicted SST, while Figure 3, showed its residual plot or plot of the bias (discrepancy between in-situ SST and predicted SST). At the bias range equal to ± 0.5°C, all data sets (100%) lay inside this range, but for more accurate prediction of ± 0.25°C, there were only 24 points (48%).  However, the range of 0.5°C is good enough for predicting the SST, and therefore, the model was applicable for mapping the SSTs in Weda Bay. A total of 108 images consisting of 69 images of Terra-and 39 images of Aqua-MODIS (Table 2) was used for making the SST maps. This table showed that the MODIS data are more available during March to May, but less in July and November.

SST maps of Weda Bay
The SST maps of Weda Bay consisted of daily, 10-day average and monthly maps of each month from January to November (Figures 1 to  22 in Appendix 1). The daily SST images in Appendix 1 showed that the SST ranges in the Weda Bay were narrow, between 28 to 30°C, and flat without any distinct fluctuation of SSTs throughout the year. There are several dates when both Terra-and Aqua images were acquired in the same day. In almost cases, the Aqua image showed SST values higher by 0.4 to 0.6°C than the Terra images. This is due to the Aqua images being acquired in the afternoon (PM 13:30 local time), which is the warmer than the SST measured in morning when the Terra satellite passed over the Weda Bay (AM 10:30 local time).
Calculation of the average values of SST in Weda Bay was done only for 10-day average and monthly SST maps. Calculations were not carried out for the entire bay, but only for a window of 290 × 200 pixels or approximately 72 × 50 km square as shown in Figure 4. The SST values from that cropped window were then extracted and averaged.  The monthly SST plots in Figure 6 shows that they distribute in narrower ranges than the daily SST and 10 days averaged SST. The fluctuation of monthly SSTs is only 0.5°C between 29.25 and 29.75°C over the whole year. In comparison with the 10-day average SST plot in Figure 5, the SST distributions are more dynamic with its range of 1.5°C, between 28.75 and 30.25°C. Therefore, in future studies it is necessary to make 10-day average SST maps rather than monthly maps, because 10-day average SST maps are more useful for monitoring purposes.

SST map applications
As already mentioned, temperature is one of the principal controls on all physical, chemical and biological processes in the environment. Thus, one of the important SST map applications is to know about and monitor upwelling in a water body such as Weda Bay. Upwelling is a very well-known physical process in the sea. During upwelling, water from the deeper layer with low temperature, high salinity, high level of nutrients (phosphorus and nitrogen) and low dissolved oxygen goes to the surface of the sea. This phenomenon is important to fisheries, because after the upwelling event, the productivity of the sea increases and this generates higher numbers of fish. In the upwelling areas, a low SST is always found compared to the higher SST of the surrounding waters. The SST differences could reach 2 to 4°C, depend on the strength of the upwelling. There are many upwelling places in the Indonesian seas such as near Banda and Flores, in the Maluku Seas and Makassar Straits, south of Java and Bali and west of Sumatra (Indian Ocean).
Judging from the SST maps only (Figures in Appendix 1,8,9), the finding in this study indicated that no upwelling phenomena occurred in Weda Bay during 2007. However, upwelling in this bay could happen in the future or may have happened in the past. Since Terraimages have been available from 2000 till now, and Aqua-MODIS from 2002 to the present, we need to conduct long-term monitoring of the SST of Weda Bay. Furthermore, SST is not the only parameter that could detect the upwelling from the space. Another parameter such as chlorophyll-a concentration, which indicates phytoplankton abundance and is known as the ocean color parameter, is another parameter that could easily be monitored too. Therefore, SST and chlorophyll-a concentration are the best parameters for effective and efficient monitoring of the upwelling phenomena in the Weda Bay from the space. This must be done in the near future in order to get a clear understanding of the characteristics of this bay.

Concluding Remarks
• Monitoring and mapping the daily, 10-day average, and monthly SST of Weda Bay for 2007 using Terra-and Aqua-MODIS satellites images have been conducted with good results. Those satellites proved their ability and usefulness as monitoring tools. Reinart et al. found the same result in monitoring lake temperatures in Sweden, where temperature varies greatly with space and time. • The SST distribution in the Weda Bay from the daily and 10-day average SST maps showed a flat and narrow ranging of SST of 2°C, between 28 and 30°C all over the year. Monthly SST maps showed a narrower range of 0.5°C, between 29.25 and 29.75°C. This finding indicated no upwelling phenomena occurred in this bay during 2007. • It is necessary to conduct long-term monitoring of Weda Bay from space, not only SST, but also for the ocean color parameter (chlorophyll-a concentration), in order to get a clear understanding of the characteristics of this bay that are useful for managing the environment.