Daily Variation of Chlorophyll-A Concentration Increased by Typhoon Activity

Abstract

The chlorophyll-a (Chla) concentration product of the Himawari-8 geostationary meteorological satellite is used to show the temporal variation of Chla owing to the passage of typhoons, namely, tropical cyclones in the western North Pacific Ocean. The daily Chla variation shows that Chla usually increases along the paths of typhoons, whereas the same observations are almost impossible when using the data of polar-orbiting satellites as shown in previous studies. This is because the temporal resolution of Himawari-8 is ten times more than that of polar-orbiting satellites, and the daily Chla distribution contains a few disturbances attributed to clouds after compositing cloud-free data. Chla usually increases on the day of typhoon arrival, but mostly, the ratio of Chla increased by a typhoon to the background Chla, ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{HIMA}}$, is less than 2. Only a few typhoons considerably increased Chla. As a whole, ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{HIMA}}$ is proportional to the maximum 10-min sustained wind speed up to 85 knots (44 m s⁻¹), namely, 0.01 mg m⁻³ knot⁻¹ (0.019 mg s m⁻⁴). However, there is no clear dependence between Chla and the wind speed in seas with higher Chla, such as the South China Sea. The result that typhoons are usually cultivating the ocean is important for studies of primary ocean productivity and carbon flux between the atmosphere and ocean.

1. Introduction

As the primary product of the ocean, phytoplankton is important for ecosystems [1] and carbon cycles [2]. Typhoons, namely tropical cyclones in the North Pacific Ocean, sometimes induce phytoplankton blooms because of upwelling and vertical mixing. For example, Shih et al. [3] performed ship-borne measurements of chlorophyll-a (Chla) concentrations and particulate organic carbon flux (POC) a week before and 4–7 days after Typhoon Jangmi passed. They showed that Chla and POC were enhanced after the typhoon passed. However, because in situ measurements around a typhoon are dangerous, following research used space-borne data to estimate Chla. Lin et al. [4] showed that typhoon Kai-Tak increased Chla in the South China Sea to a level that was 30 times greater than that before Kai-Tak passed. Chen and Tang [5] showed that the category-1 typhoon Linfa induced an eddy feature Chla bloom that lasted for 17 days. Lin [6] showed that two of 11 typhoons in 2003 increased Chla, although not the super-typhoon Maemi, whose maximum 1-min sustained wind speed was 150 knots (77 m s⁻¹). Those studies used observational data from instruments on polar-orbiting satellites, such as the Moderate Resolution Imaging Spectroradiometer (MODIS) and the Sea-Viewing Wide Field-of-View Sensor (SeaWiFS), to measure the Chla increase due to typhoons. However, because polar-orbiting satellites have relatively low temporal resolution, namely once every few days, Lin [6] used weekly Chla data, which are not suitable for measuring the daily variation of Chla distributions.

Launched in 2010, the Geostationary Ocean Color Imager (GOCI) is the first geostationary satellite that can measure ocean colors [7]. Because the temporal resolution of GOCI is 1 h, He et al. [8], for example, were able to use GOCI to show that a typhoon in a coastal area increased the total suspended particle matter for a duration of less than a day. GOCI retrieves Chla with less cloud disturbance than MODIS and SeaWiFS do. This is because even if Chla retrieval at a given moment is impossible because of cloudiness, the daily Chla can still be retrieved at another time on the same day when there is no cloud. Therefore, the daily Chla retrieved by GOCI suffers from less lack of data than that obtained by polar-orbiting satellites when compositing with cloud-free data. However, GOCI is not suitable for measuring the increase in Chla by all typhoons because the observational area of GOCI is the ocean around the Korean Peninsula, and the southern limit of GOCI measurements is close to 25°N (e.g., Figure 1 of Ryu et al. [7]).

Launched in 2014, the Himawari-8 geostationary meteorological satellite can retrieve Chla since July 2015 [9]. Its temporal resolutions are 10 min in the hemisphere, whose center is 140.7°E, and 2.5 min around Japan. That is, Himawari-8, unlike GOCI, can be used to measure the tropical regions where many typhoons occurred. Himawari-8 also has six times (24 times) more chances to measure cloud-free data than GOCI in the subtropics (around Japan). Therefore, Himawari-8 is more suitable than GOCI for measuring the increase in the Chla owing to typhoons. Shen et al. [10] used GOCI and Himawari-8 to retrieve Chla and presented a case study of the category-4 typhoon Malakas, which increased Chla and decreased sea surface temperature (SST) where it passed the northern sea of Taiwan.

This study focuses on retrieving the daily variation of Chla owing to typhoons, such as by applying statistical analysis to 4-year data to establish how the Chla increase is related to typhoon wind speed and ocean area. Since the daily variation of Chla was hardly measured by sensors on polar-orbiting satellites, the relationship between a typhoon and a phytoplankton bloom is not well known. This study also examines the daily variation of SST because the SST retrieved by Himawari-8 has better temporal and spatial resolutions than that retrieved by polar-orbiting microwave imagers such as the Advanced Microwave Scanning Radiometer 2 (AMSR-2) [11].

2. Data

The typhoon parameters that are used are those estimated by the Japan Meteorological Agency (JMA), including the lowest pressure, the maximum 10-min sustained wind speed, and the latitude and longitude of the center of each typhoon. These parameters are reported usually every 6 h (3, 9, 15, and 21 UTC) but more often when a typhoon is near Japan. The parameters at 6 UTC, 15 in local time at a longitude of 135°E, are used because we seek to analyze daytime Chla. The wind speed of typhoons of the Japan Meteorological Agency (JMA) use knot; hence, wind speed has been provided in the units of knots and m s⁻¹ in this study.

Figure 1 shows the occurrence numbers and wind speed distribution of typhoons between August 2015 and July 2019. In those four years, 103 typhoons occurred, the majority (87, 84%) between July and November (Figure 1a). The greatest maximum 10-min sustained wind speed of the typhoons was 120 knots (62 m s⁻¹), and the mode of the 10-min maximum wind speed occurrence was 35–40 knots (18–21 m s⁻¹; Figure 1b). Typhoons were present on 519 days in total (the sum of “number of days” in Figure 1b), thus their average lifetime was 5 days. Typhoons tend to exist at longitudes of 145° or less and at latitudes between 10° and 35° (not shown).

Two types of Chla data sets are used. One is Chla retrieved by the Himawari-8 geostationary meteorological satellite [12], provided by the P-tree system of Japan Aerospace Exploration Agency (JAXA), and another is Chla retrieved by the MODIS on board the Aqua satellite in a near-polar low Earth orbit, provided by the Ocean Biology Processing Group of the Ocean Ecology Laboratory of the NASA Goddard Space Flight Center [13]. Himawari-8 and MODIS retrieve Chla by using the empirical function of the reflectivities of the visible bands, thus Chla is retrievable in daytime on clear days. Murakami [12] has shown that Chla of Himawari-8 becomes noisy at 35° or higher latitude of the winter hemisphere owing to the long path of the solar light. This noise does not affect this research because most typhoons occur between July and November (Figure 1a) at 35° or lower latitude.

The temporal resolutions of Himawari-8 and MODIS are 10 min and once every few days, respectively. The horizontal distribution of Chla retrieved by Himawari-8 has less lack of data than that by MODIS in the western North Pacific, because Himawari-8 has ten times more chances to retrieve Chla than MODIS does. The horizontal resolution of the Chla products of Himawari-8 and MODIS are 0.05° (approx. 5 km) and approximately 0.04° (approx. 4 km), respectively.

This study used the daily Chla for our analysis and the monthly Chla to remove the striping noise of the Himawari-8 Chla as described in the Appendix A. Chla^HIMA denotes the Himawari-8 Chla with noise removed (see Appendix A) and Chla^MODIS denotes the MODIS Chla.

This study also used the daily data of skin and subskin SSTs retrieved respectively by Himawari-8 (SST^HIMA; [14]) and AMSR-2 (SST^AMSR2; [11]) on board the Global Change Observation Mission-W1 (GCOM-W1) that is in a sun-synchronous orbit. SST^HIMA and SST^AMSR2 are provided by the P-Tree system and the G-Portal system of JAXA, respectively. SST^HIMA and SST^AMSR2 are retrieved from the thermal emission at depths of less than 500 μm and 1 mm, respectively. For a conceptual diagram of skin and subskin SSTs, see Figure 1 of Donlon et al. [15]. The temporal and horizontal resolutions of AMSR-2 are once every few days and 35 × 62 km at a frequency of 6.9 GHz, respectively, where the vertical polarization of 6.9 GHz is used for SST retrieval. The horizontal resolutions of the SST^HIMA and SST^AMSR2 products that I used are 0.02° (approx. 2 km) and 0.1° (approx. 10 km), respectively. SST^HIMA is retrievable in the daytime on clear days, whereas SST^AMSR2 is retrievable even at cloudy nighttime. SST^AMSR2 is not retrievable in rainfall.

To compare Chla and SST at the same place, the rates of Chla increase (Equation (1a,b)) and SST decrease (Equation (2a,b)) were introduced, namely:

$${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{HIMA}}={\mathrm{Chla}}^{\mathrm{HIMA}}/{\mathrm{Chla}}_{3\mathrm{d}\mathrm{before}}^{\mathrm{HIMA}}$$

(1a)

$${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{MODIS}}={\mathrm{Chla}}^{\mathrm{MODIS}}/{\mathrm{Chla}}_{3\mathrm{d}\mathrm{before}}^{\mathrm{MODIS}}$$

(1b)

$${\mathsf{\delta }\mathrm{SST}}^{\mathrm{HIMA}}={\mathrm{SST}}^{\mathrm{HIMA}}-{\mathrm{SST}}_{3\mathrm{d}\mathrm{before}}^{\mathrm{HIMA}}$$

(2a)

$${\mathsf{\delta }\mathrm{SST}}^{\mathrm{AMSR}2}={\mathrm{SST}}^{\mathrm{AMSR}2}-{\mathrm{SST}}_{3\mathrm{d}\mathrm{before}}^{\mathrm{AMSR}2}$$

(2b)

The parameters within a bracket (e.g., 〈Chla^HIMA〉) denote the medians of the parameters (e.g., Chla^HIMA) within 100 km from the daily center at each point along the typhoon trajectory. This study uses medians instead of averages because a few outliers sometimes considerably change the average. The parameters with a subscript “3d before” (e.g., ${\mathrm{Chla}}_{3\mathrm{d}\mathrm{before}}^{\mathrm{HIMA}}$) are the same as the parameters without the subscript (e.g., ${\mathrm{Chla}}^{\mathrm{HIMA}}$) but for three days before. The parameters with a subscript “3d before” are assumed as the background values of Chla and SST.

3. Results

Figure 2 shows examples of the Chla^HIMA, Chla^MODIS, SST^HIMA, and SST^AMSR2 distributions of Typhoon Yutu, whose minimum pressure and maximum 10-min sustained wind speed on 24–30 October, 2018 were 915–985 hPa and 55–105 knots (26–54 m s⁻¹), respectively. Figure 2(a4–a7) show that Chla^HIMA increased after the typhoon passed at 18°N and 125°E–140°E. Figure 2(b7) also shows that Chla^MOIDIS increased at 18°N and 130°E; however, the Chla^MODIS time sequence is not clear as few data were retrieved because of cloudiness and lack of observations. Figure 2(c4–c7) show that SST^HIMA decreased along the path of the typhoon. SST^AMSR2 also decreased along the path, around 18°N and 130°E (Figure 2(d6,d7)). SST^HIMA is also clearer than SST^AMSR2 because the observation frequency and horizontal resolutions of AMSR-2 are sparser and coarser than those of Himawari-8. The typhoon in late October was chosen in Figure 2 to avoid unclear Chla and SST tendencies because some Chla and SST horizontal distributions from July to September have tracks of Chla increases and SST decreases induced by previous typhoons. The tendencies of the Chla increases and the SST decreases are almost common for all typhoons, but those on day 0 (the day of typhoon arrival) are sometimes unclear because of lack of data due to cloudiness. Figure 2 shows that Himawari-8 has the potential to retrieve daily variations of Chla and SST distributions.

The lowest pressure and the greatest maximum 10-min sustained wind speed were 900 hPa and 115 knots (59 m s⁻¹), respectively, around 15°N and 145°E at 18 UTC on 24–25 October 2018 (Figure 2(a1,a2)), although the increase in Chla^HIMA was not clearer than that at 18°N and 125°E–140°E as shown in Figure 2(a4–a7). This is because the background Chla around 15°N and 145°E is smaller than that at 18°N and 125°E–140°E, where the background Chla is retrieved as follows. For example, the center of the typhoon on October 27 was at 18.0°N and 133.3°E (Figure 2(a4)). The background Chla at the typhoon on October 27 was assumed to be Chla in the same area (18.0°N and 133.3°E) on October 24. The background Chla without cloud disturbance can be retrieve because Figure 2(a1) shows there was no cloud around 18.0°N and 133.3°E on October 24.

The Chla background and ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{HIMA}}$ are the greatest on 30 October, namely 0.18 mg m⁻³ and 8.4, respectively, near the coast of the Philippines along the track of Typhoon Yutu.

The wind speed dependence of ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{HIMA}}$, ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{MODIS}}$, δSST^HIMA and δSST^AMSR2 is analyzed using the daily data set of ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{HIMA}}$, ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{MODIS}}$, δSST^HIMA, δSST^AMSR2, longitude and latitude of the center of a typhoon, and the wind speed of a typhoon. Figure 3 shows the wind speed dependence of day 1 (i.e., one day after typhoon arrival) for all typhoons. The numbers of data for ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{HIMA}}$, ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{MODIS}}$, δSST^HIMA, and δSST^AMSR2 are 434, 50, 178, and 175, respectively.

Figure 3a shows that ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{HIMA}}$ is proportional to wind speed, namely 0.010 mg m⁻³ knot⁻¹ (0.019 mg s m⁻⁴), up to approximately 85 knots (44 m s⁻¹). ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{MODIS}}$ is also greater than 1, but the relationship between wind speed and ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{MODIS}}$ is not smooth, e.g., ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{MODIS}}$ has a gap at a wind speed of 60 knots (31 m s⁻¹) because of the small number of data. Stronger wind speed (above 85 knots, 44 m s⁻¹) also increases Chla, but it is likely to have less dependence on wind speed. This is consistent with the work of Lin [6], who showed that the super-typhoon Maemi (whose maximum 10-min sustained wind speed was estimated to be 105 knots (54 m s⁻¹) by the JMA) did not increase Chla. Five outliners of ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{HIMA}}$ were greater than 10; their weekly-averaged Chla may be observed by using polar-orbiting satellites.

Figure 3c,d show δSST^HIMA and δSST^MODIS on day 1. The decrease in SST^AMSR2 seems to be also proportional to wind speed up to 85 knots (44 m s⁻¹), whereas that of SST^HIMA does not depend on wind speed. δSST^AMSR2 seems to be also independent of wind speed above 85 knots (44 m s⁻¹).

Figure 4 shows the correlation coefficients between wind speed and ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{HIMA}}$ up to 85 knots (54 m s⁻¹) and background Chla^HIMA. The data are averaged for 100 km from the center of 20°N and 110°E–160°E. The correlation coefficients are around 0.8 at 130°E–160°E, while those in the west from 130°E, such as in the South China Sea and around coastal areas, are around 0.4. This suggests that the Chla increase does not depend clearly on the wind speed in seas with higher Chla.

Figure 5 shows the tangents of the regression lines for ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{HIMA}}$, ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{MODIS}}$, δSST^HIMA, and δSST^AMSR2 from day 0 to 14. Figure 5a shows that ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{HIMA}}$ and ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{MODIS}}$ have the same tendency, but ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{MODIS}}$ is not smooth because of small sample numbers. ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{MODIS}}$ and δSST^AMSR2 are not calculated for day 0 because of lack of data. The increase in Chla appeared just after day 0 and disappeared within two weeks.

Donlon et al. [15] showed that the difference between skin SST and bulk SST is around 0.17 °C when the wind speed is greater than 6 m s⁻¹. SST^HIMA and SST^AMSR2 should be the same just after a typhoon has passed. However, as shown in Figure 3c,d, SST^HIMA is higher than SST^AMSR2 at days 1 and 2. SST^HIMA starts to decrease from day 0. SST^HIMA reaches its minimum after six days from the typhoon’s arrival. Both decreases in SST also disappeared within two weeks.

Figure 3d and Figure 5b show that SST^AMSR2 decreases just after typhoon arrival. This is consistent with the results that ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{HIMA}}$ and ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{MODIS}}$ are the largest just after typhoon arrival (Figure 5a). In general, the amount of nutrients at the sea surface is smaller than that of below the thermocline because phytoplankton uses nutrients in the euphotic zone [16,17]. The increased Chla suggests that the sea surface water is transported from deeper (and colder) water, which contains more nutrients. However, the results indicating that the SST^HIMA is not decreased just after typhoon arrival (Figure 3c and Figure 5b) are inconsistent with the above results. Thus, SST^HIMA may have a warmer bias at days 0–4, possibly caused by different weather conditions. Namely, the SST^HIMA is retrieved only on clear days, whereas the SST^AMSR2 is retrieved on clear and cloudy days. The cloudier the weather, the larger the difference between SST^HIMA and SST^AMSR2. This difference should be examined in the future.

4. Conclusions

This study retrieved the empirical characteristics of Chla increase and SST decrease by using the daily Chla and SST products of Himawari-8 for 103 typhoons. The striping noise of the daily Chla of Himawari-8 was removed by using the monthly Chla of Himawari-8 and MODIS. Because the temporal resolution of Himawari-8 is greater than that of polar-orbiting satellites, the Chla and SST retrieved by Himawari-8 are clearer than those due to compositing cloud-free data. That is, the horizontal distribution of the daily Chla^HIMA (SST^HIMA) showed clearly that Chla^HIMA (resp. SST^HIMA) was increased (resp. decreased) along the typhoon path after the arrival of a typhoon.

The rate of increase in Chla retrieved by Himawari-8, namely ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{HIMA}}$, was enhanced when the wind speed of the typhoon became stronger. ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{HIMA}}$ was approximately proportional to wind speed up to approximately 85 knots (44 m s⁻¹), namely 0.010 mg m⁻³ knot⁻¹ (0.019 mg s m⁻⁴). This relationship was unclear for the Chla^MODIS because of fewer data. Chla was the largest at day 0, the day of typhoon arrival, and the increase in Chla disappeared within two weeks. There was likely to be no clear proportional expression between ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{HIMA}}$ and wind speed above 85 knots (54 m s⁻¹), although stronger wind would induce more upwelling. This may be consistent with a previous study of the super-typhoon Maemi, which did not increase the weekly-averaged Chla. Chla is usually increased by a typhoon, but mostly, ${\mathrm{R}}_{\mathrm{Chla}}^{\mathrm{HIMA}}$ is less than 2. Only a few typhoons appreciably increase Chla.

The decrease in SST^HIMA was little and not proportional to wind speed within a few days from typhoon arrival. However, SST^AMSR2 was the coldest just after typhoon arrival. SST^HIMA would have clear-sky bias because SST^HIMA was retrieved only on clear days whereas SST^AMSR2 was retrieved on both clear and cloudy days. SST^AMSR2 was also proportional to wind speed up to 85 knots (54 m s⁻¹). There is also no clear proportional relation between SST^AMSR2 and wind speed over 85 knots (54 m s⁻¹).

The above findings were not retrieved by previous polar-orbiting satellites and geostationary satellites. The next step of this work is to estimate the carbon flux by usual typhoons and super-typhoons, which will occur more often in the near future [18].