Using SPM observations derived from MERIS reflectance in a data assimilation scheme for sediment transport in the Dutch coastal zone
2007
Suspended Particulate Matter (SPM) is an important parameter affecting the marine environment by influencing light conditions in the water. Remote sensing can further our understanding of SPM in the Dutch North Sea when its products are assimilated in sediment transport models, and in situ data are also incorporated. For this aim, an optimal set of parameters consisting of SPM concentrations, error products and an approximation of optical depth was derived from MERIS data using IVM’s HYDROPT algorithm. This algorithm comprises a forward model based on inherent optical properties and radiative transfer modelling with Hydrolight, and an inverse model to estimate SPM from MERIS reflectance. These parameters were checked for: (1) accuracy of nearshore bio-optical retrieval and atmospheric correction algorithms, (2) possibility to capture change between observations under conditions of non-uniformous spatio-temporal coverage, (3) optical depth versus depth of model layers and depth of stratification. The results showed that remote sensing has much to offer, and there are ample opportunities for improved characterisation of SPM with a data assimilation scheme based on Ensemble Kalman filtering, which enables integration of SPM in a numerical sediment transport model, e.g., WL | Delft Hydraulic’s Delft3DWAQ. INTRODUCTION: NEARSHORE SPM Suspended Particulate Matter (SPM) affects the environment by attached pollutants, the transport and chemical fate of organic micro-pollutants and trace metals. SPM also influences the light conditions in the water, causing a decline of primary production with consequences for upper trophic levels. To obtain information on SPM, concentrations in the Southern North Sea have been monitored with remote sensing techniques (Eleveld et al., 2004: Fettweis et al., 2007) and modelled with sediment transport models (Gerritsen et al., 2000; Fettweis et al., 2007). Derivation of accurate SPM values for this highly dynamic coastal sea where large-scale circulation, tidal currents and riverine fresh water inputs occur is notoriously difficult. SPM retrieval from ocean colour remote sensing is dependent on good atmospheric correction, and characterisation of the high variability in Inherent Optical Properties (IOPs) in Case 2 waters. Modelling suffers from propagation of uncertainties in hydrodynamic forcing and SPM behaviour, in addition to uncertainties in the parameterization of water-bed exchange of sand-mud mixtures. Data assimilation (DA) targeted at the Dutch part of the North Sea was already discussed in (Gerritsen et al, 2000, Vos et al., 2000), but state-of-the-art developments in the characterisation of IOP variability in remote sensing science (Van der Woerd et al., 2004) and water-bed exchange of sand-mud mixtures in sediment transport modelling (Winterwerp & Van Kesteren, 2004) make an update desirable. Furthermore, future sand extraction for construction of a seaward extension of the Port of Rotterdam (‘De Tweede Maasvlakte’) might cause a temporary increase in SPM concentrations and transport along the coast, making an assessment of the T0 situation in coastal SPM concentrations necessary. The numerical sediment transport model used for the DA is presented in Blaas et al. (2007), and encouraging first results from DA are presented by El Serafy et al. (2007). This paper focuses on particular challenges and hot topics in remotely sensed nearshore SPM observations, which are: (1) The large number of scatterers (high sediment load) near the coast causes reduction of optical depth, possibly saturation of the signal and might impede the atmospheric correction (Ruddick et al., 2000); (2) The number of observations per pixel vary due to cloudiness and MERIS Level 2 quality flag settings. In the mean time major changes in SPM concentrations between observations can occur (Fettweis et al., 2007), particularly by resuspension during windy conditions (Eleveld et al., 2004); (3) Remote sensing (RS) allows estimation of SPM over a top layer of the North Sea (optical depth), in a region where salinity stratification (De Boer et al., 2006) occurs, whereas the model solves the mass balance over the full water column in 10 layers varying with water depth and incorporates exchange with bed. Information on optical depth needs to be incorporated in the DA to eliminate or decrease any possible mismatch between observed SPM concentrations (and derived mass), and predicted mass for the corresponding depth layer. This paper presents preliminary results from tackling these challenges using the advanced HYDROPT algorithm and indicates the opportunities that they offer for data assimilation. METHOD: CUSTOMISING AND ANALYSING HYDROPT MERIS SPM PRODUCTS Dataset for the data assimilation The MEdium Resolution Imaging Spectrometer instrument (MERIS) is an imaging spectrometer on board ESA’s ENVISAT spacecraft. SPM in the North Sea was studied with MERIS RR MEGS 7.4 / IPF 5.03 atmospherically corrected (Level 2) data (ESA, 2007). All MERIS RR and selected MERIS FR data covering the North Sea for 2003 were acquired and all water pixels that pass the PCD1_13 confidence checking were processed using HYDROPT (Van der Woerd and Pasterkamp, 2007). HYDROPT comprises of a forward model that generates water-leaving radiance reflectance (ρw) as a function of, a.o., the Inherent Optical Properties (IOPs) absorption (a) and scattering (b) of North Sea water and its constituents chlorophyll (CHL), SPM and coloured dissolved organic matter (CDOM) It is based on radiative transfer modelling with Hydrolight (Mobley & Sundman, 2001a and b) REVAMP IOPs (Tilstone et al., submitted) weighted (by optimisation) with the annual mean of independently collected (MWTL) in situ concentration measurements for the Dutch coast (Rijkswaterstaat, 2007). The inverse model estimates the concentrations of, a.o., SPM from MERIS water-leaving radiance reflectance ρw data at 7 optical wavelength intervals based on the Levenberg-Marquard optimisation. The inversion comprises of χ 2 fitting the modelled to the measured water-leaving radiance reflectance, and also renders standard errors (σ) with the retrieved CHL, SPM and CDOM concentrations. In addition, probability was derived from the (cumulative) distribution function for the χ 2 distribution, and ESA’s Level 2 Product Confidence Data (PCD) flags (ESA, 2007) were passed on (Van der Woerd and Pasterkamp, 2007). Additional to modelled reflectance, complementary vertical diffuse attenuation coefficient (KD) values were generated, and KD at 560 nm, which inverse can serve as an approximation of optical depth was added to customised hdf files for further analysis in a Fortran based Ensemble Kalman filtering DA toolbox (Figure 1). Nearshore coastal quality checks To support the DA process, quality checks were performed on selected near-coastal subsets from the Level 2 and its accompanying (HYDROPT-processed) Level 3 dataset. (1) Results of the ocean colour algorithm were validated by plotting SPMrs and SPMis against time (t) for all 19 coastal stations which range in distance from the coast from 2 to 235 km. Additionally, atmospheric parameters and HYDROPT SPM and error products were studied along a transect. (2) Rectified maps were subtracted to characterise spatio-temporal (ST) change between observations. (3) A first approximation of optical depth ς= 1/KD560 was calculated. Figure 1: Approach. Assimilation of SPM into the model grid. Error characteristics for all data sources are used in weighting. KD560 provides estimation of optical depth. HYDROPT MERIS SPM PRODUCTS Input for the data-assimilation The following files were generated with remote sensing for the Ensemble Kalman Filtering toolbox: • Metadata: extracted from filename, and additional Level 2 tot Level 3 processing lineage, • primary products: lat, lon, SPM, • error products: χ 2 ρw , P (cdf χ 2 ), σSPM, L2flags • KD560 Examples and nearshore characteristics of the data set are presented in the following sections. compare and optimize conc 400 500 600 700 0 0.005 0.01 0.015 0.02 observed 400 500 600 700 0.005 0.01 0.015 0.02 w avelengt h (nm) re fle c ta n c e RT Model IOPs modelled SPM, KD and error products
Keywords:
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
15
References
4
Citations
NaN
KQI