The TS data are available from the best-track typhoon dataset from Typhoon Online (www.typhoon.org.cn) (Ying et al., 2014; Lu et al., 2021), which includes the location and intensity of tropical cyclones in the northwest Pacific Ocean including the SCS every 3 to 6 h since 1949. The data include the MSW and the longitude/latitude of the center of the TS every 3 h.

We used Advanced Scatterometer (ASCAT) 6 h sea surface wind (SSW) data provided by the Remote Sensing System (www.remss.com/) at a spatial resolution of (0.25°×0.25°). We obtained daily precipitation data from the Tropical Rainfall Measuring Mission (TRMM) of the National Aeronautics and Space Administration (NASA) (https://gpm.nasa.gov/data) at a spatial resolution of 0.10°.

The data of photosynthetically active radiation (PAR) derived from four different sensors of Sea WiFS, MODIS, MERIS, and VIIRS with GSM Model (Maritorena and Siegel, 2005; Maritorena et al., 2010) are generated from GlobColour database provided by HERMES. The merged daily products of PAR with a spatial resolution of 0.04°× 0.04° are obtained for the period from July 26 to August 16.

Four buoys equipped with water quality sensors were deployed in the northern coastal areas of the Beibu Gulf from July 26 to August 16, 2019 ( Figure 1 ), respectively at the stations of S1-S4. The Chl-a concentration (i.e., fluorescence), temperature, salinity, and turbidity were continuously measured at a depth of 0.5 m below the sea surface using a YSI 6600V2 multiparameter water quality instrument (YSI Inc., Yellow Springs, USA). The dissolved inorganic nitrogen (DIN) was measured at S3 using a Systea nutrient probe analyzer (NPA; Systea S.p.A., Anagni, Italy). The above data were sampled every 30 min. After the observations were complete, the missing or invalid data were linearly interpolated to obtain a complete time series.

In situ Chl-a concentration derived from YSI 6600V2 multi-parameter water quality instrument (YSI-derived Chl-a) was measured every 30 min and first processed into daily Chl-a through 24-hr average (i.e., 0:00-24:00). The sum of Chl-a for 7 days (Chl-a _(sum)) was calculated:

where the Chl-a _(sum) was the sum of Chl-a concentration for 7 days (mg·m^-3) and n was the number of days in a week (n=7).

All the above-mentioned data analysis methods were performed in the software packages Origin, Ocean Data View, ArcGIS, and MATLAB.

The in situ observational data showed that the SST in the study region was relatively uniform high before the TS ( Figure 2A ). There was a lower SST patch near S3 during the TS ( Figure 2A ). One week after TS, there was a low nearshore SST and a high offshore in the study region. The SST increased slightly at S3 and S4, and continued to reduce at S1 and S2 1 week after TS. The SST recovered to roughly to the pre-TS SST (about 31°C-32°C) ( Figure 2A ) roughly 2 weeks later. Precipitation in the study region was significantly increased (>100 mm·day^-1) during the TS ( Figure 2B ), which may have increased the input of freshwater, leading possibly to changes in the SST, salinity, and turbidity.

The TS moved into the study region on August 1, 21:00, and left the study region on August 3, 02:00 ( Figure 1 ). The SSW time series based on the regional average for the study region (20.5-22.5° N, 107.5-110.0° E) ( Figure 3 ) showed that the SSW was < 6.0 m·s^-1 from July 26 to July 30, with a maximum SSW of about 11.5 m·s^-1 on August 1. The SSW quickly decreased after the TS had transited ( Figure 3 ).

The SST at the four stations decreased significantly, with a maximum reduction of 2-4°C 1-2 days after the passage of the TS on August 3-4 ( Figure 4A ). The decrease in the SST (< 30°C) lasted for about 6 days. The SST increased slowly and only returned to the pre-TS level on August 11. The smaller standard deviation for the four stations indicated a concentrated distribution of the SST data ( Table 1 ), which meant less SST variation at S4.

The time series of the salinity ( Figure 4B ) showed that the salinity of each station began to decrease after the TS had passed. S1 located in the landfall area of the TS was seriously affected by rainfall, and the salinity decreased by 13.5 1 day later. The salinity at S3, which was affected only by the River Nanliu, decreased by about 6.5 after the passage of TS. A low salinity appeared at S2 at the intersection of two rivers, which decreased by 8.5 2 days after the TS had passed. There was only a small, slow decrease in salinity at S4. The salinity decreased significantly at S1 and S2, resulting in large variations in their standard deviations (especially S2) ( Table 1 ).

Although S1 was located in the landfall region of the TS, the maximum turbidity here after the TS ( Figure 4C ) was lower than that at S2 and S3 and recovered to the pre-TS levels only within 3 days. The turbidity of S2 was as high as 190 NTU on August 4, about nine times the pre-TS values ( Figure 4C ), which may be related to the greater influence of increasing estuarine runoff from the Maoling and Qinjiang rivers. It took 10 days for the turbidity of S2 to recover to pre-TS levels ( Figure 4C ). Similarly, the maximum turbidity of S3, which was affected also by one river (i.e., Nanliu River), reached up to 120 NTU (about six times the pre-TS values) ( Figure 4C ). The turbidity of S3 recovered to pre-TS levels within 5 days, faster than that of S2 ( Figure 4C ). Due to the offshore location of S4, a weaker increase in turbidity was recorded with the low maximum value of 25.8 NTU during the TS (i.e., about two to three times the pre-TS values), and quickly recovered to pre-TS levels within 2 days ( Figure 4C ). Tabulated data illustrated that S2 and S3 had high turbidity and large amplitude, while relatively lower for S1 and S4 ( Table 1 ).

Our in situ data showed that the Chl-a concentration increased significantly in 1 to 2 weeks after the TS at all the stations except S2 ( Figure 5 ). There was no obvious change in the Chl-a _(sum) at S2 before and after the TS ( Figure 5A ). One week after the TS, there was a significant increase in the Chl-a _(sum) concentration of about 60% (26.34 mg·m^-3) at S1, 68% (25.62 mg·m^-3) at S3, and 69% (14.86 mg·m^-3) at S4; the changes were smaller at S2 ( Figure 5A ). Two weeks after TS, the Chl-a _(sum) concentration at S3 and S4 had decreased to roughly the levels 1 week before the TS ( Figure 5A ). At S1, the Chl-a _(sum) concentration 2 weeks after the TS was about two times that 1 week before the TS ( Figure 5A ). The time series ( Figure 5B ) shows that the average Chl-a concentration began to increase about 3 days after the TS had transited, reaching > 10 mg·m^-3 at S1, S3, and S4, especially highest at S1. The larger standard deviations of Chl-a concentration at S1, S3, and S4 indicate that the Chl-a concentration at the stations was more variable compared with S2 ( Table 1 ).

Salinity was measured at all stations, while nutrient concentration (i.e., DIN) was measured only at S3. A simple correlation analysis between DIN and salinity at S3 was made to assess the possibility of changes in nutrient concentration at all stations estimated by salinity values, to a certain degree. The scatter graph ( Figure 6 ) for DIN and salinity exhibited a sensible trend. A correlation analysis indicated that DIN was significantly negatively correlated with salinity (R² = 0.7528, p < 0.01) ( Figure 6 ).

Monitoring of Chl-a by means of traditional in-situ sampling has been difficult to be carried out under extreme weather events such as tropical storms. In situ fluorescence measurements can realize daily monitoring, hourly monitoring, and even continuous monitoring, which greatly improves the integrity and timeliness of monitoring data and helps to grasp the overall situation of the water body. Moreover, Chen and Zhao (2021) found the good relationship (R² = 0.7407, p<0.01) existed between Chl-a measured in the lab and YSI-derived Chl-a in coastal regions of the northwestern South China Sea, indicating that YSI-derived fluorescence can be used to estimate Chl-a.

There were significant changes in the SST, salinity, and turbidity ( Figure 4 ) in the study region as a result of mixing following the passage of the TS. These changes will inevitably affect the growth of phytoplankton in the water column and, in turn, the Chl-a concentration. After the TS passage, the overall Chl-a concentration at S1, S3, and S4 showed an increasing trend ( Figure 5B ), roughly consistent with earlier studies (Wetz and Paerl, 2008; Zhao et al., 2017; Liu et al., 2020).

A negative correlation between total nitrogen/total phosphorus and salinity was generally assumed in coastal regions (Lane et al., 2002; Schaeffer et al., 2012). Moreover, our study confirmed also that the high correlation (R² = 0.7528, p < 0.01) was presented between DIN and salinity at S3, suggesting that the salinity can be used to imply changes in nutrients level ( Figure 6 ) in the study. From the time series of salinity ( Figure 4B ), we knew that the nutrients at S1, S2, and S3 were full, but limited at S4. The salinity of S1, located in the landfall region of the TS ( Figure 1B ) and near the coast, decreased significantly by 13.5 in 1 day during the TS ( Figure 4B ), implying large amounts of nutrient-rich water input. The rapid (3 days) recovery in the turbidity to pre-TS levels resulted in good light conditions at S1 ( Figure 4B ), providing the conditions required for phytoplankton growth. Therefore, the two favorable conditions of the significant increase of nutrients and rapidly reduced turbidity after TS caused probably greater increase of Chl-a concentration and longer lasting time of high Chl-a concentration at S1 than those at S2, S3, and S4 ( Figure 5B ) after the passage of the TS.

S2 was located at the junction of two rivers ( Figure 1B ). The lowest salinity observed at S2 during the observational period suggested that there were large amounts of nutrient-rich freshwater input and nutrients were always available during the whole period before and after the TS ( Figure 4B ). However, the highest turbidity and slowest recovery time (10 days) limited the amount of light available at S2. There was no obvious increase in Chl-a concentration ( Figure 5 ) despite the nutrient-rich freshwater.

S3 was located in the Lianzhou Bay and was affected by the inputs from the Nanliu River ( Figure 1B ). The salinity started to decrease on August 1 and this decrease lasted for 3 days ( Figure 4B ), probably suggesting a continuous supply of nutrients. With the high turbidity restoring rapidly (within 4 days) to low turbidity after TS ( Figure 4B ), the phytoplankton grew in large numbers, resulting in an increase in Chl-a concentration.

S4 was located offshore with less influence of nutrient-rich freshwater input from rivers ( Figure 1 ). Therefore, the turbidity was lowest and the turbidity recovery time (2 days) was most rapidly at S4 ( Figure 4C ), reflecting the high availability of light in the region. The salinity at S4 was always the highest among the four stations and the decrease in salinity at S4 was less significant compared with other stations even during the passage of the TS. These suggested less of an increase of nutrient-rich freshwater at S4 after the TS and the supply of nutrients at S4 may still be the key to limiting the growth of phytoplankton. The temperature decrease ( Figure 4A ) under the influence of TS was the least significant at S4, which meant finite enhancing mixing implied after TS. The above conditions would have caused the increase in Chl-a concentration at S4 to not be the most significant ( Figure 5 ).

Previous research using remote sensing data had shown that the heavy rainfall caused by tropical cyclones increases generally the runoff from rivers in nearshore regions. This can lead to the transport of large amounts of nutrients, colored dissolved organic matter, and suspended sediments to coastal areas, contributing to the increase in Chl-a concentration (Chen et al., 2003; Zhao et al., 2009; Zhao et al., 2013). A fundamental limitation of remote sensing data is the inability to produce nearshore reliable measurements. Due to this limitation, our understanding of the response of Chl-a concentration in coastal areas to the passage of a TS is incomplete and needs to rely on the availability of in situ data. In our study, the results of in situ data showed that the growth of phytoplankton after the TS was not only affected by the nutrients brought by the increase of runoff but also related to the turbidity associated with light penetration.

Turbidity was negatively correlated with PAR (R² = 0.2284, p<0.01) ( Figure 7 ) in our study, which meant that the greater the turbidity is roughly coincided with the lower the PAR in the study. Dou et al. (2019) also found that the light attenuation coefficient decreases when the water transparency increases. While turbidity was an important restrictive factor affecting the transparency of water, which can be verified by the logarithmic fitting equation based on the relationship between suspended solids and transparency (Zhang et al., 2003). Therefore, in our study, the rapid recovery of the turbidity after TS would mean more favorable light conditions, and vice versa, the slow recovery of the turbidity would limit light. With the decrease of turbidity (increase of transparency), PAR increased, resulting in the increase of phytoplankton photosynthesis and higher Chl-a concentration. The recovering time of turbidity to pre-TS levels was 3 days at S1, 10 days at S2, 5 days at S3, and 2 days at S4 (recovery time: S2>S3>S1>S4) ( Figure 4B ), respectively, suggesting the favorable light conditions soon after TS at S1, S3, and S4, as well as the bad light conditions at S2. According to the conditions of salinity ( Figure 4B ), nutrients were always available during the whole period at S1, S2, and S3, but limited at S4 (nutrients input: S1>S2>S3>S4). The intensity of the phytoplankton blooms was the strongest at S1, followed by S3, then S4 (Chl-a: S1>S3>S4); the increase in Chl-a concentration was less significant at S2 ( Figure 5 ). These may imply that nutrients had a more important role in the growth of phytoplankton at S1, S3, and S4 along with the favorable light conditions suggested by the turbidity rapidly recovering to pre-TS levels. Therefore, the most signifcant increase of nutrients along with favorable light conditions had probably induced the highest increase of Chl-a concentration at S1 after TS ( Figure 5 ). Hagy et al. (2006) also reported that the input of freshwater from Hurricane Ivan led to one similar increase in the biomass of phytoplankton for several weeks in Pensacola Bay after the event.

In turbid coastal environments, light availability is usually the main driver of phytoplankton growth (Grobbelaar, 1990; Alpine and Cloern, 1992; Cloern, 1996; Kocum et al., 2002). The slow recovery of turbidity to pre-TS levels at S2 and the highest turbidity after TS suggested that the unfavorable light conditions into the water column, which was an important elementfor about 10 days after TS. Heavy rainfall during the TS ( Figure 2B ) led probably to large amounts of suspended particulate matter in the input from the Maoling and Qinjiang rivers, which may have reduced the transparency of the water at S2. Ma and Zhao (2021) have also found that, under normal circumstances, the Chl-a concentration in the nutrient-rich Pearl River estuary is low as a result of the high turbidity. Hence, the lack of available light limited the growth of phytoplankton in the water column at S2, although there were large amounts of nutrient-rich freshwater input implied by the salinity decreasing after TS. Moreover, the strong dilution from tropical rainfall during the storm and the mixing effects of the river input with high suspended matter limited also the accumulation of phytoplankton (Song et al., 2011), leading to a similar pattern of Chl-a concentration at S2 ( Figure 5 ) before and after the TS. In other words, the input from rivers can change the hydrodynamic conditions to be unsuitable for phytoplankton growth, meaning that it took longer for levels to recover to those before the TS. Fu et al. (2009) showed that Chl-a concentration near the Pearl River estuary decreased by about 35% as a result of the high content of suspended matter after the passage of a typhoon. In these examples, light associated with the turbidity was the most important factor for the increase in Chl-a concentration at S2 under the conditions of sufficient nutrients, but a lack of light to promote phytoplankton growth.

A previous study indicated that the most significant response of the ocean to the TS was a decrease in the sea temperature (Yang and Tang, 2010). The data collected for this research indicated that the whole study area had cooled down after the passage of the TS ( Figure 2A ). The decrease in the SST (< 30 °C) caused by the TS lasted for about 6 days, but the level of the decrease in the SST was different at all four stations ( Figure 4A ). The decrease in temperature indicated the degree of mixing of the water column. The mixing induced by the TS forced the redistribution of surface stratification, which caused the decrease of SST and entrainment of nutrients into the upper layer of oceans (Wang et al., 1998; Shao et al., 2015). The cooling of S1 by about 4.3°C and S2 by about 5.0°C ( Figure 4A ), and the decrease in the SST lasted for 1 week, which meant that they were fully mixed. During the TS, on August 2, although the cooling of S3 was the fastest, its temperature quickly recovered 1 week later ( Figure 2A ). In contrast, the degree of mixing of S3 was not as significant as that of S1 and S2. S4 had slow cooling and fast recovery, and the mixing of this station was weak. The less of an increase of Chl-a concentration at S4 may be related to weak mixing of water column, to a certain degree.