The total value of reconstructed bilateral trade flows matched the corresponding total imports in FishstatJ (Figure 1B). At the same time, the number of nodes in ITN-Fish was consistently higher than the numbers of reporters in the UN Comtrade and traders recorded in FAO FishstatJ (Figure 1A), which indicated that ITN-Fish was able to identify economies involved in fish trade that did not report trade data to the UNSD or FAO. The number of nodes in ITN-Fish had fluctuated over the years, as some countries had merged or dissolved, and some entered or exited the fish trade. The number increased from 193 in 1990 to 227 in 2000, mainly due to the dissolution of the former Soviet Union and the Southern African Customs Union, and has remained around that value since 2000 (Figure 1A).

Figure 2 shows the trend in ITN-Fish of developing countries (including other developing countries, ODC and least developed countries, LDC) and developed countries (DC) in 1990, 2000, 2010, and 2018. Developing countries (ODCandLDC) have increased their share of fish trade from 43% to 52% of global export value and from 10% to 31% of import value from 1990 to 2018 (see Supplementary material). The share of trade within “other developing countries” (ODC → ODC) has increased and it is in the range of 13%−18%, while the share of trade within developed countries (DC → DC) has decreased and it is in the range of 15%−37%. Moreover, in terms of trade shares, developed countries have expanded their exports to developing countries, while developing countries have contracted their exports to developed countries (Figure 2).

Figure 3 shows the network nodes' static properties in 2018 and the dynamic properties from 1990 to 2018. Nodes with higher kⁱⁿ generally had higher k^out (Figure 3A). Average in-degree 〈kⁱⁿ〉, out-degree 〈k^out〉, and undirected-degree 〈k〉 were increasing (Figure 3B), implying that participants in the fish trade had been actively establishing trade connections with more traders during 1990–2018. The probability of finding a more significant degree decayed exponentially (Figure 3C), which indicates that most nodes had low degrees and a small number of edges were connected to them. In contrast, the few nodes with higher degrees were the hubs of ITN-Fish. In 2018, the λ of in-degree distribution was higher than that of out-degree distribution (Figure 3A). The λ of in-degree remained stable, while the λ of out-degree showed a decreasing trend, indicating the heterogeneity in the distribution of fish export channels diminishing from 1990 to 2018.

In-strength and out-strength distributions approximate a lognormal (Figure 3D). The lognormal distribution of out-strength was left skewed, and the negative skewness was decreased. In contrast, the skewness of in-strength lognormal distribution fluctuated around zero. The concentration of exports remained stable, with the top five exporters accounting for about 30% of total exports and the top 10 accounting for approximately 50%; the concentration of imports was declining but still higher than that of exports (Figure 3E). An important reason for the decline of import concentration is the dispersion of import share. In other words, the share of major importing countries tends to decline. Exporting countries are exporting aquatic products to more countries, rather than concentrating on the original major importing countries, such as the US, Japan, and Spain, which indicates the expansion of trade relations. However, this decentralized imagination has not yet been reflected in exports, and export concentration has remained stable.

Figure 3F shows the in, out, and net strength of the top 50 nodes sorted by tot-strength in 2018. The US was the largest fish trader, the largest importer, and the largest net importer; China (CHN), the largest exporter and third largest importer, had almost equal imports and exports in 2018, after becoming a net exporter in 1990, 2000, and 2010 (Figure 2); Japan was the second largest importer and the second largest net importer in 2018; and Norway was the second largest exporter and the largest net exporter in 2018. Among the major players in the fish trade, only a few have almost balanced fish trade (imports equal exports), such as China, Netherlands (NLD), Denmark (DNK), and Poland (POL), yet fish trade was unbalanced in most economies.

In 2018, the gap as a share of total trade (TNR) was 49%, a percentage that had declined since 2000 (Figure 3G); while the average of each trader's gap as a share of their respective total trade (ANR) had remained ~65%. Since TNR is a weighted form of ANR and major traders hold most of the trade weights, that is, a few countries have most of the trade value, while most countries have only a small amount of trade value. Based on this, we can infer that the fish trade imbalance of small traders is higher than that of large traders. In addition, the imbalance of large traders is decreasing.

Additionally, both in- and out-degree distributions showed extended tails, so exponentially truncated power-law degree distributions could characterize the degree distributions more precisely (Achard and Bullmore, 2007; Fornito et al., 2016a). However, as fewer nodes are in the tails, the corresponding functions were not fitted in this study.

Figures 3H, I show the in-centrality and out-centrality of the top five nodes ranked by the mean of in-centrality and out-centrality during 2010–2018, respectively, as well as their ranking in 1990, 2000, 2010, and 2018 (see Supplementary material for the centrality and ranking of all nodes in each year). In terms of in-centrality, the US and Japan had been in the top two positions; while China's in-centrality had proliferated and jumped to the third. As for out-centrality, China and Norway had been in the top two positions since 2005; Vietnam's out-centrality had grown rapidly and had jumped to the third position.

The directed and undirected densities of ITN-Fish were increasing (Figure 4A), indicating that the connectivity of the network had been improved. The increasing densities also indicated that traders were actively establishing new trade relations, consistent with the trend reflected in the average degrees (Figure 3B).

The evolution of the network was moderate and continuous. Although only 18% of links and 29% of weights overlapped in the 1990 and 2018 networks (Figure 4B), 62–74% of links and 78–86% of weights overlap in two adjacent years (i.e., 26–38% of links and 14–22% of weights changed each year) (Figure 4C). Furthermore, in the weighted or unweighted network, the stability of ITN-Fish was improving, in other words, the variability of the network was slowing down.

Due to the number of nodes in the network caused by the dissolution of the former Soviet Union, the path length increased in the initial years. Then, it showed a decreasing trend, especially the unweighted path length, which had declined steadily since 1993 (Figure 4D). In 2018, the unweighted path length was 1.75, meaning that any two fish traders, on average, needed only 1.75 steps (or 0.75 extra traders as mediator) to be connected. Moreover, the mediators were always the hubs (Fornito et al., 2016a), such as the US and China, that had connections with most traders (Figure 3A) and were at the center of the network (Figures 3H, I).

The network was highly clustered. The unweighted clustering coefficient decreased during 1990–2000. It had increased since 2000 (Figure 4E). But even at the lowest point in 2000, the coefficient was as high as 0.73, i.e., there was a 73% probability that was also a trade relation between any two trade partners of a trader. The trend of the weighted clustering coefficients was not stable and differed from that of the unweighted in some periods (Figure 4E). For example, during 1993–1998, the weighted increased while the unweighted was stable, and during 2004–2016, the weighted increased and then decreased, with a peak in 2009, while the unweighted slowly increased.

The small-world index was close to zero, so ITN-Fish was a small-world network. The index increased in 1990–2000 and decreased in 2004–2018 (Figure 4F). But the shift in the index trends did not imply a reversal of fish globalization.

In 1990, the network was highly clustered, i.e., there was a high probability that there was also a trade relationship between any two trade partners of a trader. Despite the sparse links between traders, and hubs, such as the US and Japan, they still allowed the network to achieve connectivity, with an unweighted path length of only 1.87. Thus, the small-world index, derived from the unweighted clustering coefficient and the unweighted path length, was close to zero, implying trade groups' presence (Palla et al., 2007). Within these groups, nodes and groups were connected through hubs, forming a small-world network.

During 1990–2000, the network density was increasing, and path length was decreasing, while the clustering coefficient was decreasing, i.e., the newly established trade relations made it less likely that two trade partners of a trader also had a trade relation. This indicated that these new relations were established between nodes from different trading groups rather than within the original groups. As a result, the small-world index increased, and the network showed some random characteristics in this period.

During 2004–2018, the continuously increasing density and decreasing path length indicated higher connectivity, while the clustering coefficient and small-world index shifted to increase, which indicated that most of the new relations were established within the groups. In the previous period, nodes within the original group had been connected and then were connected to out-group nodes, which led to a decrease in the clustering coefficient. Therefore, in this period, the groups were merging, as the out-group nodes were connected to more nodes within the original groups and were integrated into the groups, which led to an increase in the clustering coefficient and decreasing small-world index.

Therefore, from a network topology perspective, the globalization of fish trade thrived during 1990–2018, and the globalization process was characterized by different patterns in periods 1990–2000 and 2004–2018, with 2001–2003 as a transition period.

The international trade network constructed in this study, ITN-Fish, contains more traders than those reported in FAO and UN trade data. That is, ITN-Fish identifies those economies that do not report trade data at FAO and the United Nations, and the problem of missing fish and fisheries product trade data in the study of Gephart and Pace is improved (Gephart and Pace, 2015). To sum up, most countries and territories in the world are involved in the fish trade. The number of traders participating in the international trade network has been relatively stable in recent years. In 2018, ITN-Fish identified 229 traders and trade links increased.

In terms of import influence, since 2010, the US, Japan, China, Spain, and France have been imported trade hubs, and China's import centrality has been rapidly increasing. In terms of export influence, China, Norway, Thailand, the US, and Vietnam are export trade hubs, and the export centrality of China and Vietnam has increased rapidly. In general, the proportion of imports and exports of developing countries in the corresponding global share continues to rise. The above research results are consistent with the following published research. Yamao's 2006 research results show that Japan, the European Union, the US, and East Asia are the four major consumer markets in the global fish market (Yamao, 2006). In the study published by Gephart and Pace in 2015, the global ocean trade network from 1994 to 2012 was constructed, and the results showed that the trade influence of developing countries increased (Gephart and Pace, 2015). The observed changes in international trade flows may be the combined effect of increased production and reprocessing capacity (Natale et al., 2015). Taking China as an example, China increasingly absorbs imported fish and fish products from all over the world, some of which are consumed domestically and some are reprocessed for export (Zhang, 2014), which has driven neighboring countries to develop a large number of fish and fisheries products for export (Wilkinson and Rocha, 2009; Golub and Varma, 2014) and trade enhancement.

During the period 1990–2000, the network density increased, the path length between countries decreased, indicating that new trade links were established between countries, and the clustering coefficient of the network decreased, indicating that this connection crossed the boundaries within the group. From 2004 to 2018, new trade links were also established between the countries. The difference is that the external links of the group are gradually integrated into the internal links, and the various groups tend to integrate—the transformation of the development model—in the process of globalization mentioned above. The global fish trade helps consumers access natural resources across borders (Berkes et al., 2006). Based on the import, export, and net strength results of our study, fish trade is likely to be more balanced in most economies, with smaller traders more unbalanced than large traders. Further to the findings of Gephart et al. (2016) countries with lower per capita GDP are vulnerable to external shocks from the global fish trade network. Thus, the process of globalization eliminates imbalances and promotes sustainable development.

Comprehensive node and network structure evolution, this study shows that ITN-Fish is becoming increasingly tight, and the globalization of trade in fish and fisheries products will continue. Based on this, research on food security issues and sustainable development of aquatic products should be based on a global rather than a local perspective (Oosterveer and Sonnenfeld, 2012). Taking economically underdeveloped countries as an example, the contradictions mainly focus on export-oriented fishing behavior for economic benefits, which brings about domestic protein supply risks. The decline of fisheries in Ghana and West Africa stems from this (Atta-Mills et al., 2004). In addition, external turbulent factors such as trade disputes will affect a more expansive area through the international trade network (Maggi and Staiger, 2018; Wessells and Wallström, 2019). Considering this, aquatic product trade may need to incorporate a broader socio-environmental perspective.