PV-LAC-COAST:
Proba-V New Coastal Products

   
Minimize PV-LAC Introduction

The PV-LAC project aims at the development, testing, and validation of two advanced algorithms for the scientific exploitation of the PROBA-V mission, one focusing on atmospheric correction and the other on coastal products.

In the first Activity, an Optimal Estimation (OE) method is developed for the atmospheric correction of the 1 km resolution PROBA-V data. This method relies on the joint  aerosol optical depth and surface reflectance retrieval through the inversion of a physically-based coupled surface-atmosphere radiative transfer model. PROBA-V data are accumulated during several days to form a multi-angular and multi-spectral observation vector. It is assumed a greater stability of surface reflectance parameters with respect to aerosols . Additonally, surface reflectance retrieval of the previous processed accumulated period is used as prior information for the inversion. This method,  was originally developed for geo-stationary satellite observations, which provide very high temporal sampling. It is now adapted for the processing of PROBA-V 1 km data.

The second  Activity within PV-LAC focuses on the exploration of PROBA-V for the generation of coastal products, making use of only the PROBA-V central camera, that provides data at 100 m spatial resolution with global coverage every 5 days. Although PROBA-V was not conceived as an ocean colour mission, its specifications (spectral bands and signal-to-noise ratio) are suitable to monitor some of the key ocean colour parameters, in particular the Suspended Particulate Matter (SPM). Especially in estuarine environments, the higher spatial detail of the products with a sufficiently high repeat frequency could be an added value to other sensors.

For both activities, the implications for the Sentinel-3 mission will also be investigated.

More information on PROBA-V can be found at http://proba-v.vgt.vito.be/

The project started in January 2016 and will end in July 2017. It consistes of four tasks:

  1. The scientific review and requirements consolidation (April 2016)
  2. The algorithm definition (November 2016)
  3. Algorithm evaluation and validation (April 2017)
  4. Recommendations and roadmap (July 2017)
Minimize Methodological approach

Atmospheric correction identification

While typical marine atmospheric correction schemes assume a zero NIR water leaving reflectance in order to retrieve aerosol information from the image, coastal atmospheric correction algorithms have to use different hypotheses to determine the aerosol, or to correct for a non-zero NIR reflectance due to the presence of the suspended particles. In recent years, several algorithms have been developed trying to deal with the complexity of the atmospheric correction for coastal waters. First, a literature review has been made of atmospheric correction algorithms published in the past years for coastal waters and their applicability to PROBA-V data has been assessed  (see the Requirements Baseline Document). Taking into account the specifications of PROBA-V, the following two approaches were selected:

  • A/C based on spatial extension of aerosol information retrieved from nearby land (referred to as "iCOR land-based")

     

  • A/C based on extending the "black pixel" approach to the SWIR (referred to as "iCOR SWIR-based"). The SWIR black pixel approach assumes that the contribution of in-water constituents is zero, due to the high absorption of pure water in the SWIR. The signal in the SWIR can thus be assumed to be entirely atmospheric and can therefore be employed for the aerosol determination.

 

TSM or Turbidity retrieval algorithm identification

In order to select the most suitable TSM or Turbidity algorithm a literature review has been made of suitable algorithms (see Requirements Baseline Document). Based on the literature review, the following three algorithms were selected in the first stage to test:

1. The Nechad et al. (2010) single band semi-analytical TSM algorithm is selected because of its strong theoretical background and successful applications in the literature. However, we should bear in mind that the algorithm might perform less in regions where particle type and composition vary significantly from the calibration region. We propose to apply a wavelength switching to include both clearer and turbid waters.

2. A band ratio algorithm combining the PROBA-V RED and NIR bands, because this algorithm might be less sensitive to changes in particle type and composition and less sensitive to errors in the atmospheric correction.

3. The Dogliotti et al. (2015) semi-analytical turbidity algorithm is almost insensitive to the sediment type and composition. A TSM product can be derived from the turbidity product by combining the PROBA-V turbidity data with a regionally calibrated Turbidity-TSM relationship.

 

After preliminary performance analyses using in-situ datasets, the semi-analytical turbidity algorithm has been selected as the most suitable algorithm for PROBA-V. The performance of the turbidity algorithm outperformed the TSM algorithm.  Besides this, the RED/NIR band ratio algorithm showed to be only suitable for very turbid waters. Furthermore, both the one-band TSM algorithm and the RED/NIR ratio algorithm requires a site specific calibration,  which limits its global and/or operational applicability.

Minimize Algorithm Validation

Testing and Validation the A/C algorithm

For evaluating the performance of both AOT retrieval approaches (i.e., iCOR land-based versus iCOR SWIR-based) the approaches were applied to PROBA-V 100 m data acquired from the Belgian coastal waters. The retrieved AOT (550 nm) values were compared against the AOT (550 nm) values measured by the Aerosol Robotic Network Ocean Colour (AERONET-OC) CIMEL instruments located at the MOW1 platform (51.362°N; 3.120°E) near Zeebrugge harbor and the more offshore Thornton_C-power platform (51.5329°N; 02.9549°E). A threshold for the temporal offset between the time of the PROBA-V overpass and AERONET measurement was set to ± 30 minutes. Figure 6 gives the scatter plots of AERONET AOT versus PROBA-V retrieved AOT (550 nm) values for the land-based and the SWIR based A/C approach, respectively. The SWIR-based approach seems to systematically overestimate the AERONET-OC AOT values, while the land-based approach overestimates the AOT at low AOT-values, but underestimates at low AOT values. Overall the SWIR-based approach with an R² of 0.46 performs better than the land-based approach with a R² of 0.22.

 


Figure 6: Scatter plots of PROBA-V retrieved AOT (550 nm) versus AERONET AOT values for (a) the land-based and (b) the SWIR based A/C approach. The solid black line is the 1:1 line, while the dotted RED line is the regression line. The Y-error bar gives the standard deviation of PROBA-V derived AOT values within a 1km x 1km box around the Aeronet location. The X-error bar gives the standard deviation of the Aeronet AOT within ±30 minutes of the PROBA-V acquisition.
 

Testing and Validation the selected Turbidity algorithm

The retrieved PROBA-V turbidity is validated using in-situ Turbidity data from the CEFAS SmartBuoys (Mills et al, 2003). These autonomous systems are moored, automated, multi-parameter recording platforms used to collect marine environmental data. Turbidity data are typically collected every 30 minutes at 1 m water depth. The data are freely available for research purposes. In the North Sea study area there are three Buoys currently in operation. These are: 1) Warp (TH1) NMMP in the turbid waters of the Thames; 2) West Gabbard (mid-2016 replaced by West Gabbard2) and 3) Dowsing, located more offshore. One more buoy outside the study area is used, i.e., the Liverpool Bay Coastal Observatory buoy. For each PROBA-V overpass (over the test site) the turbidity of a nominal pixel containing the location of the field data measurement was extracted. The threshold for the temporal offset between the time of the PROBA-V overpass and field data measurement was set as ±1 hour.
The validation results using the turbidity from the CEFAS smartbuoys are shown in Figure 7. A nice correlation with an R² of 0.73 between the PROBA-V derived Turbidity values and the in-situ turbidity values is found for the PROBA-V data corrected on the basis of the iCOR SWIR-based approach.

 

Figure 7 Regression plots between in situ turbidity and PROBA-V turbidity for all the Smartbuoy locations values for respectively (a) PROBA-V data corrected with the land-based approahc and (b) the SWIR based A/C approach.. The Y-error bar gives the standard deviation of PROBA-V derived turbidity within the 1km x1km around the Smartbuoy location. The X-error bar gives the standard deviation of the buoy turbidity measurements performed within max. 1 hour of the PROBA-V acquisitions. The solid black line is the 1:1 line , the dotted RED line is the regression line
To complete the performance analysis, PROBA-V turbidity maps were compared to MODIS turbidity maps, as is illustrated in Figure 8. Similar turbidity levels and patterns are found. In the Scheldt estuary the advantage of the higher spatial resolution of PROBA-V (i.e., 100 m compared to 1000 m from MODIS) is clearly visible.

 

 


Figure 8: Turbidity map from the Belgian Coastal Waters/Scheldt estuary on 3 April 2017: (a) MODIS acquisition time 13:00 UTC and (b) PROBA-V acquisiton time 11:19 UTC.

A more extensive description of the validation results can be found in the Validation Report.

Minimize References
  • Dogliotti, A.I., Ruddick, K.G., Nechad, B., Doxaran, D., Knaeps, E. (2015). A single algorithm to retrieve turbidity from remotely-sensed data in all coastal and estuarine waters. Remote sensing of environment, 156: 157-168.
  • Nechad, B., Ruddick, K.G., Park, Y. (2010). Calibration and validation of a generic multisensory algorithm for mapping of total suspended matter in turbid waters. Remote sensing of environment 156:157-168