Accuracy assessment of wave data from altimeter near the coast

Ocean Engineering 178 (2019) 229–232 Contents lists available at ScienceDirect Ocean Engineering journal homepage: www.elsevier.com/locate/oceaneng Accuracy assessment of wave data from altimeter near the coast ∗ Patricia López-García , Jesús Gómez-Enri, Juan J. Muñoz-Pérez T CASEM, Universidad de Cádiz, Pol. Rio San Pedro s/n, 11510, Puerto Real, Spain A R T I C LE I N FO A B S T R A C T Keywords: Satellite Altimetry Wave height Propagation Coast This study presents a methodology to verify the notion that wave measurements could be less accurate near the coast when they are provided by altimetry radars (on-board satellites). The methodology is based on comparing significant wave height (Hs) data from the RA-2 altimetry radar on-board ENVISAT satellite (European Space Agency) against points from a Gulf of Cadiz database that were propagated through SMC software (Sistema de Modelado Costero). Nine years of SWELL waves were analysed (2002–2010). The results show a RMSE of only 0.19 m (correlation coefficient of R ≥ 0.96) in deep waters, but this decreases within 10 km of the coastline, reaching a value of 1.12 m (R = 0.16) at 5 km from the coast. 1. Introduction Measurements of significant wave height (Hs) and its variability in coastal areas are used for many purposes (e.g. sediment transport analysis, coastal wave setup, storm surges), including validation/calibration of models for wave forecasting, atmospheric modelling, and ocean circulation. These applications are useful for a wide range of purposes, such as structural design engineering on the high seas, coastal zone protection, improvements in ship route security, and planning of ocean operations. Satellite radar altimetry is designed to provide accurate information on sea surface height, Hs, and wind speed at the sea surface over the open ocean. However, Vignudelli et al. (2011) (and references therein) suggested that altimetry measurements could be less accurate in coastal areas. These issues are due to two main factors: the contamination of waveforms due to land or very calm waters entering the radar footprint (Gómez-Enri et al., 2010); and inaccurate tidal and wet tropospheric corrections (Caballero et al., 2014). Access to accurate information on coastal sea conditions is of great importance because of the socioeconomic and strategic interest of the coastal zone. This calls for new processing strategies to generate optimised altimetric products suited to the diverse applications in challenging conditions. Many studies have focused on validation of the Hs data given by satellite radar altimetry using in-situ observations (e.g. Faugere et al., 2006; Passaro et al., 2015; Kudryavtseva and Soomere, 2017), but few have focused on the influence of the proximity to the coast, although it has been set out that altimeter measurements are of lower accuracy near the coast (Queffeulou and Bentamy, 2007; Bouffard et al., 2008; Vignudelli et al., 2011; Hareef Baba Shaeb et al., 2015). With the objective of proving the reliability of altimetry data near the coast, the aim of this paper is to ∗ present a methodology for assessing comparisons between the Hs data given by altimetry data and that propagated from numerical models of virtual buoy data. This study analyses a particular area of the Gulf of Cadiz (SW of Spain) over nine years. 1.1. Study area The study area is located in the southwest of the Iberian Peninsula, in the Gulf of Cadiz (Spain) (Fig. 1a and b). It is a mesotidal area, with a tidal range of about 4 m and significant waves from the W-SW quadrant (Laiz et al., 2013). Even though the winds from E-SE have high speed and persistence, they lack the fetch length necessary to generate waves of similar heights to the W-SW waves (Muñoz-Perez and Abarca, 2009). Fig. 1c shows the wave rose of the study area (obtained from the “ROM 0.3–91 Wave Climate on the Spanish Coast (Ministry of Public Works, 1991)”). Higher probability of waves that come from the fourth quadrant (W-NNW) can be seen in this wave rose. The RA-2 altimetry radar (on-board ENVISAT European satellite) has two tracks that cross at the continental shelf of the Gulf of Cadiz (Fig. 1b); one corresponds to the descending pass 223 from land to sea (from north to south) and the other to the ascending pass 187 from sea to land (from south to north). 2. Database and methodology 2.1. Altimetry data An estimated Hs based on the reprocessed altimetry radar signal (RA-2 of the ESA ENVISAT satellite) was used in this study. Data were obtained from the new full mission reprocessing: FMR V3.0 (ftp://ra2- Corresponding author. E-mail address: [email protected] (P. López-García). https://doi.org/10.1016/j.oceaneng.2019.03.009 Received 7 May 2018; Received in revised form 14 January 2019; Accepted 4 March 2019 0029-8018/ © 2019 Elsevier Ltd. All rights reserved. Ocean Engineering 178 (2019) 229–232 P. López-García, et al. 2.4. Wave propagation Sistema de Modelado Costero (SMC) software was used to perform the wave propagation. This software was made by Hidraulic Institute of Cantabria, supervised by the Spanish Ministry of the Environment. In this way, the Hs data from the SIMAR database could be propagated to the points located along the satellite passes shown in Fig. 1, to carry out a comparison with the altimetry data. The SMC software creates propagation nets in the study area, to allow the software to calculate where and in which direction the wave propagates, and in how many data points it provides information. To that end, data from SIMAR (the Hs, average period, and average wave direction) were added to the software on a wave editor. Once the wave is calculated and propagated, the SMC connects with another software, Surfer 8, that provides a graphical output of the study area with wave height data. Fig. 1. (a) Location of the study area (South of Spain). (b) Position of SIMAR point (virtual buoy), also the two tracks performed by the satellite in this area are shown: descending pass (from north to south) and ascending pass (from south to north). The points marked from 1 to 6 and 1′ to 7′ are the zones where comparisons between the significant wave height (Hs) data given by the altimeter and the propagated data from virtual buoy were performed. (c) Wave rose of the study area. 2.5. Statistical analysis We used two statistical parameters to compare both datasets: (i) the linear correlation coefficient (R) between altimetry Hs and propagated Hs values; and (ii) the root mean square error (RMSE). 3. Results and discussion ftp-ds.eo.esa.int), Geophysical Data Record product. The along track spatial resolution of the dataset was 1 Hz, equivalent to a distance between two consecutive measurements of ∼7 km. The temporal resolution was 35 days. Table 1 shows the enumeration assigned to each point of the satellite passes, as well as the average point depth and its average distance to the coastline in kilometres. The points were enumerated according to the distance to the coastline, and were classified into 6 zones. Fig. 2 compares “altimetry Hs – propagated Hs” in the different zones. The correlation coefficients (R) are represented in Fig. 3 as a function of the distance to the coastline, with a confidence interval of 95%, of the linear adjustments of the compared data in Fig. 2. Fig. 3 shows that R is high and significant far from the coast (from 10 km to deep waters). Close to the coast (zone A), R is low. The results presented in this study agree with those by Caballero et al. (2014), who noted that the accuracy of satellite measurements is 0.5 m. It can be seen that there is a ± 0.5 m difference between the data provided from altimetry and the propagated SIMAR data. More than 95% of all the data fall into the range of ± 0.5 m, except for those nearest to the coastline (zone A), where only 60% of the data are within that range. Furthermore, the slope of the linear regression is 0.36 (67% 2.2. Virtual buoy data SIMAR database comes from the Hindcast of dynamic processes of the ocean and coastal areas of Europe (HIPOCAS) project (see the “Puertos del Estado de España” website at www.puertos.es). This wave database was obtained through numerical wave modelling (WAM) from wind time series. WAM is a third-generation spectral model that solves the equation of energy balance without a priori hypotheses about the spectral shape of the waves. Therefore, the data are virtual and do not come from direct measures of the nature. However, this database has been validated by numerous studies and has been used in many practical applications along the Spanish coasts (Tomas et al., 2004). The SIMAR database contains data generated every hour since 1958. The wave data are given in open water and undefined depths conditions, where they comply with the following relationship: h > Lo / 2 = 0.78 T2 Table 1 Grouped points of the study area for both direction satellite passes and the corresponding zone of each one. The last two points of the ascending track (6′ and 7’) were grouped (F). (1) where h is depth, T is period, and Lo is wave length in deep water (offshore). The SIMAR point with the code 5034017 was selected for this study, which can be found freely at: www.puertos.es. The point location is 6°67′ West and 36°67′ North (Fig. 1). The parameters chosen to carry out the wave propagation were the Hs (m), the average period (T, s), and the average wave direction (θ). ZONE Point Coastline distance (km) Depth (m) Descending (from land to sea) A B C D E F 2.3. Data filtering The altimetry database contained data for 140 passes of the satellite (from 2002 to 2010). Of these, 47 were not considered because of null values. The hourly points included SEA and SWELL data wave. Thus, of the 93 valid data points in the 2002–2010 period, the 53 passes that corresponded to SWELL waves were selected (43 descending and 10 ascending passes). We used the ocean depth/land elevation field available in the product and considered only ocean measurements. 1 2 3 4 5 6 3.5 10.5 16.9 22.7 27.7 32.3 2 17 35 54 70 85 Ascending (from sea to land) A B C D E F F 230 1′ 2′ 3′ 4′ 5′ 6′ 7′ 5.8 11.1 16.4 21.6 26.5 30.9 31.1 4 8 31 47 61 70 75 Ocean Engineering 178 (2019) 229–232 P. López-García, et al. Fig. 2. Comparison between altimetry Hs and propagated Hs for each zone (zone A is nearest to the coast). A range of ± 0.5 m is indicated by the two lines parallel to the linear regression. 4. Conclusions underestimation) in Zone A while it is 0.78 in Zone F (only 22% underestimation). Finally, the RMSE decreases as the distance to the shore increases, from 1.12 m in zone A to 0.19 m in deep waters. Therefore, the accuracy of the estimation for data in zones further the coastline is much better than in the nearest area. The accuracy of altimetry radar measurements in zones near the coast has been assessed in previous works (see Vignudelli et al., 2011 and references therein). In this work, SIMAR virtual buoy data was 231 Ocean Engineering 178 (2019) 229–232 P. López-García, et al. ‘coastal’ retracking approaches would be an asset. References Bouffard, J., Vignudelli, S., Cipollini, P., Menard, Y., 2008. Exploiting the potential of an improved multimission altimetric data set over the coastal ocean. Geophys. Res. Lett. 35. Caballero, I., Gómez-Enri, J., Cipollini, P., Navarro, G., 2014. Validation of high spatial resolution wave data from Envisat ra-2 altimeter in the gulf of Cadiz. Geophys. Res. Lett. IEEE 11 (1), 371–375. Faugere, Y., Dorandeu, J., Lefevre, F., Picot, N., Femenias, P., 2006. ENVISAT ocean altimetry performance assessment and cross-calibration. In: In: Chen, G., Quartly, G.D. (Eds.), Special Issue on “Satellite Altimetry: New Sensors and New Application", vol. 6. pp. 100–130 Sensors, (3). Gómez-Enri, J., Vignudelli, S., Quartly, G.D., Gommenginger, C.P., Cipollini, P., Challenor, P., Benveniste, J., 2010. Modeling Envisat RA-2 waveforms in the Coastal Zone: case study of calm water contamination. IEEE Geosci. Remote Sens. Lett. https://doi.org/10.1109/LGRS.2009.2039193. Hareef Baba Shaeb, K., Anand, Arur, Joshi, A.K., Bhandari, Satyendra M., 2015. Comparison of near coastal significant wave height measurements from SARAL/ AltiKa with wave rider buoys in the Indian region. Mar. Geodes. 38 (S1), 422–436. Kudryavtseva, N., Soomere, T., 2017. Satellite altimetry reveals spatial patterns of variations in the Baltic Sea wave climate. Earth Syst. Dynam. 8, 697–706. Laiz, I., Gómez-Enri, J., Tejedor, B., Aboitiz, A., Villares, P., 2013. Seasonal sea level variations in the gulf of Cadiz continental shelf from in-situ measurements and satellite altimetry. Cont. Shelf Res. 53, 77–88. Ministry of Publics Works, 1991. ROM 0.3-91 Waves, Annexe 1. Wave Climate on the Spanish Coast. Madrid. 80 pp. Muñoz-Pérez, J.J., Abarca, J.M., 2009. Influencia del viento y de las variaciones de la presión atmosférica en el nivel del mar de marismas y estuarios. Rev. Obras Publicas 3505, 21–32. Passaro, M., Fenoglio-Marc, L., Cipollini, P., 2015. Validation of significant wave height from improved satellite altimetry in the German Bight. IEEE Trans. Geosci. Remote Sens. 53 (4), 2146–2156. Queffeulou, P., Bentamy, A., 2007. Analysis of wave height variability using altimeter measurements: application to the Mediterranean Sea. J. Atmos. Ocean. Technol. 4 (12), 2078–2092. Tomas, A., Mendez, F.J., Medina, R., Losada, I.J., Menendez, M., Liste, M., 2004. Bases de datos de oleaje y nivel del mar, calibración y análisis: El cambio climático en la dinámica marina en España. El Clima entre el Mar y la Montaña. Asociación Española de Climatología y Universidad de Cantabria, Santander Serie A, n 4. Vignudelli, S., Kostianoy, A., Cipollini, P., Benveniste, J., 2011. Coastal Altimetry. Springer-Verlag, Berlin/Heidelberg, Germany. Fig. 3. RMSE and linear correlation coefficients (R) between altimetry Hs and propagated Hs, as a function of the distance to the coastline, for each zone (A–F). propagated through numerical modelling with SMC software, and compared to Hs data from altimetry measurements in the Gulf of Cadiz. The linear correlation coefficient between the altimetric and propagated Hs data was much lower near the coastline (R = 0.16). In this particular case, the correlation coefficient increased at distances further than 10 km to the coastline with R ≥ 0.96 in deep waters. It was observed that 95% of these data were within ± 0.5 m of the adjusted linear correlation “altimetry Hs – propagated Hs”, whereas only 60% of the data from the zone near the coast was inside the ± 0.5 m range. The RMS error between the Hs measurements from ENVISAT and propagated numerical model of Hs decrease from 1.12 m in coastal area to only 0.20 m in deep waters. Our analysis was focused on using an official product at a low posting rate: 1 Hz (about 7 km between two consecutive measurements). A more detailed study at a higher alongtrack spatial resolution (20 Hz) and using products from dedicated 232