The C-band synthetic aperture radar (SAR) data from the Bohai Sea of China, the Labrador Sea in the Arctic and the Weddell Sea in the Antarctic are used to analyze and discuss the sea ice full polarimetric information reconstruction ability under compact polarimetric modes. The type of compact polarimetric mode which has the highest reconstructed accuracy is analyzed, along with the performance impact of the reconstructed pseudo quad-pol SAR data on the sea ice detection and sea ice classification. According to the assessment and analysis, it is recommended to adopt the CTLR mode for reconstructing the polarimetric parameters σHH0,σW0,H,and α, while for reconstructing the polarimetric parameters σHV0,ρH-V,λ1 and λ2, it is recommended to use the π/4 mode. Moreover, it is recommended to use the π/4 mode in studying the action effects between the electromagnetic waves and sea ice, but it is recommended to use the CTLR mode for studying the sea ice classification.
The neon flying squid, Ommastrephes bartramii, is one of the most important cephalopod with great potential for economic development, widely distributed over the Pacific Ocean (Roper et al., 1984). The life history stages of O. bartramii are affected by the ambient oceanographic regimes and the epipelagic environment (Alabia et al., 2016; Igarashi et al., 2017), and its spatial and temporal distributions are highly related to the variability in various oceanographic variables (Yu et al., 2015). Additionally, based on the relationships between the environmental variables and the distribution of the O. bartramii, the potential fishing ground and the habitat suitability of the O. bartramii can also be detected and assessed (e.g., Cao et al., 2009; Chen et al., 2011; Nishikawa et al., 2014). As such, understanding of the relationship between the oceanographic environmental factors and the spatio-temporal distributions of O. bartramii is essential for predicting its potential habitat pattern in the Pacific Ocean.
Several oceanographic variables, such as chlorophyll a concentration (Chl a) (e.g., Chen et al., 2010; Yu et al., 2017; Nishikawa et al., 2014) and sea surface temperature (SST) (e.g., Chen et al., 2007; Yatsu et al., 2010; Yu et al., 2020) are demonstrated to affect the habitat variations of O. bartramii in the Northwest Pacific Ocean. In order to deduce the distribution of O. bartramii, previous studies (e.g., Gong et al., 2012; Alabia et al., 2015; Wang et al., 2015, 2016; Yu et al., 2016a, 2016b, 2021) usually build the model between the distribution of O. bartramii and the environmental factors. Generally, traditional models are directly built based on the satellite-based oceanographic environment variables (e.g., Chl a and SST). However, the Chl a and SST data products cannot fully describe the spectrum characteristics of the oceanic surface. On one hand, Chl a and SST are not the only indicators for the ocean water. On the other hand, uncertainties in Chl a and SST remain after making several corrections during the data processing (e.g., Cui et al., 2020; Gentemann and Hilburn, 2015). As a result, the connection between conventional satellite-based oceanographic variables and distribution of O. bartramii may be not able to accurately represent the habitation of O. bartramii under different oceanic conditions.
In fact, the Chl a and SST measurements are not the raw information of the satellite observations. The Chl a can be estimated based on the ratio (O’Reilly et al., 1998; O’Reilly and Werdell, 2019) or difference (Hu et al., 2012, 2019) of spectral remote sensing reflectance (Rrs) at blue and green bands. In addition, the SST can be retrieved with different algorithms (e.g., Shibata, 2006; Wentz and Meissner, 2007; Meissner and Wentz, 2012; Merchant et al., 2008, 2009) based on the brightness temperature (BT). As such, the Rrs and BT measurements are the more neglected remote sensing information than the Chl a and SST, respectively.
In this study, the neglected remote sensing Rrs and/or BT data are firstly introduced to simulate and predict the distribution of O. bartramii with the feed-forward back propagation (BP) artificial neural network (ANN) model in the Northwest Pacific Ocean. In order to assess the performance of Rrs and/or BT on representing the distribution of O. bartramii, the ANN- stimulated and -predicted CPUE of O. bartramii are compared with the nominal CPUE from in situ daily fishery logbook data. Moreover, in order to clarify the superiority of the neglected remote sensing data to the conventional oceanographic variables, the performance differences between them on predicting the distribution of O. bartramii are also investigated.
2.
Materials and methods
2.1
Data sources
The O. bartramii daily fishery logbook data were obtained from the Chinese Squid-Jigging Technology Group of Shanghai Ocean University from July to December during 2004–2018. These data include fishing dates, daily catch (tonnes), fishing effort (days fished) and fishing locations (latitude and longitude) for the Chinese commercial squid fishery operating on the traditional fishing ground between 35°–50°N and 145°–175°E in the Northwest Pacific Ocean. The western stock of winter–spring O. bartramii accounted for most of the catch in the western Pacific Ocean with no bycatch. Chinese squid-jigging fishing vessels were equipped with almost identical engine, lamp and fishing power. These data were compiled into monthly data and grouped using 1°×1° grid cells. As a result, a total of 416 grids (26 columns by 16 rows) are generated. The monthly nominal catch per unit effort (CPUE) in one fishing unit of 1°×1° can then be calculated by
where $ {\mathrm{C}\mathrm{P}\mathrm{U}\mathrm{E}}_{y,m,i} $ is the monthly nominal CPUE, $ {C}_{y,m,i} $ is the total catch for all the fishing vessels within a fishing grid, $ {F}_{y,m,i} $ is the number of fishing vessels within one fishing grid, i is fishing unit at 1°×1° grids, m is month and y is year. In this paper, the derived monthly nominal CPUE was used as a reliable index of squid abundance, as well as the response variable to assess the performance of the prediction model.
The Level-3 Moderate Resolution Imaging Spectroradiometer (MODIS) Chl a and the neglected Rrs monthly data, collected by both Terra and Aqua from 2004 to 2018, were acquired from National Aeronautics and Space Administration (NASA) Goddard Space Flight Center (http://oceancolor.gsfc.nasa.gov). Both Chl a and Rrs data are at a 9 km×9 km (at nadir) spatial resolution. The MODIS Chl a is calculated using an empirical relationship derived from in situ measurements and Rrs in the blue-to-green region of the visible spectrum. Specifically, Rrs at 465 nm, 555 nm and 645 nm spectral regimes are used to estimate the near-surface MODIS Chl a product via merging the standard OC3/OC4 (OCx) band ratio algorithm (O’Reilly et al., 1998) and the color index of Hu et al. (2012). Therefore, only MODIS Rrs data at 465 nm, 555 nm and 645 nm spectral regimes are incorporated for further analysis in this paper. In addition, after averaging the monthly Chl a (and Rrs) data from Terra and Aqua platforms, the averaged Chl a (and Rrs) data were resampled according to the location of the monthly nominal CPUE grids (i.e., at a horizontal resolution of 1°).
SST and the neglected BT measurements were retrieved from Advanced Microwave Scanning Radiometer for Earth Observing System (AMSR-E) onboard Aqua launched in May 2002 and decommissioned in October 2011, as well as the follow-on instrument AMSR2 flown on the Global Change Observing Mission (GCOM-1) launched in May 2012. In this paper, BT measurements at 6 GHz horizontal (H) and vertical (V) polarization, 10H, 10V, 23 V, 37H and 37V are used for comparisons with SST. Both AMSR-E and AMSR2 Level-3 monthly data products, with a spatial resolution of 25 km and 10 km, respectively, were obtained from Japanese Aerospace Exploration Agency (JAXA; https://sharaku.eorc.jaxa.jp/AMSR/index.html). In order to obtain the spatiotemporally synchronized matchups between CPUE and SST (and BT) data, SST and BT data are also resampled at 1°×1° grid cells.
2.2
CPUE modelling
In this paper, we use the ANN to build the prediction model between the CPUE of O. bartramii and oceanographic information. The ANN is a model that is motivated by the biological neural network of the human brain and is used to simulate the processes that depends on a huge number of unknown inputs (Priddy and Keller, 2005). ANN is the collections of interconnected neurons that interchanges information among each other, and the connections are weighted and adjusted to get appropriate results. ANN contains mainly input layer, hidden layer and output layer. Neurons in the input layer accept the inputs for further processing. Neurons in the hidden layer accept the input from input layer with allotted weights, as well as forward the output to the output layer. In the output layer, neurons are represented with expected attribute values to the external world as output.
Back propagation (BP) algorithm is utilized in the layered feed-forward ANN in this study to build the prediction model of O. bartramii in the Northwest Pacific Ocean. The feed-forward BP neural network is a supervised learning ANN and based on the learning rule for decreasing error till the ANN becomes skilled at the data training (Wang et al., 2015). In addition, the Levenberg–Marquardt algorithm (trainlm) training algorithms is selected during the implementation of the feed-forward BP ANN (Zhang et al., 2015). Trainlm training function is based on Levenberg–Marquardt optimization and updates the bias to weight values. Trainlm falls in the supervised training algorithm category and serves as the fastest BP algorithm (Sangwan et al., 2020). The disadvantage of trainlm is that it takes more memory than other algorithms.
In this study, the input layer of the feed-forward BP ANN may include one or more kinds of remote sensing data, and the output layer is the CPUE of O. bartramii. Additionally, we implement the ANN training at each grid cell in each month. In order to assess the performance of conventional oceanographic variables and neglected remote sensing information on building the prediction model of O. bartramii in the Northwest Pacific Ocean, six schemes are designed to build the feed-forward BP ANN model in this paper (see Table 1). The schemes differ from each other by using different data sources as the model input. Schemes Ⅰ, Ⅱ and Ⅲ use the conventional Chl a, SST and their combination as the input, and Schemes Ⅳ, Ⅴ and Ⅵ use the corresponding neglected Rrs, BT and their combination as the input. Additionally, performances of the schemes also provide us an opportunity to examine the potential of neglected remote sensing information on improving our understanding of the distribution of O. bartramii. Moreover, both the O. bartramii fishery data and remote sensing measurements are split into two temporal groups for model training and validating, respectively. The first group in July–December of 2004–2013 is used to build the prediction model between the remote sensing data and the CPUE of O. bartramii, and estimate the internal coincidence precision of the model by comparing the model output with the monthly nominal CPUE of O. bartramii. The other group in July–December of 2014–2018 is used to validate the performance of the built model in terms of external coincidence precision via comparing the model predictions with the nominal CPUE
Table
1.
The input and response data for the feed-forward back propagation artificial neural network (BP ANN) model between Ommastrephes bartramii and oceanographic information from July to December during 2004–2018 in the Northwest Pacific Ocean
In order to assess the performance of the ANN-derived distribution of CPUE of O. bartramii in the Northwest Pacific Ocean, we analyzed the precision of the prediction model of O. bartramii. Figure 1 displayed the root mean square error (RMSE) of the ANN-simulated CPUE of O. bartramii from July to December during 2004–2013 with the conventional oceanographic variables (Figs 1a–c) and the neglected remote sensing data (Figs 1d–f) as input. Uncertainties in the ANN-simulated CPUE of O. bartramii with Chl a as input (Fig. 1a) generally exhibited similar spatial distribution to those with Rrs as input (Fig. 1d). Additionally, the overall RMSE of the ANN-simulated CPUE in Schemes Ⅰ and Ⅳ was approximately 0.49 t/d and 0.47 t/d, respectively, indicating both Chl a and Rrs were suitable to simulate the CPUE of O. bartramii with the feed-forward BP ANN. Moreover, the overall RMSE of the ANN-simulated CPUE of O. bartramii with both SST (Fig. 1b) and BT (Fig. 1e) as input was about 0.45 t/d. When the conventional Chl a and SST from July to December during 2004–2013 were combined as the input to the feed-forward BP ANN (i.e., Scheme Ⅲ), the ANN-simulated CPUE of O. bartramii agreed well with the nominal CPUE (Fig. 1c). Moreover, the combined Rrs and BT also successfully simulated the CPUE of O. bartramii (i.e., Scheme Ⅵ). The overall RMSE of ANN-simulated CPUE was approximately 0.46 t/d and 0.42 t/d in Schemes Ⅲ and Ⅵ, respectively. In general, when the conventional oceanographic variables were input to build the prediction model (i.e., Schemes Ⅰ, Ⅱ and Ⅲ), the RMS of the ANN-simulated CPUE of O. bartramii was greater than the simulation results with the corresponding neglected remote sensing data as input (i.e., Schemes Ⅳ, Ⅴ and Ⅵ) (Fig. 2).
Figure
1.
Spatial distribution of root mean square error (RMSE) of the artificial neural network (ANN)-simulated catch per unit effort (CPUE) of Ommastrephes bartramii from July to December during 2004–2013. a. Scheme I with chlorophyll a concentration (Chl a) as input; b. Scheme Ⅱ with SST as input; c. Scheme Ⅲ with Chl a and SST as input; d. Scheme Ⅳ with remote sensing reflectance (Rrs) as input; e. Scheme Ⅴ with brightness temperature (BT) as input; f. Scheme Ⅵ with Rrs and BT as input.
Figure
2.
Spatial distribution of the differential (diff.) root mean square error (RMSE) of the artificial neural network (ANN)-simulated catch per unit effort (CPUE) of Ommastrephes bartramii from July to December during 2004–2013. a. Scheme I minus Ⅳ; b. Scheme Ⅱ minus Ⅴ; c. Scheme Ⅲ minus Ⅵ.
When the conventional oceanographic variables are inputted to the ANN, remarkable uncertainties (e.g., RMSE > 7.0 t/d) were observed in the predicted CPUE of O. bartramii (Figs 3a–c). After replacing with the neglected remote sensing data as input, the uncertainties of the ANN-predicted CPUE of O. bartramii were obviously mitigated (Figs 3d–f). The overall RMSE of the ANN-predicted CPUE with conventional oceanographic variables as input was approximately 1.55 t/d, 1.29 t/d and 1.51 t/d in Schemes Ⅰ, Ⅱ and Ⅲ, as well as 1.48 t/d, 1.18 t/d and 1.30 t/d with the neglected remote sensing data as input in Schemes Ⅳ, Ⅴ and Ⅵ, respectively. Although the incorporation of the neglected remote sensing data also worsened the RMS of the ANN-predicted CPUE of O. bartramii at some grids, the RMS improvements were more widely observed in the Northwest Pacific Ocean (Fig. 4).
Figure
3.
Distribution of root mean square error (RMSE) of artificial neural network (ANN)-predicted catch per unit effort (CPUE) of Ommastrephes bartramii from July to December during 2014–2018. a. Scheme I with chlorophyll a concentration (Chl a) as input; b. Scheme Ⅱ with SST as input; c. Scheme Ⅲ with Chl a and SST as input; d. Scheme Ⅳ with remote sensing reflectance (Rrs) as input; e. Scheme Ⅴ with brightness temperature (BT) as input; f. Scheme Ⅵ with Rrs and BT as input.
Figure
4.
Spatial distribution of the differential (diff.) root mean square error (RMSE) of the artificial neural network (ANN)-predicted catch per unit effort (CPUE) of Ommastrephes bartramii from July to December during 2014–2018. a. Scheme I minus Ⅳ; b. Scheme Ⅱ minus Ⅴ; c. Scheme Ⅲ minus Ⅵ.
The distribution of fish species is highly related with the environmental variations (Chen, 2004). Hence, accurately building the relationship between fish species and ambient environment is essential to understand the habitat preferences of fish species and predict the dynamics of the fish population (Dickey, 2003; Ishikawa et al., 2009; Nakada et al., 2014). The ANN model has been used to predict the distributions of capelin (Mallotus villosus) (Huse, 2001), European eel (Anguilla anguilla) (Laffaille et al., 2003, 2004), Eurasian perch (Perca fluviatilis) (Brosse and Lek, 2002), neon flying squid (O. bartramii) (Wang et al., 2015) and skipjack tuna (Katsuwonus pelamis) (Wang et al., 2018). The input parameters of the ANN model in previous studies were mostly the conventional oceanographic variables, such as Chl a, SST, and/or sea surface height (SSH). However, these conventional oceanographic variables are not representative of the oceanic environment. In this study, the neglected Rrs and/or BT data were firstly proposed to simulate and predict the spatio-temporal distributions of O. bartramii in the Northwest Pacific Ocean based on the feed-forward BP ANN model.
The robust connections between the distributions of O. bartramii and SST have also been demonstrated by previous studies (e.g., Chen and Tian, 2005; Chen et al., 2007, 2008). Wang et al. (2015) also suggested that the favourable range of SST for O. bartramii was 11–18°C in the Northwest Pacific Ocean based on the BP ANN model. This article further demonstrated that the neglected BTs have consistent effects with SST on simulating the CPUE of O. bartramii with the feed-forward BP ANN in the Northwest Pacific Ocean (Figs 1b, e and 2b; Table 2). Moreover, we also found that BT was better than SST in predicting the distribution of O. bartramii, since the RMSE of the CPUE of O. bartramii is decreased by approximately 9% for the former (Figs 3b, e and 4b; Table 2).
Table
2.
The overall mean root mean square (RMS) of artificial neural network (ANN)-derived CPUE of Ommastrephes bartramii with different schemes (Unit: t/d)
Considering that the Chl a is a good indicator of the food availability for squid (Nishikawa et al., 2014), it is an important environmental variables that significantly affects the distribution of O. bartramii (Xu et al., 2004). This study confirmed the importance of Chl a during the simulation and prediction of the CPUE of O. bartramii in the Northwest Pacific Ocean with the BP ANN (Figs 1a and 3a). Additionally, we also found that the Rrs measurements at 465 nm, 555 nm and 645 nm were more suitable than the Chl a to simulate and predict the distribution of O. bartramii (Figs 1d, 2a, 3d and 4a; Table 2). As such, the neglected Rrs (and BT) could be a prefer data source than the conventional Chl a (and SST) in studying the habitat suitability of O. bartramii in the Northwest Pacific Ocean.
While the RMSEs of the simulated and predicted CPUE of O. bartramii with the combined Chl a and SST (Rrs and BT) as the model input were less than those with Chl a (Rrs) as the model input, they were greater than those with SST (BT) as the model input (Table 2). Moreover, the uncertainties in the simulated and predicted CPUE of O. bartramii with the Chl a (Rrs) as model input were remarkably larger than those with the SST (BT) as model input (Table 2). This indicated that the RMSE improvements with the combined parameters as input were mainly owe to the SST (BT), and that the SST (BT) was better than the Chl a (Rrs) in simulating and predicting the CPUE of O. bartramii in the Northwest Pacific Ocean. The results were also consistent with Wang et al. (2015), who found that the SST was the most important environmental factor in the formation of fishing grounds and it had the greatest influence on the prediction model.
It is also worth mentioning that despite only the CPUE of O. bartramii in the Northwest Pacific Ocean is simulated and predicted in this paper, the neglected Rrs (and BT) data could be further popularized to build the prediction model of other marine species over other sea areas. Furthermore, the corresponding neglected remote sensing information to other conventional oceanographic variables (e.g., SSH, sea surface salinity and wind stress curl) can be also further explored for studying the habitat suitability of the marine species.
Acknowledgements:
The authors thank two anonymous reviewers for their constructive suggestions and insightful criticisms that substantially improved the quality of our work.
Boularbah S, Ouarzeddine M, Belhadj-Aissa A. 2012. Investigation of the capability of the compact polarimetry mode to Reconstruct Full Polarimetry mode using RADARSAT-2 data. Advanced Electromagnetics, 1(1): 19-28
Charbonneau F J, Brisco B, Raney R K, et al. 2010. Compact polari-metry overview and applications assessment. Canadian Journ-al of Remote Sensing: Journal Canadien de Télédétection, 36(S2): S298-S315
Cloude S R, Pottier E. 1997. An entropy based classification scheme for land applications of polarimetric SAR. IEEE Transactions on Geoscience and Remote Sensing, 35(1): 68-78
Collins M J, Denbina M, Atteia G. 2013. On the reconstruction of quad-pol SAR data from compact polarimetry data for ocean target detection. IEEE Transactions on Geoscience and Remote Sensing, 51(1): 591-600
Dabboor M, Geldsetzer T. 2014. Towards sea ice classification using simulated RADARSAT Constellation Mission compact polari-metric SAR imagery. Remote Sensing of Environment, 140: 189-195
Dabboor M, Howell S, Shokr M, et al. 2014. The Jeffries-Matusita dis-tance for the case of complex Wishart distribution as a separab-ility criterion for fully polarimetric SAR data. International Journal of Remote Sensing, 35(19): 6859-6873
Dierking W, Skriver H, Gudmandsen P. 2003. SAR polarimetry for sea ice classification. In: Lacoste H, ed. Proceedings of the Work-shop on POLinSAR-Applications of SAR Polarimetry and Polar-imetric Interferometry (ESA SP-529). Frascati, Italy: CDROM, 18-22
Geldsetzer T, Yackel J J. 2009. Sea ice type and open water discrimin-ation using dual co-polarized C-band SAR. Canadian Journal of Remote Sensing, 35(1): 73-84
Gu Wei, Liu Chengyu, Yuan Shuai, et al. 2013. Spatial distribution characteristics of sea-ice-hazard risk in Bohai, China. Annals of Glaciology, 54(62): 73-79
Guo Hao, Fan Qing, Zhang Xi, et al. 2014. Multifeature fusion for po-larimetric synthetic aperture radar image classification of sea ice. Journal of Applied Remote Sensing, 8(1): 083534
Huynen J R. 1970. Phenomenological theory of radar targets [disser-tation]. Rotterdam, NW: Drukkerij Bronder-Offset
Jardon F P, Vivier F, Vancoppenolle M, et al. 2013. Full-depth desalin-ation of warm sea ice. Journal of Geophysical Research: Oceans,118(1): 435-447
Lee J S, Grunes M R, Ainsworth T L, et al. 1999. Unsupervised classi-fication using polarimetric decomposition and the complex Wishart classifier. IEEE Transactions on Geoscience and Re-mote Sensing, 37(5): 2249-2258
Lee J S, Pottier E. 2009. Polarimetric radar imaging: from basics to ap-plications. Boca: CRC Press
Nghiem S V, Kwok R, Yueh S H, et al. 1995. Polarimetric signatures of sea ice: 2. Experimental observations. Journal of Geophysical Research: Oceans, 100(C7): 13681-13698
Nord M E, Ainsworth T L, Lee J S, et al. 2009. Comparison of compact polarimetric synthetic aperture radar modes. IEEE Transac-tions on Geoscience and Remote Sensing, 47(1): 174-188
Ochilov S, Clausi D A. 2012. Operational SAR sea-ice image classifica-tion. IEEE Transactions on Geoscience and Remote Sensing, 50(11): 4397-4408
Prinsenberg S J, Peterson I K, Holladay J S, et al. 2012. Labrador shelf pack ice and iceberg survey, March 2011: Canadian technical report of hydrography and ocean sciences. v 275. Goose Bay, Labrador: Canadian Helicopters Ltd, 1-44
Raney R K. 2007. Hybrid-polarity SAR architecture. IEEE Transac-tions on Geoscience and Remote Sensing, 45(11): 3397-3404
Salberg A B, Rudjord O, Solberg A H S. 2014. Oil spill detection in hy-brid-polarimetric SAR images. IEEE Transactions on Geoscience and Remote Sensing, 52(10): 6521-6533
Scheuchl B, Cumming I, Hajnsek I. 2005. Classification of fully polari-metric single- and dual-frequency SAR data of sea ice using the Wishart statistics. Canadian Journal of Remote Sensing, 31(1): 61-72
Souyris J C, Imbo P, Fjortoft R, et al. 2005. Compact polarimetry based on symmetry properties of geophysical media: The π/4 mode. IEEE Transactions on Geoscience and Remote Sensing, 43(3): 634-646
Stacy N, Preiss M. 2006. Compact polarimetric analysis of X-band SAR data. In: Proceedings of EUSAR 2006. Dresden, Germany: EUSAR
Wakabayashi H, Matsuoka T, Nakamura K, et al. 2004. Polarimetric characteristics of sea ice in the Sea of Okhotsk observed by air-borne L-band SAR. IEEE Transactions on Geoscience and Re-mote Sensing, 42(11): 2412-2425
Yang Guojin. 2000. Sea Ice Engineering Science (in Chinese). Beijing: Petroleum Industry Press
Yin J J, Yang J, Zhou Z S. 2013. New parameters in compact polari-metry for ocean target detection. In: Proceedings of IET Inter-national Radar Conference 2013. Xi'an: IEEE, 1-6
Zhang Xi, Dierking W, Zhang Jie, et al. 2015. A polarimetric decom-position method for ice in the Bohai Sea using C-band PolSAR data. IEEE Journal of Selected Topics in Applied Earth Observa-tions and Remote Sensing, 8(1): 47-66
Sahar Ebrahimi, Hamid Ebadi, Amir Aghabalaei. Forest Classification Using Simulated Compact Polarimetry Data and Deep Learning Networks. Journal of Geospatial Information Technology, 2023, 11(1): 19. doi:10.61186/jgit.11.1.19
2.
Mohammed Shokr, Mohammed Dabboor. Polarimetric SAR Applications of Sea Ice: A Review. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2023, 16: 6627. doi:10.1109/JSTARS.2023.3295735
3.
Hangyu Lyu, Weimin Huang, Masoud Mahdianpari. A Meta-Analysis of Sea Ice Monitoring Using Spaceborne Polarimetric SAR: Advances in the Last Decade. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2022, 15: 6158. doi:10.1109/JSTARS.2022.3194324
4.
Ajeet Kumar, Anup Das, Rajib Kumar Panigrahi. Hybrid-pol Decomposition Methods: A Comparative Evaluation and a New Entropy-based Approach. IETE Technical Review, 2020, 37(3): 296. doi:10.1080/02564602.2019.1613937
5.
Mohammed Dabboor, Mohammed Shokr. Compact Polarimetry Response to Modeled Fast Sea Ice Thickness. Remote Sensing, 2020, 12(19): 3240. doi:10.3390/rs12193240
6.
Xiaochen Wang, Yun Shao, Fengli Zhang, et al. Comparison of C- and L-band simulated compact polarized SAR in oil spill detection. Frontiers of Earth Science, 2019, 13(2): 351. doi:10.1007/s11707-018-0733-9
7.
Wenli Qiao, Jinbao Song, Hailun He, et al. Application of different wind field models and wave boundary layer model to typhoon waves numerical simulation in WAVEWATCH III model. Tellus A: Dynamic Meteorology and Oceanography, 2019, 71(1): 1657552. doi:10.1080/16000870.2019.1657552
8.
Mohsen Ghanbari, David A. Clausi, Linlin Xu, et al. Contextual Classification of Sea-Ice Types Using Compact Polarimetric SAR Data. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(10): 7476. doi:10.1109/TGRS.2019.2913796
9.
Xiaochen Wang, Yun Shao, Wei Tian, et al. An Investigation into the Capability of Compact Polarized SAR to Classify Multi-Sea-Surface Characteristics. Canadian Journal of Remote Sensing, 2018, 44(2): 91. doi:10.1080/07038992.2018.1461554
10.
Xiaochen Wang, Yun Shao, Lu She, et al. Ocean Wave Information Retrieval Using Simulated Compact Polarized SAR from Radarsat-2. Journal of Sensors, 2018, 2018: 1. doi:10.1155/2018/1738014
11.
Mohammed Dabboor, Benoit Montpetit, Stephen Howell. Assessment of the High Resolution SAR Mode of the RADARSAT Constellation Mission for First Year Ice and Multiyear Ice Characterization. Remote Sensing, 2018, 10(4): 594. doi:10.3390/rs10040594
12.
Martine Espeseth, Camilla Brekke, A. Johansson. Assessment of RISAT-1 and Radarsat-2 for Sea Ice Observations from a Hybrid-Polarity Perspective. Remote Sensing, 2017, 9(11): 1088. doi:10.3390/rs9111088
13.
Yongxun Wang, Xiaochen Wang. Analysis of the Reconstruction Accuracy of Compact Polarized SAR Pseudo Quad-Pol Data for Use in Ocean Surface Observations. 2023 International Applied Computational Electromagnetics Society Symposium (ACES-China), doi:10.23919/ACES-China60289.2023.10249585
Table
1.
The input and response data for the feed-forward back propagation artificial neural network (BP ANN) model between Ommastrephes bartramii and oceanographic information from July to December during 2004–2018 in the Northwest Pacific Ocean
Table
2.
The overall mean root mean square (RMS) of artificial neural network (ANN)-derived CPUE of Ommastrephes bartramii with different schemes (Unit: t/d)
DownLoad:
DownLoad:
DownLoad: