1. Introduction
Safe marine navigation depends crucially on the forecast of ocean state, especially along the ship route. Prior information of the ocean state would highly benefit mariners to ensure the maximum safety and crew comfort, minimum fuel consumption, minimum time underway, or any desired combination of these factors. In recent times, like all other oceans, the Indian Ocean (IO) also experienced an increase in maritime activities, such as transportation, fishing, and sailing, which in turn increased the demand for a reliable and accurate ocean state forecast (OSF) along the ship routes across the IO. To satisfy this increasing demand, the Earth System Science Organisation (ESSO)–Indian National Centre for Ocean Information Services (INCOIS) started a new service, namely, Ocean State Forecast Along Ship Routes (OAS) in March 2013. The forecast is extensively used by the ships plying the Indian Ocean region (INCOIS 2015). Efforts are being made to develop an interactive ocean forecast and ship routing system for mariners using the two-way communication system of the Indian National Satellite (INSAT).
Wave height and wind speed are the major factors affecting smooth voyage of the ship. Moreover, the optimum routing is normally considered attained if the effects of wind and seas can be optimized (Bowditch 2002). On that account, the OAS service mainly provides forecasts of significant wave height (Hs) and ocean surface wind, 5 days in advance and at 3-hourly intervals, with daily updates. These forecasts are routinely validated and calibrated using coastal and open-ocean buoys (Balakrishnan Nair et al. 2013). These in situ datasets are sparse and do not cover the entire ship routes. The most suitable way to ensure the reliability and accuracy of the forecast along the ship route is the validation, at least for sample test cases, of the forecast using the ship-mounted wave height meter (SWHM), which is mounted on the same ship. An SWHM was mounted on the Oceanographic Research Vessel (ORV) Sagar Nidhi, which plies fairly long routes covering the three oceanic regimes, namely, the tropical northern Indian Ocean (TNIO; north to the equator), tropical southern Indian Ocean (TSIO; equator to the Tropic of Capricorn), and the extratropical southern Indian Ocean (ETSI; the Tropic of Capricorn to 60°S).
A comprehensive evaluation of the OAS is carried out using the measurements from SWHM for the period 2011–12, and is presented in this article. Harikumar et al. (2013) presented a detailed validation using similar collocation methodology, but for surface meteorological parameters including wind, using ship-mounted INCOIS Real-Time Automatic Weather Stations (I-RAWS). Though the ORV Sagar Nidhi covers the entire IO meridionally (as shown in Fig. 2), it does not cover all the national and international ship routes across the IO. Another limitation of the above-mentioned validation is that, it does not give a complete seasonal picture of each region, since the observing time and space are limited by the ship’s schedule and track. As far as marine navigation is concerned, the seasonal validation of the wave characteristics is essential. Because of the spatial and temporal limitations of validation using only the SWHM data, we have also carried out an alternative evaluation of wave height characteristics using those derived by altimeter, with dense coverage, across the entire IO and for all seasons. The quality of radar altimeter data is arguable, so a thorough validation of satellite altimetry information is needed (Caballero et al. 2011). The reliability and accuracy of satellite-derived Hs have been established in global oceans and coasts (Hithin et al. 2015). Moreover, a considerable sample of global radar altimetry data is available to support the spatial validation of Hs (Staabs and Bauer 1997). To this end, a comparison of altimeter-derived Hs and available ESSO-National Institute of Ocean Technology (NIOT) moored-buoy-derived Hs has been carried out, as explained in the forthcoming section. This comparison ensured the validity of using altimeter wave data as a reference standard against which to evaluate OAS-forecasted Hs in the IO.
The objective of the present study is to carry out an extensive evaluation of the OAS in the IO using available SWHM measurements and satellite data during the period 2011–12.
2. Experimental techniques and data used
a. ESSO-INCOIS OAS service
OAS forecasts are generated by a suite of state-of-the-art numerical models customized to simulate and predict the IO features realistically. Atmospheric forecast products from the European Centre for Medium-Range Weather Forecasts (ECMWF) and ESSO-National Centre for Medium Range Weather Forecasting (NCMRWF) are used for forcing the models. The Hs forecast is done by using the third-generation spectral wind wave model MIKE 21 SW, which is based on unstructured meshes and is developed by the Danish Hydraulic Institute. A short description of its source term can be found in Sørensen et al. (2004). The operational model setup at INCOIS is well explained in Balakrishnan Nair et al. (2013).
b. ESSO-INCOIS SWHM
SWHM is a small, inexpensive, and simply structured microwave Doppler radar sensor having 10 mW of power and 10.525 GHz of frequency. It provides an accurate method of measuring wave amplitudes and periods while the ship is streaming (Akio et al. 1985). The complete SWHM system consists of a sensor unit, an accelerometer, a connection box, and a signal processor (Fig. 1a). The sensor head is mounted directly over the waves to be measured. Signals from it in conjunction with that from the accelerometer, which is used to remove ship motion from the wave amplitude measurements, are sent to the connection box and then to the signal processor unit, which derives the Hs (Tsurumi-Seiki Co. 2009). The specifications of the SWHM unit are shown in Table 1. The SWHM was first tested by means of measuring waves generated in a wave tank. Then, a field test was carried out in which the wave record was analytically compared with that measured simultaneously by an ultrasonic wave height meter. The results suggest that the SWHM agreed reasonably well (>95%) with the ultrasonic wave height meter (Akio et al. 1985).
(a) Schematic representation of the components of SWHM. (b) SWHM on board ORV Sagar Nidhi. (c) Closeup shot from the ship deck.
Citation: Journal of Atmospheric and Oceanic Technology 32, 11; 10.1175/JTECH-D-15-0047.1
SWHM system specifications.
ORV Sagar Nidhi SWHM-measured hourly Hs data for the period 2011–12 were used for the validation of OAS Hs.
c. Satellite altimeters
Cryosat-2 carries the Synthetic Aperture Radar Interferometric Radar Altimeter (SIRAL) built by the European Space Agency (ESA) and launched in 2010. It orbits at an inclination of about 92° and at an altitude of 717 km. SIRAL is a Ku-band (13.575 GHz) instrument providing high-quality Hs data. Repetivity of Cryosat-2 is 369 days with a 30-day subcycle. The reason for selecting Cryosat-2 altimeter Hs data for validation is its dense spatial coverage. The Cryosat-2 data were downloaded from the GlobWave FTP server (IFREMER 2014). Values of Hs from Jason-1 and Jason-2 satellites are also used. The accuracy of Jason altimeters is 4.2 cm (NASA 2014).
d. Moored buoys
As mentioned earlier, 3-hourly Hs data measured by moored buoys, deployed by the ESSO-NIOT, have been used for validating the altimeter-measured and SWHM-derived Hs. The locations of buoys are shown in Fig. 2, and the data availability for the selected years is given in Table 2.
Ship tracks of ORV Sagar Nidhi during January 2011–December 2012 drawn as per the data received from SWHM. The TNIO, TSIO, and ETSI are labeled. Black filled circles represent the buoy locations.
Citation: Journal of Atmospheric and Oceanic Technology 32, 11; 10.1175/JTECH-D-15-0047.1
Buoy data availability.
3. Methodology
Discriminating power for the statistical tests is enhanced by the ample size of the datasets and the stringency of the quality control measures imposed upon the data. Beyond the standard procedures, we have employed a procedure on the SWHM data similar to that of Caballero et al. (2011), discarding all observations for which Hs > 20 m or Hs < 0.1 m, as well as any observation not falling within the ship’s track.
Data collocation
Both forecast data and observational data should be prepared in a similar manner for a meaningful comparison. SWHM data were 1-h averages and essentially point measurements, while the model forecast data were 3-hourly averages over a 0.10° × 0.10° area. Hence, for collocation purposes, the forecast data corresponding to the grid(s) through which the ship track passed through were extracted at a 3-hourly interval. SWHM data, averaged over the past 3 h, were used to compare the values at the forecast output time interval. When the ship passed through more than one model grid during the averaging period, the area- and time-weighted averages of the forecast data were compared with the SWHM data. Thus, a collocated dataset was prepared. For a regionwise analysis, the IO was divided into TNIO (0°– 30°N, 30°–120°E), TSIO (0°–23.5°S, 30°–120°E), and ETSI (23.5°–60°S, 30°–120°E) (Fig. 2) on the basis of major swell generation areas suggested by Alves (2006).
The window size is established as X = 50 km longitude, Y = 50 km latitude, and T = 30 min.
Although the forecast model extends up to 30°N–60°S, 30°–120°E, for comparison we have considered the region 55°S–25°N, 35°–115°E; that means the outer regions in the model setup have been treated as a sponge layer, wherein the forecasts are not considered to be very accurate and therefore are eliminated. Based on the confidence in the accuracy of GlobWave Cryosat-2 and Jason altimetric data as inferred from comparison with buoys, we used these altimetry sources to further evaluate the OAS. The data from the forecast model grids coinciding with the satellite track are extracted first. Satellite data in a particular 1° × 1° grid are averaged, and then the OAS data at that particular region are computed using the same methodology adopted for the OAS-SWHM Hs comparison. These collocated datasets are used for seasonal and regional evaluation of the forecasted Hs.
4. Result and discussions
Various statistical measures like standard deviation (SD), mean bias (MB), root-mean-square error (RMSE), scatter index (SI), and correlation coefficient (R) are used to assess the OAS performance using buoys, SWHM, and altimetry. The Student’s t test was used throughout this study to compare datasets at the 95% confidence level. All test statistics in the study were found to be statistically significant. More formally stated, for each comparison the hypothesis that the datasets had unequal means was rejected with 95% confidence.
a. Intercomparison of SWHM, buoy, and altimeter Hs
To verify the SWHM-measured Hs, it is compared with buoy Hs and altimeter-derived Hs. In the SWHM–buoy comparison, it is clear that there is a slight (average = 0.2 m) underestimation of SWHM Hs (Fig. 3). The number of collocation points obtained is 32. As mentioned above, it was found that all the test statistics are significant at the 95% confidence level, though the collocation data points are fewer. This gives confidence in using the data from the SWHM for further evaluation of OAS. Altimeter Hs data from Cryosat-2, Jason-1, and Jason-2 are compared with SWHM-measured Hs in the IO for the study period 2011–12. Sabique et al. (2013) compared these altimeter-derived Hs data with that from available ESSO-NIOT moored-buoy-derived Hs and ensured the quality of the altimeter wave data in the IO. Figure 4 represents the scatterplot of the comparison between SWHM Hs and altimeter Hs. For the entire altimeter data during the study period, 67 collocation points are obtained with SWHM data, in different regions of the IO. Statistical comparison for these collocation points revealed that altimeter Hs is in very good agreement with the SWHM Hs with a high correlation of 0.97 and a mean bias of only 0.15 m. As SWHM measurements agree well with buoy measurements and altimeters observations, SWHM Hs can confidently be used for the validation of the OAS forecast.
Comparison of SWHM Hs and buoy Hs in the IO region. Dashed lines indicate the 30% error zone.
Citation: Journal of Atmospheric and Oceanic Technology 32, 11; 10.1175/JTECH-D-15-0047.1
Comparison of SWHM Hs and altimeter Hs in the IO region. Dashed lines indicate the 30% error zone.
Citation: Journal of Atmospheric and Oceanic Technology 32, 11; 10.1175/JTECH-D-15-0047.1
b. Validation of OAS forecast with SWHM
The OAS and SWHM Hs are compared for the period 2011–12. There were intermittent data gaps for the considered period, and the available periods are shown in Fig. 2. It is worth mentioning here that the data availability is 55% during the months June and July, when the sea is mostly rough. Figure 5 shows the time series comparison of OAS day 1 forecast Hs and SWHM Hs for each cruise period. The comparison along the route shows very good agreement in Hs variability in all three oceanic regimes (regions through which the ship transits are marked in the time series itself for better clarity in Fig. 5). Though these regions differ greatly in their wave characteristics, the good agreement obtained in the pattern of variability of Hs is quite encouraging. However, the analysis of the time series shows ~10% overestimation of forecasted Hs in the low wave heights. These are the preliminary interpretations from the time series comparison of Hs. Further assessment on the skill of the OAS forecast is done by computing the error statistics for the three regions (TNIO, TSIO, and ETSI) separately (Table 3). The performance of OAS Hs, forecasted up to 5 days, in these three regions of IO is validated. Error statistics analyses of the three regions suggest a strong correlation (R = 0.8) throughout the period with low scatter index (SI < 0.25), except for TNIO. As mentioned earlier, a slight overestimation of Hs is seen in the low wave heights. Among the three regions, low wave heights are mainly seen in TNIO, and hence the bias obtained for this region was high, with a high value of SI. Because wave growth and dispersion are determined by the actions of wind, most of the wave model errors are attributed to inaccuracies in representing winds. The recent study by Remya et al. (2014) shows that ECMWF-analyzed wind has a positive bias in the low range of wind speeds across the IO. The MIKE 21 SW model, which is used for generating the OAS forecast, is forced with ECMWF-forecasted winds, and so overestimation in the lower values of Hs is expected. From Table 3, it can be seen that as the forecast lead time increases, the quality of the forecast deteriorates slightly. As the routes of the ORV Sagar Nidhi ship cover most of the national and international ship routes (Fig. 2), it can be said that OAS forecasts are reliable along these routes based on the error statistics (Table 3). To study the reliability and accuracy of the OAS wave height forecast, we compared values of OAS Hs with SWHM Hs and with wave heights estimated from altimeters on board the Cryosat-2, Jason-1, and Jason-2 satellites for the period January 2011 to December 2012. The results are shown in Fig. 6. From the time series plot (Fig. 6), it is clear that SWHM-measured Hs and altimeter-derived Hs are in good agreement, and both agree well with the OAS Hs also. To get a complete picture of the spatial and seasonal performance of the OAS forecast, spatial and seasonwise evaluations, respectively, are carried out using altimeter-derived Hs as discussed in the following sections.
Time series plots of SWHM Hs and OAS first-day forecasted Hs. OAS first-day forecasted Hs are represented by gray-filled circles, and black-filled circles represent the observed Hs by SWHM.
Citation: Journal of Atmospheric and Oceanic Technology 32, 11; 10.1175/JTECH-D-15-0047.1
Error statistics for the OAS forecast–SWHM Hs comparison.
Time series plot showing the comparison between the OAS Hs (gray), SWHM Hs (black), and satellite altimeter Hs (star).
Citation: Journal of Atmospheric and Oceanic Technology 32, 11; 10.1175/JTECH-D-15-0047.1
c. Validation of altimeter-derived Hs with in situ measurements
Cryosat-2, Jason-1, and Jason-2 altimeter-derived Hs are compared with measured buoy data for the period 2011–12. The purpose of this comparison was to reconfirm whether the altimeter datasets can provide realistic estimates of Hs. The validation is done for both the Arabian Sea and for the Bay of Bengal by selecting six buoys from each basin (Fig. 2). A scatterplot of collocations for altimeters with the buoys shows that satellites agree well with the buoy data (Fig. 7). Less scatter is observed for the entire range of Hs with a high R value of 0.94 and a mean bias of only 0.06 m.
Scatterplot of buoy Hs vs altimeter-derived Hs for 51 collocations obtained in the Arabian Sea and in the Bay of Bengal. Dashed lines indicate the 30% error zone.
Citation: Journal of Atmospheric and Oceanic Technology 32, 11; 10.1175/JTECH-D-15-0047.1
d. Validation of seasonal and spatial OAS forecast with altimeter measurements
With the confidence of good agreement of altimeters with buoys, now the altimeter data, which are spatially complete (compared to point-based buoy data), have been utilized to evaluate the OAS forecast. As far as the marine navigation across the IO is concerned, the seasonal changes in the wind and its effects on waves are very important. We can classify the seasons as February–May (FMAM), June–September (JJAS), and October–January (ONDJ).
During February–May (Fig. 8a), more than 95% of OAS- and altimeter-derived Hs collocations are inside the 30% error zone. The forecast wave parameter with SI of less than 30% is extensively accepted by the user community for operational planning (Woodcock and Greenslade 2007). The OAS forecast across TNIO shows a clear overestimation, as altimetry-derived wave heights are mostly in a low range of 1–2 m (Fig. 8a). An overestimation of ~10% is seen in Hs at low wave heights in the comparison between OAS and SWHM Hs also, and the possible reason is already discussed in section 4b. This, indirectly, shows that both SWHM- and altimetry-derived Hs are in good agreement.
Scatterplot of altimeter vs first-day forecasted Hs (m) in the TNIO, TSIO, and ETSI during (a) FMAM, (b) JJAS, and (c) ONDJ. The number of collocations of each 0.1-m bin has been contoured. Dashed lines indicate the 30% error zone.
Citation: Journal of Atmospheric and Oceanic Technology 32, 11; 10.1175/JTECH-D-15-0047.1
Forecasted Hs and altimeter-observed Hs agree well (bias = −0.07 m) in the TSIO region. The error statistics analyses of the ETSI region, where high waves are seen throughout the year, show very good agreement with altimeter-derived Hs. As expected, the quality of the forecast deteriorates slightly with the increase in the forecast lead time (Table 4). The forecasted Hs shows good agreement with a high value of R (>0.75) and low RMSE (<1.0 m) for all forecasts with 1–5-day lead times, which shows the capability of OAS forecast for predicting Hs accurately well in advance.
Error statistics for the OAS forecast-altimeter Hs comparison for the FMAM season. ALT denotes altimeter.
The months of June–September (Indian summer monsoon season) are very challenging for marine navigation across the IO due to the very rough condition of the sea. Hence, a reliable and accurate forecast of Hs is of paramount importance for voyage safety. Validation of Hs shows excellent performance of the OAS forecast in all three regions (Fig. 8b). Higher R and lower SI are observed in all three regions of the IO during June–September compared to other seasons. The first-day forecasted Hs validation shows extremely good agreement with low SI (<15%) and high R (~0.9) (Table 5). Excellent agreement of mean and standard deviation also can be seen in the case of the TNIO. For the TNIO and TSIO, the comparison of all collocations is inside the 30% error zone. But in the ETSI few collocation points are outside, however, which are very close to the 30% error zone (Fig. 8b). The quality of the forecast deteriorates marginally while the forecast lead time increases, as expected, but during June–September, the measure of deterioration is less compared to other seasons (Table 5).
Error statistics for the OAS forecast–altimeter Hs comparison for the JJAS season.
During the October–January season, the performance of the OAS forecast in all three regions was good (Fig. 8c; Table 6). Overestimation of lower Hs (bias = 0.14 m) is visible in TNIO, similar to February–May. As seen in other seasons, OAS-forecasted Hs and altimeter Hs collocations are within the 30% error zone for the TNIO and TSIO regions during October–January (Fig. 8c). In this season also, the forecast for the TSIO region is very accurate (bias = 0.05 m, SI = 0.14) with a high correlation of 0.89. The ETSI region shows good agreement in the pattern of variability of Hs (R = 0.94), but with an underestimation of 0.21 m.
Error statistics for the OAS forecast–altimeter Hs comparison for the ONDJ season.
To get a spatial picture of the quality of forecasted Hs across the entire area of IO, test statistics (R, MB, RMSE, SD, and SI) are computed every 1° bin and plotted spatially for the entire period 2011–12 (Fig. 9). A very high correlation is obtained (>0.85) with an exception of a few localized regimes in the equatorial IO region. The correlation coefficient R in the Arabian Sea is found to be high compared to the Bay of Bengal. In the Arabian Sea, an overestimation of 0.4 m is obtained. In the region near the western boundary of the ETSI, forecasted Hs is highly underestimated (>1 m). A possible reason might be the closed boundary of the model in the west, which blocks the propagation of swells from the Atlantic Ocean to the domain. The SI for the entire region is below 30%, and within most of the region it is under 20%.
Spatial test statistics (a) R, (b) mean bias, (c) RMSE, and (d) SI derived for the altimeter Hs and OAS-forecasted Hs comparison.
Citation: Journal of Atmospheric and Oceanic Technology 32, 11; 10.1175/JTECH-D-15-0047.1
5. Conclusions
OAS is an advisory service of ESSO-INCOIS, which helps mariners to ensure safe navigation in the IO in all seasons and under all weather conditions. In this study, we conducted an extensive evaluation of OAS Hs in the Indian Ocean using SWHM, buoys, and altimeters.
The comparison of Hs variability along the ship routes with that from the SWHM shows very good agreement in all three oceanic regimes, namely, TNIO, TSIO, and ETSI. Even though the forecast errors in all these regions were very close to the acceptable limit of 30%, both first-day and second-day forecasts show good agreement with SWHM in all three regions. As the forecast lead time increases, the quality of the forecast deteriorates slightly.
Assessment of the seasonal and spatial performance of the OAS forecast is carried out using altimeter-derived values of Hs, which were initially validated with open-ocean buoys and were found to be in good agreement over the three different regions in the IO. During February–May, the OAS forecast over the TNIO shows slight overestimation (mean bias = 0.14 m), with a possible reason being errors in the forecasted winds. The OAS forecast is found very reliable for the three regions during June–September, a season that demands particularly thorough advance deliberation with regard to route planning. High R and low SI are observed in all three regions of the IO during June–September compared to other seasons. During October–January, overestimation of the lower range of Hs is evident in the TNIO region, similar to February–May. In general, the performance of the OAS forecast in all three regions is good. OAS shows good correlation, with small bias and small SI, for the entire 5-day forecast range. The analyses carried out and discussed above clearly show the OAS reliability over IO for all the seasons. The consistency of OAS in delivering accurate and reliable forecasts for all seasons and regions of the IO would greatly assist mariners in planning for voyage safety and efficiency well in advance of departure.
Now, ESSO-INCOIS is designing an interactive tool for transmitting the forecast service directly to ships at sea using the Indian National Satellite (INSAT) system. The forecast model setup still needs to incorporate boundaries from the global wave model to improve the forecasting in the Southern Ocean and south Indian Ocean.
Acknowledgments
We thank Kaviyazhahu, Rakhi Kumari, and Remya, and ESSO–INCOIS for providing support during various stages of this work. Rajashekhar and the VMC team from ESSO-NIOT, Chennai, India, are thanked for providing the necessary help for the installation and maintenance of the wave height meter on board the ORV Sagar Nidhi. We also thank the Earth System Science Organisation, Ministry of Earth Sciences, for the financial support. The moored buoy data for the analysis were provided by ESSO-NIOT. Cryosat-2 data for the analysis were provided by GlobWave, and Jason data were provided by NASA. We thank the anonymous reviewers for their suggestions and comments, which led to remarkable improvements of this manuscript.
REFERENCES
Akio, Y., Kuwashima S. , and Kanai Y. , 1985: A shipborne-type wave-height meter for oceangoing vessels, using microwave Doppler radar. IEEE J. Oceanic Eng., 10, 138–143, doi:10.1109/JOE.1985.1145094.
Alves, J. H. G. M., 2006: Numerical modeling of ocean swell contributions to the global wind-wave climate. Ocean Modell., 11, 98–122, doi:10.1016/j.ocemod.2004.11.007.
Balakrishnan Nair, T. M., and Coauthors, 2013: Performance of the Ocean State Forecast system at Indian National Centre for Ocean Information Services. Curr. Sci., 105, 175–181.
Bowditch, N., 2002: The American Practical Navigator. Paradise Cay Publications, 896 pp.
Caballero, I., Gómez-Enri J. , Navarro G. , and Villares P. , 2011: Towards a validation of Envisat RA-2 high rate significant wave height in coastal systems: Case study of the Gulf of Cadiz. Proc. Fifth EARSeL Workshop on Remote Sensing of the Coastal Zone, Prague, Czech Republic, European Association of Remote Sensing Laboratories, 120–130. [Available online at http://www.earsel.org/SIG/CZ/5th-workshop/manuscripts/012.pdf.]
Dinesh Kumar, E., Sannasiraj S. A. , Sundar V. , and Polnikov V. G. , 2013: Wind-wave characteristics and climate variability in the Indian Ocean region using altimeter data. Mar. Geod., 36, 303–318, doi:10.1080/01490419.2013.771718.
Durrant, T. H., Greenslade D. J. M. , and Simmonds I. , 2009: Validation of Jason-1 and Envisat remotely sensed wave heights. J. Atmos. Oceanic Technol., 26, 123–134, doi:10.1175/2008JTECHO598.1.
Harikumar, R., Balakrishnan Nair T. M. , Bhat G. S. , Nayak S. , Reddem V. S. , and Shenoi S. S. C. , 2013: Ship-mounted real-time surface observational system on board Indian vessels for validation and refinement of model forcing fields. J. Atmos. Oceanic Technol., 30, 626–637, doi:10.1175/JTECH-D-11-00212.1.
Hithin, N. K, Remya P. G. , Balakrishnan Nair T. M. , Harikumar R. , Kumar R. , and Nayak S. , 2015: Validation and intercomparison of SARAL/AltiKa and PISTACH-derived coastal wave heights using in-situ measurements. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., doi:10.1109/JSTARS.2015.2418251, in press.
IFREMER, 2014: Globe wave data. IFREMER, accessed 12 January 2014. [Available online at ftp://eftp.ifremer.fr/waveuser/globwave/data.]
INCOIS, 2015: Forecast along ship routes. INCOIS, accessed 4 January 2015. [Available online at http://115.113.76.77/shipforecast/shipforecast.html.]
Kessler, W. S., and McCreary J. P. , 1993: The annual wind-driven Rossby wave in the subthermocline equatorial Pacific. J. Phys. Oceanogr., 23, 1192–1207, doi:10.1175/1520-0485(1993)023<1192:TAWDRW>2.0.CO;2.
Monaldo, F., 1988: Expected differences between buoy and radar altimeter estimates of wind speed and significant wave height and their implications on buoy-altimeter comparisons. J. Geophys. Res., 93, 2285–2302, doi:10.1029/JC093iC03p02285.
NASA, 2014: Sea level. Jet Propulsion Laboratory, California Institute of Technology, accessed 2 February 2014. [Available online at https://sealevel.jpl.nasa.gov/missions/jason/.]
Remya, P. G., Kumar R. , and Basu S. , 2014: An assessment of wind forcing impact on a spectral wave model for the Indian Ocean. J. Earth Syst. Sci., 123, 1075–1087, doi:10.1007/s12040-014-0450-z.
Sabique, L., Balakrishnan Nair T. M. , Srinivas K. , and Nayak N. S. , 2013: Comparison of grid averaged altimeter and buoy significant wave heights in the northern Indian Ocean. Mar. Geod., 36, 72–85, doi:10.1080/01490419.2012.699499.
Sørensen, O. R., Kofoed-Hansen H. , Rugbjerg M. , and Sørensen L. S. , 2004: A third-generation spectral wave model using an unstructured finite volume technique. Coastal Engineering 2004: Proceedings of the 29th Conference on Coastal Engineering, J. McKee Smith, Ed., World Scientific, 182–186.
Staabs, C., and Bauer E. , 1997: Statistical comparison of global significant wave heights from Topex and ERS-1 altimeter and from operational wave model WAM. Phys. Chem. Earth, 23, 581–585.
Tsurumi-Seiki Co., 2009: Shipborn wave height meter. User manual, 4 pp. [Available online at http://www.tsk-jp.com/tska/pdf/SWHM.pdf.]
Woodcock, F., and Greenslade D. J. M. , 2007: Consensus of numerical model forecasts of significant wave heights. Wea. Forecasting, 22, 792–803, doi:10.1175/WAF1021.1.
