1 INTRODUCTION

From 2005 to 2012, the average annual production of tuna in Indonesia was 1.035 million tons, accounting for more than 16% of global tuna production, and with the development in recent years, in 2016, Indonesia has become the country with the largest catch of tuna, occupying an important position in the world's fishery resources (Muawanah et al., 2018; Khan et al., 2020; Wiryawan et al., 2020). The waters around South Java, which benefit from extensive upwelling and the distribution of large fishing ports, are one of the important fishing areas and are known as productive tuna fisheries (Koropitan et al., 2021).

Upwelling is one of the most important vertical physical processes in marine systems. It transports nutrient-rich deep waters to the surface, greatly increasing the primary productivity of the water column. Upwelling is therefore of great importance to local fisheries resources (Wyrtki, 1961; Siswanto, 2010). There are several upwelling areas in the Indian Ocean, such as the area around Somalia and Oman. The Somali and Omani upwelling regions experience phytoplankton blooms that are prominent with net primary production (NPP) exceeding 435 gC/(m²·a), and are the most densely distributed areas of fish (Liao et al., 2016; Sreeush et al., 2018). In the upwelling areas around the Gulf of Omani and the Arabian Sea, catches account for over 60% of the entire Western Indian Ocean, dominated by Indian sardine, croakers, Indian mackerel, anchovies, and other fish that move in the pelagic waters (FAO, 2011). There are at least the following types of upwelling: coastal upwelling, large-scale winddriven upwelling, eddy-related upwelling, topographyrelated upwelling, dynamic upwelling, and extensive diffusive upwelling in the ocean interior (Chang et al., 2018). Wind-driven coastal upwelling is one of the best-known types of upwelling. It is also most relevant to human activities, as it supports some of the most productive fishing zones in the world, such as the Arabian Sea catch area (Chang et al., 2018). The South Java sea upwelling is a wind-driven upwelling caused by the monsoon (Purba and Khan, 2019; Horii et al., 2020).

The velocity of upwelling is slow and difficult to be observed directly. Upwelling transports cold nutrient-rich water from the deep ocean to the surface, causing a decrease in sea surface temperature (SST) and an increase in chlorophyll-a (Chl-a) concentration, so the spatial and temporal distribution of thermocline depth, SST, and Chl-a concentration are common methods for observing upwelling (Siswanto, 2010; Purba and Khan, 2019). When upwelling occurs, divergence occurs in flow field at ocean surface, making the sea surface height in the upwelling zone lower than the vicinity. Therefore, sea surface height anomaly (SSHA) data can also be used to observe the upwelling distribution (Moore II and Marra, 2002).

During the annual southeast monsoon (June– October), southeasterly winds along the South Java coast take surface waters away from the coast and are replenished by deep cold water. The processes of surface water transport and deep cold water upwelling are called Ekman transport and Ekman pumping, respectively, and the joint effect of the two forms the phenomenon of upwelling (Wirasatriya et al., 2020). Upwelling moves northwestward with the development of the southeast monsoon until it is terminated around November by the arrival of the northwest monsoon (Susanto et al., 2001).

The coastal upwelling of South Java shows different intensity in terms of interannual variation. Wirasatriya et al. (2020) confirmed that the South Java upwelling weakened in 2010 and 2016, while it strengthened in 2007, 2011, and 2015. The interannual variation in upwelling is a guideline for fisheries production assessment and therefore it is important to investigate the causes of interannual variation in upwelling. It has been suggested that climatic events such as El Niño/La Niña-Southern Oscillation (ENSO) and Indian Ocean Dipole (IOD) have a significant impact on the interannual variation of upwelling along the coastal of South Java (Susanto et al., 2001; Iskandar et al., 2009; Atmadipoera et al., 2020; Wirasatriya et al., 2020). ENSO is a seaair phenomenon that occurs in the tropical Pacific and has two modes: El Niño and La Niña (Fang and Xie, 2020). During El Niño, the upwelling along the South Java coast strengthens and during La Niña, it weakens (Atmadipoera et al., 2020). IOD is one of the major climate events occurring in the tropical Indian Ocean and is also divided into positive IOD and negative IOD, with the SST becoming lower in the eastern Indian Ocean and higher in the western Indian Ocean during the positive IOD, whereas the opposite is true during the negative IOD period (Behera et al., 2006). Wirasatriya et al. (2020) suggested that negative IOD has a weakening effect on the South Java upwelling and positive IOD enhances the upwelling. However, there are still different understandings on the main controlling factors and mechanisms of interannual variation of upwelling due to climate events (Susanto et al., 2001; Atmadipoera et al., 2020; Wirasatriya et al., 2020). Susanto et al. (2001) confirmed a correlation of 0.6 between the thermocline depth (22 ℃ as standard) and the Southern Oscillation Index (SOI), an indicator of ENSO, for the period 1981–1999, and proposed that ENSO events have a major controlling effect on upwelling by altering the local ocean current system, the Indonesian Throughflow (ITF). Atmadipoera et al. (2020) also highlighted the role of ENSO in controlling upwelling in South Java through model data analysis. In contrast, Chen et al. (2016) proposed that IOD remotely regulates South Java upwelling through equatorial latitudinal winds and that IOD has a stronger control on upwelling than ENSO through analysis of observations and model experiments. This was later supported from Wirasatriya et al. (2020) who analyzed the multiple correlations between the Oceanic Niño Index (ONI), the Indian Ocean Dipole Index (DMI), and SST, Chl-a concentration, and they found that IOD plays a stronger role in controlling upwelling than ENSO. Xu et al. (2021) proposed that the propagation of coastal-caught Kelvin waves is the cause of upwelling anomalies along the Sumatra and South Java coasts. Therefore, there is a controversy over the degree of control and mechanism of ENSO and IOD on the South Java upwelling.

Few studies have been conducted on the effects of ENSO and IOD on fisheries in coastal marine areas of South Java. Handayani et al. (2019) and Wiryawan et al. (2020) used data on skipjack tuna at Malang Harbor and large tuna at West Nusa Tenggara and proposed that environmental changes in SST and thermocline depth caused by El Niño would lead to increased landings of tuna, and attributed the increasing catches in 2012 and 2015 to the occurrence of El Niño. Lumban-Gaol et al. (2021), based on the investigation results in Palabuhanratu Bay, proposed that in 2016, when the negative IOD event occurred, the number of small pelagic fish in the sea off Java also decreased due to the weak upwelling, while in 2019, when the positive IOD event occurred, the upwelling was strong, so fish catches also increased significantly. The relationship between climatic events, upwelling, and fishery resources, and the mechanism of their interaction is unclear. Further research is needed on the impact of climatic events on fishery resources off South Java.

Based on previous studies, we present a comprehensive analysis of climate and environmental data, remote sensing data and fishery resources data in the coastal waters of South Java over a 17-year period (2003–2020). The seasonal variation, interannual variation, and influencing factors of upwelling in this sea region were clarified, and the relationships between ENSO/IOD climatic events, upwelling and fishery resources on a larger scale were investigated, and the physical mechanisms of upwelling affecting fishery resources under different climatic events are explained. The results can provide a basis for further studies on the evaluation of fishery resources and their variability in the region.

2 MATERIAL AND METHOD 2.1 Study area

Upwelling is very strong in the coastal waters of South Java (Kunarso et al., 2020). We set this sea area as the main study area (2°S–16°S, 102°E– 117°E; Fig. 1) to investigate the upwelling processes and their main controlling factors. One of the main catch areas in the South Java Sea region is Box1 (7.5°S–15°S, 105°E–116°E) (Wiryawan et al., 2020), which we set as a sub-study area and discuss the effects of variation in upwelling intensity on fish catches in Box1 in Section 4.3.

2.2 Sources of environmental data

We use the ONI and SOI from January 2003 to December 2020 to describe ENSO, and the DMI in the same time period to describe IOD. ONI and SOI were obtained from the NOAA Climate Prediction Center (https://origin.cpc.ncep.noaa.gov/products/analysismONItoring/ensostuff). The ONI is based on the three-month SST anomalies of Niño3.4 area to indicate ENSO phenomena. ENSO can be further classified as weak (|0.5|–|0.9|), moderate (|1.0|–|2.0|), and strong (>|2.0|) events based on the ONI threshold (Wirasatriya et al., 2020). The SOI calculates the sea level pressure difference at the Tahiti station in the eastern Pacific Ocean and the Darwin station in the western Pacific Ocean, and can also be used to indicate ENSO phenomena, which are usually considered normal when the SOI is between -2 and 2, and El Niño (La Niña) when it is greater than 2 (less than -2) (Susanto et al., 2001; Atmadipoera et al., 2018). The DMI indicating IOD was obtained from the NOAA website (https://stateoftheocean.osmc.noaa.gov/sur/ind/DMI.php), which is defined as the difference in the SST anomalies between the western (50°E–70°E, 10°S–10°N) and eastern parts (90°E–110°E, 10°S–0°) of the tropical Indian Ocean (Saji et al., 1999; Horii et al., 2020). IOD has three states, i.e., positive (DMI>0.5), neutral (-0.5≤DMI≤0.5), and negative phase (DMI<-0.5) (Supriyadi and Hidayat, 2020).

The SST data and Chl-a data are from the L3 dataset of the MODIS Aqua satellite, both with a spatial resolution of 4 km. As the MODIS Aqua satellite data started in July 2002, we started downloading from 2003 until 2020. Downloads are available from Asia Pacific Data Research Centre (APDRC; http://apdrc.soest.hawaii.edu/data/data.php). Both the SST and Chl-a data are monthly averaged in Section 3.1, and annually in Section 4.3. SSHA data for January 2003–December 2020 is from the AVISO website (https://las.aviso.altimetry.fr/las/UI.vm) with a spatial resolution of 0.25° and a temporal resolution of 6 days; in Section 3.2 we process it as monthly averages and in Section 4.2 retain the original temporal resolution. The monthly-averaged 20 ℃ isotherm data (D20) is European Centre for Medium-range Weather Forecasts reanalysis (ECMWF) data with a spatial resolution of 1°, also downloaded from the APDRC (http://apdrc.soest.hawaii.edu/data/data.php), and is currently only updated to December 2018, so we use the data for the period January 2003– December 2018. The wind field data for 2010, 2013, and 2015 are the Cross-Calibrated Multi-Platform wind vector analysis data (CCMP, http://www.remss.com/measurements/ccmp/) in spatial resolution of 0.125° and temporal resolution of monthly averaged in Section 4.2.1 and daily averaged in Section 4.2.2.

2.3 Fishery data source

The fishery data used in this paper for the period 2007– 2016 is from Statistics of marine and coastal resources, including annual catches, number of fulltime anglers and number of fishing vessels, published by the Indonesian National Statistical Institute (https://www.bps.go.id/). Since data such as catches are annual, we re-processed the SST, Chl-a concentration data as annual averages in Section 4.3, when discussing the effect of upwelling on annual catches. Fishery data are currently published in the bulletin information only up to the year 2016. We selected seven major economic fish species (including golden banded mackerel, flying fish, mackerel, skipjack tuna, marlin, flatfin sailfish, and dorado fish) active in the pelagic waters during 2007–2016 to explore the impact of upwelling on fishery resources in the study area.

2.4 Analytical method

We used monthly averaged SST, Chl-a, and SSHA data for the period 2003–2020 to characterize the upwelling processes in the study area. The method of Benazzouz et al. (2014) was referred to using the Temperature Upwelling Index (TUI) obtained based on the SST and modified to characterize the variation in upwelling intensity. TUI is calculated from the difference between onshore and offshore SSTs at the same longitude. We chose the onshore location at point A (110°E, 8°S) and the offshore location at point B (110°E, 12°S). A high value of TUI indicates a strong upwelling, while its low value (>0) indicates a weak upwelling, and negative TUI indicates downwelling (Benazzouz et al., 2014). Since we chose a 4° difference in latitude between points A and B, we calculated the gradient of SST variation between them to obtain a TUI indicative of upwelling strength.

TUISST(lon, time) = [SST_offshore(lon, time) – SST_onshore (lon, time)]/4.

We have calculated the Ekman transport using wind field data and first need to calculate the sea surface wind stress (Wirasatriya et al., 2020):

τ=[τ_x, τ_y]=ρ_aC_DV₁₀[u₁₀, v₁₀],

where ρ_a is the air density (1.2 kg/m³); V₁₀ is the magnitude of the wind speed at 10 m above the sea surface; u₁₀ is the latitudinal wind speed; v₁₀ is the longitudinal wind speed; and C_D is the drag coefficient, calculated using the following (Trenberth et al., 1989):

The Ekman transport calculation uses a two-parameter model based on the following equation (Lagerloef et al., 1999):

where ρ is the seawater density (1.02×10³ kg/m³); r is the friction coefficient (2.15×10^-4 m/s); h_md is the mixed layer depth (32.5 m); and f is the Coriolis parameter (Wirasatriya et al., 2020). In Section 4.2.1, to remove the effect of the seasonal signals from the equatorial latitudinal wind and SSHA, we processed using low-pass filtering, with the sampling frequency set to 1/90 for the equatorial latitudinal wind and 7/90 for the SSHA frequency.

On fishery resource data, as the Indonesian government has increased its support for fisheries in recent years and the number of fishermen has increased, so catch per unit of effort (CPUE) was used as an index of fishery resources in order to eliminate the influence of human factors (Chodriyah and Wiyono, 2011).

CPUE=catch/effort,

where catch is total annual catch, effort is the effort of fishing activities, and in article is the number of full-time fishers (in Section 3.3 for details).

3 RESULT 3.1 Upwelling variability and climate events

The series of monthly averaged SST and Chl-a concentration in the main study area show a clear seasonal cycle (Fig. 2a). SST starts to decline significantly around June, reaches its lowest value in August–September, and starts to rebound after October. Chl-a concentration is lowest in February and gradually increases from April. Chl-a concentration increases rapidly around May, reaches a maximum around September and starts to decrease after October (Fig. 2a). Chl-a concentration and SST showed a clear trend of variability influenced by upwelling. The monthly averaged SSHA and TUI time series also show seasonal variation (Fig. 2b). The trends they show reflect that upwelling begins to increase around June, with July–September being the strongest period of the year for upwelling, and after October upwelling begins to diminish, with upwelling largely disappearing by November. After November, the study area is dominated by downwelling (Fig. 2b).

In terms of interannual variability, the minimum values of SST in 2005, 2010, and 2016 were anomalously high compared with other years, while the minimum values of SST in 2003, 2006, 2007, 2011, 2015, and 2019 were even lower than the surrounding years, and the years in which the interannual variation in Chl a showed anomalies are essentially the same as those reflected in the interannual variation in SST (Fig. 3a). This indicates that upwelling weakened in 2005, 2010, and 2016, but strengthened in years such as 2003 and 2006. The SSHA and TUI time series respond in terms of interannual variability with upwelling weakening in 2005, 2010, and 2016, and strengthening in 2003, 2006, 2007, 2011, 2015, and 2019 (Fig. 3b–c). It should be noted that the TUI time series also exhibits extremely high values in 2008, but we did not consider 2008 in terms of interannual variability, considering that it is not reflected in the other index series.

According to the ONI time series for the period 2003–2020 (Fig. 4a), strong El Niño events occurred in late 2009 and 2015, and strong La Niña events occurred in late 2007, 2010, and late 2020. Another type of climatic event with considerable impact, the IOD, occurs more frequently compared to ENSO from 2003 to 2020 (Fig. 4b). Typical years of positive IOD events are 2006, 2012, 2015, 2017, 2018, and 2019, with negative IOD events occurring in 2010 and the second half of 2016.

Major positive IOD events occurred in 2006 and 2019, when upwelling showed anomalously enhanced, and briefly in 2003, 2007, 2011 when the DMI exceeded the 0.5 criterion, and in 2015 when both positive IOD and El Niño events occurred. Negative IOD events occurred in 2005 and 2016, when upwelling was anomalously weak, and in 2010, when both La Niña and negative IOD events occurred. This demonstrates the important role of climate events in regulating interannual variability in upwelling, with El Niño/positive IOD typically enhancing upwelling and La Niña/negative IOD weakening it (Susanto et al., 2001; Atmadipoera et al., 2020; Wirasatriya et al., 2020). To specifically discuss the role of typical climatic events on upwelling, in Section 4.2 we select the year 2010 (La Niña coincided with negative IOD), 2013 (Climate event indices fluctuated in the normal range), and 2015 (El Niño coincided with positive IOD), as representative years for analysis.

3.2 Correlation of upwelling with climatic events

We analyzed the correlations between the indices of the two types of climatic events and the intensity of upwelling separately (Fig. 5). With statistical significance at the 95% confidence level, the significant correlation between SOI and D20 is approximately 0.41, and and the significant correlation between ONI and D20 is approximately 0.39 (α= 0.05, P < 0.01). The correlation between ENSO and upwelling strength is relatively weak. In contrast, the significant correlation between DMI and D20 is approximately -0.57 and with TUI is approximately 0.59 (α=0.05, P < 0.05), which is at a moderate correlation level, indicating a closer relationship between IOD and upwelling than ENSO.

3.3 Interannual variability in fishery resources

Upwelling provides a large amount of nutrients that are important for fishery resources. The total annual catches of the seven fish species mentioned in the Fisheries Data Source were used to study the impact of inter-annual variability in upwelling on fishery resources. Total annual catches did not change much until 2010 (Fig. 6a–b). However, the total annual catch fluctuated considerably between 2010 and 2014 due to the impact of overfishing and the enactment of the fisheries ban (Satria et al., 2018).

To eliminate as much human interference as possible, we use the number of fishing boats and the number of full-time anglers as indicators of human activity. We analyzed the correlations between total annual catch and fishing vessel numbers, and between total annual catch and full-time fishers, with correlation coefficients of 0.34 and 0.82 at 95% confidence level (α =0.05, P < 0.01), respectively. As there was a more significant correlation between catches and the number of full-time fishers, it indicates that the number of full-time anglers is a better proxy for the impact of human activity on catch than the number of fishing vessels. We therefore defined the number of full-time anglers as the Effort parameter to calculate the interannual variation in CPUE over the period 2007–2016 (Fig. 7).

The CPUE decreased year by year from 2007 to 2009, reaching its lowest value in 2010. By 2011, however, CPUE had grown rapidly, increasing by 86.7% compared to 2010, and reaching a maximum in 2012. From 2013 to 2016, CPUE was in an overall declining trend, but in 2015 CPUE increased by 11% compared to 2014. Over the 10-year period 2007–2016, the CPUE shows an increase or decrease in variation over the years. Therefore, we discuss the relationship between climate events, upwelling and fishery resources in Section 4.3.

4 DISCUSSION 4.1 Relationship between climate events and upwelling variations

Based on the correlation analysis of the climatic event index with D20, TUI in the results, we suggest that IOD has a stronger control on the South Java coastal upwelling than ENSO, which is supported by Chen et al., (2016) and Wirasatriya et al. (2020). Susanto et al. (2001) analyzed the correlation (R= 0.6) between the thermocline depth (22 ℃ as standard) sequence and SOI from 1970 to 2000, suggesting that upwelling processes are mainly regulated by ENSO and may be partially regulated by internal Indian Ocean dynamical processes. Atmadipoera et al. (2020) took 2010 (La Niña/ negative IOD occurred simultaneously) and 2015 (El Niño/positive IOD occurred simultaneously) as examples to propose that ENSO events strongly regulate the intensity of south Java upwelling. It should be noted, however, that they did not consider the important role of IOD in this process. Susanto et al. (2001) confirmed that upwelling along the Java-Sumatra coast intensified during El Niño 1997 and was weak during La Niña 1998. Coincidentally according to the DMI indication on the NOAA website (https://stateoftheocean.osmc.noaa.gov/sur/ind/dmi.php), a positive IOD occurred at the same time as El Niño in 1997, and a negative IOD event occurred during La Niña in 1998. ENSO and IOD climate events also occurred simultaneously in 2010 and 2015 as proposed by Atmadipoera et al. (2020). In the work of Chen et al. (2016), who analyzed the correlation between wind stress, SST, thermocline depth, and climate indices for the Sumatra and southern Java seas from 2001 to 2011, respectively, DMI showed better correlation with environmental parameters such as wind stress than the ENSO index, indicating that IOD may have a stronger control on upwelling than ENSO.

After analyzing the relationship between climate events and upwelling over the period 2003– 2020, we found that: positive IOD and La Niña occurred simultaneously in 2011; positive IOD and normal ENSO in 2008 and 2012; and positive IOD and El Niño in 2015; the strength of upwelling described by the TUI was stronger in all these years than in normal years. This suggests that the enhancement of upwelling along the South Java coast by positive IOD is consistent regardless of ENSO conditions. In contrast, in 2010, when negative IOD and La Niña occurred simultaneously, and in 2016, when both negative IOD and normal ENSO occurred, upwelling strength was significantly weaker. This suggests that negative IOD has a weakening effect on upwelling. In other words, IOD has a stronger control on upwelling in the study area than ENSO, which is also supported by Chen et al. (2016) and Wirasatriya et al. (2020).

4.2 Mechanism of climate events affecting interannual variations of upwelling

To understand the mechanism of climate events on upwelling, three representative years (2010, 2013, and 2015) mentioned above were selected for targeted discussion, the specific reasons have been explained in Section 3.2. The monsoon controls the formation of upwelling along the coast of South Java (Wyrtki, 1962; Moore II and Marra, 2002), and several scholars proposed that the Kelvin wave propagation in the interior of the ocean would have an impact on upwelling by causing changes in thermocline depth (Cai et al., 2009; Nagura and McPhaden, 2010; Chen et al., 2016; Xu et al., 2021). Therefore, we analyzed the spatial and temporal distribution of wind stress, Ekman transport, and Kelvin wave propagation in the main study area.

4.2.1 Effect of monsoon on interannual variations of upwelling

In 2013, Compared to 2010 and 2015, the overall wind stress was stronger along the coast (north of 9°S) from July to September (Fig. 8). The strongest wind stress in July was about 0.12 N/m. In August, the center of wind stress moved slowly to the northwest, with areas of wind stress greater than 0.1 N/m mainly in southeastern Sumatra. In September, the strength and impact of wind stress weakened. The coastal wind stress in 2010 was the weakest of the three years, with the strongest wind stress of approximately 0.10 N/m occurring in August. The wind stress in 2015 was stronger than that in 2010, but still weaker than that in 2013, and the coastal wind stress was the strongest in August, about 0.12 N/m. By analyzing the correlation between wind stress and SST, Atmadipoera et al. (2020) showed that wind stress is an important factor influencing upwelling. In terms of interannual variability in the wind field, coastal wind stress was stronger in 2013 than in 2010 and 2015, but upwelling was the most intense in 2015, followed by 2013. This is not consistent with the interannual variability of upwelling. This suggests that wind stress may not be a major control on the interannual variability of upwelling along the South Java coast.

We calculated the spatial and temporal distribution of Ekman transport along the South Java coast during the main upwelling period (July–September) in 2010, 2013, and 2015 (Fig. 9). Ekman transport along the South Java coast was less intense in 2010 than in 2013 and 2015. The maximum transport is about 3 m²/s in August 2010. Ekman transport was more intense in normal 2013, with a maximum of about 5.5 m²/s along the coast. This indicates a weakening of the capacity of the wind stress to transport water masses in 2010 compared to 2013, which is consistent with the interannual variability of upwelling. However, Ekman transport was roughly the same spatial and temporal distribution in 2015 as 2013, even showing weaker transport near 9°S during August and September than in 2013 (Fig. 9h & l). This variation is similar to the spatial-temporal distribution of wind stress (Fig. 8). The fact that upwelling was stronger in July-September 2015 than in 2013 is inconsistent with the spatial and temporal distribution of Ekman transport and wind stress. Chen et al. (2016) quantified the remote and local forcing that leads to interannual variability in equatorial EIO upwelling from 2001 to 2011, and they proposed that interannual variability in the local wind field contributes importantly to interannual variability in SST. However, remote regulation of the equatorial Indian Ocean wind field plays a more important role in the interannual variability of local upwelling, which explains why the local wind stress caused by Ekman transport is not consistent with the interannual variability of upwelling. Therefore, we do not believe that interannual variability in the local wind field was the main cause of the anomalous intensification of the South Java upwelling in 2015.

4.2.2 Effect of Kelvin waves on interannual variability of upwelling

The equatorial Indian Ocean latitudinal winds control the propagation of Kelvin waves, i.e., equatorial easterly winds (U < 0) excite upwelling Kelvin waves and westerly winds (U > 0) excite downwelling Kelvin waves (Rao et al., 2009; Yuan and Liu, 2009; Chen et al., 2015). We analyzed the interannual variability of equatorial latitudinal winds (U) and SSHA (Fig. 10) in the above three typical years, and studied the influence of Kelvin wave propagation on the interannual variability of upwelling.

During the major upwelling period (July–October), there were significant differences in the spatial and temporal distribution of equatorial latitudinal winds. In 2013, easterly winds dominated the area between 50°E– 60°E, while lower westerly winds dominated in the area east of 60°E. A few areas of easterly wind fields developed near 90°E. In 2010, the easterly wind was relatively weak in the 50°E–60°E region. And the westerly wind was strong from the east of 60°E to the sea area off Sumatra. Therefore, a downwelling Kelvin wave was excited in 2010 compared with in 2013, resulting in shallower SSHA in the coastal region of South Java. In 2015, the equatorial Indian Ocean latitudinal wind was mostly easterly, and weak westerly wind only existed around 80°E. Therefore, the easterly wind excited an upwelling Kelvin wave in the equatorial Indian Ocean in 2015 compared with in 2013, resulting in a negative SSHA. The above results show a good correspondence between the different types of Kelvin waves excited by the latitudinal wind and the variability of upwelling intensity in the coastal of South Java. A salinity barrier layer is evident in the study area (Horii et al., 2020), which prevents the upwelling of deep cold water. The upwelling Kelvin waves can lift the thermocline and bring deep water to the surface (Thompson et al., 2006), where it is adequately mixed with surface water in the presence of local wind fields (Chen et al., 2015). The downwelling Kelvin wave does the opposite function.

Many works have demonstrated that the westerlies over the equatorial Indian Ocean usually disappear while the easterlies strengthen during the positive dipole, and the westerlies over the equatorial Indian Ocean become stronger during the negative dipole than positive dipole, there is a strong relationship between equatorial latitudinal winds and IOD (Grunseich et al., 2011). During the IOD period, different types of Kelvin waves are excited by equatorial latitudinal wind, which remotely regulate upwelling in the coastal area of South Java. We believe that this is the main factor causing the interannual variation of upwelling.

4.3 Relationship between climatic events, upwelling and fishery resources

We reprocessed the SST and Chl-a concentration data from Box1 for the 2007– 2016 period to show the annual average results of their time series (Fig. 11). The correlations of CPUE with SST and Chl-a concentration are -0.52 and 0.46 (α =0.05, P < 0.01), respectively. This indicates a relationship between SST, Chl-a concentration in Box1 and local fishery production. In 2013 and 2014, the climate indices fluctuated within the normal range, so we used the annual means of SST and Chl-a concentration for these two years to represent the intensity of upwelling during a normal year (Fig. 3). It should be noted that in 2014, Indonesia imposed a ban on fishing (Satria et al., 2018), so the CPUE maintained a downward trend after 2013. Although we have used the number of full-time fishers as a proxy for fishing effort to remove the effects of human activity, our discussion of the relationship between upwelling and CPUE is only between adjacent years in order to remove the effects of other factors such as policy. In 2011, low SST and high Chl-a concentrations indicate enhanced upwelling, and in 2010, SST and Chl-a concentration indicate weakened upwelling. CPUE was higher in 2011 than in 2010. Similarly, CPUE was higher in 2015 when upwelling was enhanced than in 2016 when upwelling was weakened (Fig. 9). This provides evidence that fish catch is largely influenced by the upwelling. This is due to the strong upwelling bringing in large amounts of nutrient salts that favour local primary productivity, ultimately resulting in higher Chl-a concentration. Sufficient plankton biomass to sustain commercial fisheries can only be found at Chl-a concentration greater than 0.2 μg/L (Sachoemar, 2012). In contrast, weaker upwelling in 2010 and 2016, lower annual averaged Chl-a concentration ultimately led to a considerable decrease in CPUE. Sachoemar (2012) also suggested that there was a positive correlation between seasonal variations in catches and Chl-a concentration, and the high Chl-a concentration was usually caused by upwelling (Tito and Susilo, 2017; Subarna, 2018; Khan et al., 2020). Therefore, variations in upwelling due to climate events can affect changes in catches in the study area.

In addition, variations in upwelling can also have an impact on fish distribution. Osawa and Julimantoro (2010) investigated the distribution of tuna in South Java and found that the area of fish activity was located near 112°E–115°E, 12°S–15°S. Temperature fronts are formed at the junction of upwelling and shallow water, where nutrients are retained, providing suitable temperatures, and feeding grounds for fish (Lumban-Gaol et al., 2002, 2015). When upwelling along the South Java coast is enhanced, the thermocline is lifted and the high nutrient waters are brought closer to the surface, allowing for increased primary productivity and shallower fish depths (Fig. 12b). On the contrary, in years such as 2010, When upwelling is weakened, high nutrient waters cannot rise to the surface and fish tend to dive deeper, which reduces the productivity of fisheries using traditional operations such as long-line fishing (Fig. 12a; Lan et al., 2012).

Handayani et al. (2019) and Wiryawan et al. (2020) used skipjack tuna data from Malang harbor and large tuna data at West Nusa Tenggara, respectively, to propose that environmental changes in seawater temperature and thermocline depth caused by El Niño increase tuna landings, and attributed the increase in catches in 2012 and 2015 to the El Niño occurrence. Handayani et al. (2019) argued that tuna production increased when strong El Niño (2015) occurred and decreased during strong La Niña (2016) (Handayani et al., 2019; Suniada, 2020), and Wiryawan et al. (2020) also argued that a weak El Niño in 2012 and a strong El Niño in 2015 resulted in enhanced upwelling, causing an increase in catches. We argue that the upwelling has a direct effect on fishery resources. In addition, we should use the variations of upwelling intensity when discussing the impact of climate events on fishery resources. In Section 4.2, we have showed that IOD events have a stronger impact on interannual variability of upwelling than ENSO. Therefore, more attention should be paid to the impact of IOD on fishery resources in the coastal waters of South Java. It is a fact that positive IOD events occurred in 2012 and 2015, when catches increased. And it was the year of negative IOD event, when catches decreased in 2016 (Handayani et al., 2019; Wiryawan et al., 2020). This result further validates the assertion that more attention should be paid to IOD affecting local fishery resources.

5 CONCLUSION

Upwelling in the coast of South Java has obvious seasonal variation. The upwelling is strong from July to September, and starts to weaken after October. Climate events such as ENSO and IOD have important effects on the interannual variability of upwelling. However, the influence of IOD on the interannual variability of upwelling is greater than ENSO. Positive IOD has a strengthening effect on the upwelling, and negative IOD has a weakening effect on the upwelling.

Monsoon and Kelvin wave propagation processes are the main ways in which climatic events affect upwelling intensity in the coast of South Java. Although the monsoon has a controlling role in upwelling formation, wind field variability is not the main controlling factor for the interannual variability of upwelling. During the IOD period, equatorial latitudinal wind variability excites different types of Kelvin waves, which have a remote forcing effect on the upwelling and is the main mechanism of action affecting the interannual variability of upwelling in the study area.

The study of the effects of climate events on fishery resources should focus on the role of climate events on upwelling. The interannual variability of upwelling caused by climate events not only affects the catches in the coast of South Java, but also affects the distribution of commercial fish stocks. Compared with ENSO, the impact of IOD on interannual variability of fishery resources in the study area is more significant. Catches were higher in years with positive IOD, and catches reduced in years with negative IOD. The results of this study can provide a scientific basis for analyzing and predicting the variations of fishery resources in the coast area of South Java and guiding fishery production.

6 DATA AVAILABILITY STATEMENT

Remote sensing data supporting the results of this study are available from the International Pacific Research Center (IPRC)/Asia Pacific Data Research Center (APDRC) and the corresponding open web site, detailed at Data Source. The fishery data in this paper are obtained from the publication 'Statistics of Marine and Coastal Resources' issued by the Indonesian National Bureau of Statistics.

7 ACKNOWLEDGMENT

The author would like to thank the scientific research institutions mentioned in "Data sources" for providing data support to us, and would like to thank the editors and reviewers for their efforts.