Open AccessArticle

Sensitivity Assessment on Satellite Remote Sensing Estimates of Primary Productivity in Shelf Seas

Xiaolong Zhao

^1,2,

Jianan Sun

³,

Qingjun Fu

^4,*,

Xiao Yan

⁴ and

Lei Lin

⁴

North China Sea Marine Forecasting Center, Ministry of Natural Resources of the People’s Republic of China, Qingdao 266061, China

Shandong Key Laboratory of Marine Ecological Environment and Disaster Prevention and Mitigation, Qingdao 266061, China

Ningbo Yonghuanyuan Environmental Engineering and Technology Co., Ltd., Ningbo 315000, China

⁴

College of Ocean Science and Engineering, Shandong University of Science and Technology, Qingdao 266590, China

Author to whom correspondence should be addressed.

J. Mar. Sci. Eng. 2024, 12(12), 2146; https://doi.org/10.3390/jmse12122146

Submission received: 28 October 2024 / Revised: 21 November 2024 / Accepted: 22 November 2024 / Published: 25 November 2024

(This article belongs to the Section Marine Environmental Science)

Download

Browse Figures

Versions Notes

Abstract

The vertically generalized production model (VGPM) is one of the most important methods for estimating marine net primary productivity (PP) using remote sensing. However, different data sources and parameterization schemes of the input variables for the VGPM can introduce uncertainties to the model results. This study compared the PP results from different data sources and parameterization schemes of three major input variables (i.e., chlorophyll-a concentration (

C_{o p t}

), euphotic depth (

Z_{e u}

), and maximum photosynthetic rate (

P_{o p t}^{B}

)) and evaluated the sensitivity of VGPM in the Yellow and Bohai Seas on the inputs. The results showed that the sensitivity in the annual mean PP was approximately 40%. Seasonally, the sensitivity was lowest in the spring (35%), highest in the winter (70%), and approximately 60% in the summer and autumn. Spatially, the sensitivity in nearshore water (water depth < 40 m) was more than 60% and around two times higher than that in deep water areas. Nevertheless, all VGPM results showed a decline trend in the PP from 2003 to 2020 in the Yellow and Bohai Seas. The influence of

P_{o p t}^{B}

and

C_{o p t}

was important for the magnitude of annual mean PP. The PP seasonal variation pattern was highly related to the parameterization scheme of

P_{o p t}^{B}

, whereas the spatial distribution was mostly sensitive to the data sources of

C_{o p t}

Keywords:

shelf sea; marine primary productivity; satellite remote sensing; chlorophyll-a

1. Introduction

Marine primary productivity is defined as the rate at which marine phytoplankton synthesize organic matter via the process of photosynthesis. Phytoplankton absorb carbon dioxide and release oxygen through photosynthesis, which is the driving force of the marine ecosystem cycle. Therefore, marine primary productivity plays an important role in the global carbon flux and carbon cycle [1,2]. Although the shelf seas account for less than 10% of the global oceans, they provide more than one-third of the marine primary productivity and thus play a crucial role in the global carbon cycle [3,4]. Therefore, it is important to quantify the primary productivity in shelf seas. At the same time, marine primary productivity is also an important parameter for the assessment and management of the abundance of marine biological resources and for measuring the status and quality of the marine ecological environment, which provide important reference values for the management and planning of offshore marine ecology and fisheries.

The traditional method for estimating marine primary productivity is ship survey sampling combined with field measurement methods, such as the ¹⁴C tracer method [5], chlorophyll fluorescence method, and black-and-white bottle oximetry [6,7]. However, field measurements are often costly, and they are insufficient for addressing the requirements of large-scale spatiotemporal data research due to the limited number of on-site observation stations. Compared to the traditional determination and estimation methods of primary productivity from onboard surveys, satellite remote sensing has the advantage of obtaining long time series and large-scale marine environmental parameters. Therefore, the calculation of marine primary productivity using satellite remote sensing has become a common and important method for monitoring and evaluating marine primary productivity.

Remote sensing inversion models of marine primary productivity can be divided into two types: empirical and physiological process models. Smith et al. [8] and Eppley et al. [9] obtained chlorophyll concentration data using remote sensing and established a statistical model of marine primary productivity to study marine primary productivity. However, this model had strong regional limitations and low accuracy; therefore, relevant parameters of phytoplankton physiological characteristics were added to the light–biological model of Platt [10] and Morel [11] to improve the accuracy in model estimation. Subsequently, an increasing number of improved primary productivity models have been proposed, such as the Bedfor Productivity Model (BPM), Laboratoire de Physique et Chimie Marines (LPCM), and Vertically Generalized Production Model (VGPM) [12,13,14]. The VGPM was proposed by Behrenfeld and Falkowski based on satellite data to estimate the vertical integrated net primary productivity in the euphotic layer (the primary productivity hereafter referred to as net primary productivity). All the input data variables for the VGPM can be obtained from the remote sensing data. This model has an excellent potential for application and is widely used in studies on marine primary productivity.

Considering the important contribution to the primary productivity of shelf seas, many studies have applied the VGPM and satellite remote sensing data to estimate and study the primary productivity of different shelf sea areas, such as the Bering Sea, Northeastern Arabian Sea, and China seas [15,16,17]. Data, such as chlorophyll concentration, euphotic depth, and maximum photosynthetic rate, have been needed for the VGPM. However, global sea surface chlorophyll products, which have been used in previous studies, were mainly developed for marine Case-I waters. In shelf seas, especially in some coastal areas with high turbidity, the accuracy of the products will be affected by suspended substances in water, which can lead to chlorophyll-a concentration errors of more than 100% [18]. The euphotic depth data estimated by traditional satellite remote sensing can also have large errors for shelf seas [19]. In addition, the maximum photosynthetic rate in the VGPM was obtained as a seventh-order polynomial of sea surface temperature. However, Carr et al. [20] and Yoon et al. [21] showed that the optimal parameterization schemes of the maximum photosynthetic rate in different sea areas could be significantly different. As the primary productivity obtained from VGPM is the product of several factors, the accumulation of the errors of different types of remote sensing data and parameterization schemes may lead to large deviations and uncertainty in the primary productivity estimates of the VGPM in shelf seas. However, the magnitude of the sensitivity to the data sources and parameterization schemes has not yet been quantitatively assessed. Therefore, in this study, sea surface chlorophyll concentration and euphotic depth data from different sources and the maximum photosynthetic rate obtained by different parameterization schemes were input into the VGPM to estimate the primary productivity of the Yellow and Bohai Seas (YBS). By comparing the differences in the VGPM results, the sensitivity of the VGPM estimates of primary productivity to the data sources and parameterization schemes was evaluated, and the influence of different factors on the results was analyzed. This paper serves as a reference for future applications of the VGPM in shelf seas.

2. Materials and Methods

2.1. Study Area

The YBS are continental shelf marginal seas in the Northwest Pacific Ocean, with an area of ~450,000 km² and an average water depth of ~40 m (Figure 1). The YBS exhibit high levels of primary productivity and significant deposition rates, which are critical to the carbon budget in China’s shelf seas [22]. Tan and Shi [17] estimated the primary productivity of the China shelf seas using the VGPM and found that the annual mean of primary productivity was approximately 536.47

mgC / (m^{2} \cdot d)

in the Southern Yellow Sea, 363.08

mgC / (m^{2} \cdot d)

in the Northern Yellow Sea, and 564.39

mgC / (m^{2} \cdot d)

in the Bohai Sea. In contrast, Yang [23], using the VGPM, estimated the annual mean primary productivity in the Yellow Sea to exceed 1000

mgC / (m^{2} \cdot d)

. Cong [24] also employed the VGPM alongside data from MODIS and SeaWiFS satellite, revealing that primary productivity in most regions of the Yellow Sea reached 500–700

mgC / (m^{2} \cdot d)

during the spring, peaking in the summer at 800–1000

mgC / (m^{2} \cdot d)

. However, Jia et al. [25] estimated the summer primary productivity in the Yellow Sea using the same model to be only 530

mgC / (m^{2} \cdot d)

. According to Ding [26], the VGPM results indicated that the maximum primary productivity in the YBS exceeded 2000

mgC / (m^{2} \cdot d)

during the summer, with a bimodal distribution observed in the Yellow Sea, characterized by two peaks in June and September. Additionally, Li et al. [27] reported that the primary productivity in the Bohai Sea was relatively low from January to April, averaging only 677

mgC / (m^{2} \cdot d)

in January, while the net primary productivity (NPP) reached 5265

mgC / (m^{2} \cdot d)

in August and 825

mgC / (m^{2} \cdot d)

in December. The above results suggest a strong uncertainty in the VGPM estimates of primary productivity in the YBS. This uncertainty could result from the different data sources and parameterization schemes used in different studies. Therefore, by comparing the VGPM results of different data sources and parameterization schemes, this study evaluated the sensitivity of the VGPM estimates of primary productivity in the YBS to different data sources and parameterization schemes.

2.2. VGPM

The equation of the VGPM for calculating the vertical integral primary productivity in the enphotic zone is as follows [14]:

P P_{e u} = 0.66125 \times P_{o p t}^{B} \times [\frac{E_{0}}{E_{0} + 4.1}] \times C_{o p t} \times Z_{e u} \times D_{i r r},

(1)

where

P P_{e u}

is the integrated net primary productivity in the euphotic zone (

mg / (m^{2} \cdot d)

P_{o p t}^{B}

is the maximum photosynthetic rate within a water column (

mg / (mg \cdot d)

Z_{e u}

is the euphotic zone depth (

m

E_{0}

is the sea surface photosynthetically active radiation (PAR) (

mol / (m^{2} \cdot d)

), and

C_{o p t}

is the corresponding chlorophyll concentration at

P_{o p t}^{B}

(

mg / m^{3}

), which can be conveniently replaced by the sea surface chlorophyll concentration (

C_{s a t}

) multiplied by 0.9899 [14].

D_{i r r}

is the photoperiod (

h

), which can be calculated from the latitude of the study area and solar declination.

The parameterization scheme of

P_{o p t}^{B}

given by Behrenfeld and Falkowski [14], according to the empirical relationship between

P_{o p t}^{B}

and the monthly mean sea surface temperature (SST), is as follows (hereafter,

P_{o p t}^{B}_B

P_{o p t}^{B}_B = \{\begin{array}{l} 1.13, T \leq - 1.0; \\ 4.00, T \geq 28.5; \\ P_{o p t}^{B'}, - 1.0 < T < 28.5 \end{array},

(2)

where

T

is the SST (

℃

), and the expression of

P_{o p t}^{B'}

is expressed as follows:

P_{o p t}^{B'} = 1.2596 + 2.749 \times 10^{- 1} T + 6.17 \times 10^{- 2} T^{2} - 2.05 \times 10^{- 2} T^{3} + 2.462 \times 10^{- 3} T^{4} - 1.348 \times 10^{- 4} T^{5} + 3.4132 \times 10^{- 6} T^{6} - 3.27 \times 10^{- 8} T^{7} .

(3)

Yoon et al. [21] proposed a parameterization scheme for

P_{o p t}^{B}

for the Yellow Sea based on field observation data of the Yellow Sea, and the expression is as follows (hereafter,

P_{o p t}^{B}_Y

P_{o p t}^{B}_Y = - 0.013 \times S S T^{2} + 0.27 \times S S T + 3 .

(4)

2.3. Data Sources

This study focused on variations in the primary productivity in the YBS from 2003 to 2020. The monthly average satellite data from 2003 to 2020 were used in the VGPM, including sea surface PAR, SST, chlorophyll-a concentration, diffuse attenuation coefficient (490 band) (Kd490), and Zeu data. Monthly average PAR data were obtained from the MODIS/Aqua and MODIS/Terra satellites of the National Space Agency (NASA) (https://oceandata.sci.gsfc.nasa.gov/, accessed on 15 June 2022). SST data were obtained from the monthly mean SST product of MODIS (https://oceandata.sci.gsfc.nasa.gov/, accessed on 15 June 2021). Different maximum photosynthesis rates could be obtained by substituting SST data into the parameterized schemes

P_{o p t}^{B}_B

and

P_{o p t}^{B}_Y

, respectively. Sea surface chlorophyll-a concentration data from MODIS modified by the GAM developed by Wang et al. [18] (marked as

C_{o p t}_M O D I S

) and a multisource satellite data product created by the European Space Agency (ESA) (https://www.oceancolour.org/, accessed on 15 June 2022) (marked as

C_{o p t}_E S A

) were used. Two types of euphotic depth data were used: one was the fusion of three kinds of satellite data products from VIIRS/SNPP, MODIS/Terra, and MODIS/Aqua on NASA (https://oceancolor.gsfc.nasa.gov/, accessed on 15 June 2022) adopting the IOP inversion algorithm developed by Lee et al. [28] (marked as

Z_{e u}_I O P

), and the other was calculated using theKd490 of seawater (marked as

Z_{e u}_K d

). The calculation formula is as follows [29]:

Z_{e u}_K d = 2 \ln \frac{10}{K_{d}} = \frac{4.605}{K_{d 490}},

(5)

where Kd490 is the monthly average data from NASA MODIS/Aqua and MODIS/Terra satellites from 2003 to 2020 (https://oceandata.sci.gsfc.nasa.gov/). The spatial resolution of the data is ~4 km. The spatiotemporal missing values were reconstructed using the DINEOF interpolation method.

2.4. Experimental Design

To evaluate the sensitivity of the VGPM estimates of primary productivity and study the influence of different data sources and parameterization schemes on the VGPM results,

Z_{e u}

and

C_{o p t}

obtained from two different data sources and

P_{o p t}^{B}

obtained from two different parameterization schemes, were substituted into the VGPM, and the estimation results of the eight experiments were obtained and compared in this study. The data and parameterization schemes for the eight cases are listed in Table 1.

2.5. Data Analysis

In this study, the standard deviation and coefficient of variation (CV) of the VGPM results of the eight experiments were computed to quantify the sensitivity of the VGPM to different sources of data and parameterization schemes. The coefficient of variation is the ratio of the standard deviation to the mean value, which represents the statistics of the degree of variation of a variable.

Empirical mode decomposition (EMD) is a method to deal with non-stationary signals proposed by Huang et al. [30], which essence is to smooth a signal, decompose the fluctuation or trend of different scales in signals, and produce a series of modes and residuals with different characteristic scales, and its residuals can be used to characterize the long-term trend of the data. The EMD method was used in this study to decompose the time series of primary productivity of the eight experiments, and EMD residuals were obtained to characterize the long-term trend of primary productivity in the YBS for different experiments.

3. Results

3.1. Seasonal Variations

The primary productivity of the VGPM for the eight experiments was significantly different in magnitude (Table 2). The annual mean of the primary productivity estimated by the eight experiments was quite different, with an average value of 1015.4

mgC / (m^{2} \cdot d)

, a standard deviation of 406.8

mgC / (m^{2} \cdot d),

and a CV of 40%, indicating that different data sources and parameterization schemes brought 40% sensitivity to the estimation of primary productivity in the YBS. Table 2 shows that the average annual mean of Experiments 1, 2, and 3 was significantly lower than the average of the eight experiments, while that of Experiment 8 was the highest, reaching 1778.3

mgC / (m^{2} \cdot d)

. Meanwhile, the values of Experiments 1–4 were lower than those of Experiments 5–8 on the whole, which could be related to the different

P_{o p t}^{B}

schemes used in the VGPM. Under the conditions of using the same

P_{o p t}^{B}

parameterization scheme, the annual mean of the primary productivity of Experiments 4 and 8 was significantly higher than that of other experiments in Experiments 1–4 and Experiments 5–8, respectively, which may be due to the combined effect of

C_{o p t}

and

Z_{e u}

. In addition, based on the results, the average relative differences in annual primary productivity caused by the

P_{o p t}^{B}

C_{o p t}

, and

Z_{e u}

were 38%, 35%, and 29%, respectively.

The primary productivity of the eight experiments exhibited different seasonal variation characteristics (Figure 2). The seasonal variations in the primary productivity of Experiments 1–4 were similar, showing bimodal distributions with two peaks in spring (May) and autumn (from October to December). In addition, the primary productivity of Experiments 1–4 was relatively low in summer and showed little difference in magnitude. The seasonal variations in Experiments 5–8 were unimodal, with a relatively high primary productivity from June to October. The rank of the magnitude of the primary productivity in different seasons of the eight experiments was also different, being spring > summer > autumn > winter for Experiments 1–2, spring > winter > autumn > summer for Experiment 3, winter > spring > autumn > summer for Experiment 4, and summer > spring > autumn > winter for Experiments 5–8.

The average of the eight experiments showed that the primary productivity of the YBS was higher in the spring and summer, with an average of 1260.3 and 1223.8

mgC / (m^{2} \cdot d)

, respectively, and the lowest was in the winter, with an average of 684.3

mgC / (m^{2} \cdot d)

. The monthly mean standard deviations of the eight experiments were higher than 550

mgC / (m^{2} \cdot d)

from July to October, especially in July and August, up to 700

mgC / (m^{2} \cdot d)

, while those in April and May were lower than 400

mgC / (m^{2} \cdot d)

(Figure 3a). In addition, the coefficients of variation (CV) from April to June were lower than 40%, whereas those in January, February, August, and December were higher than 60%. The coefficients of variation were approximately 35% in the spring, 70% in the winter, and 60% in other seasons (Figure 3b). The CV results suggest that the sensitivity was relatively small in the spring and high in the summer, autumn, and winter when using different sources of data and parameterization schemes in the model to estimate the primary productivity in the YBS.

3.2. Interannual Variations

The primary productivity of the YBS in the eight experiments showed different interannual variation patterns (Figure 4). The primary productivity of Experiments 1–4 was relatively low in 2004, 2010, and 2016, but it was high in 2007, 2013, and 2019; the productivity for Experiments 5 and 6 was relatively low in 2007, 2013, and 2019 but high in 2004, 2009, and 2016, which is opposite to the interannual variation patterns of Experiments 1–4; the productivity for Experiments 7 and 8 was relatively low in 2005, 2011, and 2017 but high in 2009 and 2014. Combined with Table 1, it can be speculated that the interannual variations were significantly affected by

P_{o p t}^{B}

, followed by

C_{o p t}

The results of the EMD analysis show that the primary productivity estimation results of the eight experiments display a decreasing trend; however, the rates of decline and corresponding years of change are different in the eight experiments (Figure 5). The average decline rates of Experiments 1, 3, and 7 were 4.71

mgC / (m^{2} \cdot d) / year

, 0.25

mgC / (m^{2} \cdot d) / year,

and 4.56

mgC / (m^{2} \cdot d) / year

, respectively; the primary productivity of Experiments 2, 5, and 6 declined rapidly after 2010 at a rate of> 20

mgC / (m^{2} \cdot d) / year

; Experiments 4 and 8 displayed a decreasing trend at first, then increased slightly and reached a minimum approximately in 2014, with a decline rate of 3–4

mgC / (m^{2} \cdot d) / year

on the whole. The same

Z_{e u}

were used in Experiments 1, 3, and 7; the same

C_{o p t}

were used in Experiments 2, 5, and 6; and the same

Z_{e u}

and

C_{o p t}

were used in Experiments 4 and 8; therefore, it can be speculated that

Z_{e u}

and

C_{o p t}

had a relatively large influence on interannual variations.

3.3. Spatial Distribution

The spatial distributions of the climatological annual mean primary productivity for the eight experiments in the YBS are shown in Figure 6. In general, the primary productivity of all the eight experiments is higher in the coastal waters and lower in the central Yellow Sea. However, there were still some differences in magnitude and spatial distribution patterns of primary productivity among the eight experiments. In the entire area, the primary productivity of Experiments 1, 2, and 3 was lower than that of the other experiments, especially in the coastal areas, and that of Experiment 8 was significantly higher among the eight experiments. In addition, in the Jiangsu Coastal water, the primary productivity of Experiments 1, 2, 5, and 6 was lower than that of the other areas, whereas Experiments 3, 4, 7, and 8 obtained a high value for the primary productivity. The main difference between the two groups was the use of different

C_{o p t}

data. Therefore, the chlorophyll-a data source could influence the magnitude and spatial distribution pattern of nearshore primary productivity in the VGPM.

The standard deviations in the shallow waters (water depth < 40 m), such as the Bohai Sea, Changjiang River Estuary, and Coast of Jiangsu Province, were relatively high (>500

mgC / (m^{2} \cdot d)

), and the coefficients of variation were also high (>60%) in these regions. However, the standard deviations and CVs were relatively low (<30%) in the offshore waters (water depth > 40 m) of the Yellow Sea (Figure 7). The results of the standard deviation and CV suggest that the influence of different data sources and parameterization schemes on the estimates of primary productivity of the VGPM in shelf seas is strongly correlated with water depth: shallower water has a stronger sensitivity in the VGPM, while the sensitivity in the deep water areas is relatively small.

4. Discussion

4.1. Influences of $C_{o p t}$ , $Z_{e u}$ , and $P_{o p t}^{B}$ on the VGPM Results

The aforementioned results suggest that the results of the VGPMs with different data sources and parameterization schemes had significant differences in magnitude, seasonal and interannual variations, and spatial distributions for the primary productivity, indicating that different data sources and parameterization schemes brought uncertainties to the application of the VGPM in shelf seas. In this study, we compared the values of

C_{o p t}

Z_{e u},

and

P_{o p t}^{B}

for different data sources and parameterization schemes to understand their influence on the VGPM results.

As shown in Figure 8a–c,

C_{o p t}

and

Z_{e u}

of the two data sources and

P_{o p t}^{B}

of the two parameterization schemes were different for each month.

C_{o p t}

of the two sources was significantly different in the autumn and winter, and the average

C_{o p t}_E S A

was approximately three times that of

C_{o p t}_M O D I S

, which was much larger than the difference in the spring and summer. The difference is primarily attributed to the influence of the high concentration of suspended sediment in the autumn and winter and the lack of calibration of the MODIS satellite data [18]. Therefore, the influence of different sources of

C_{o p t}

on the VGPM results in shelf seas is stronger in the autumn and winter than in the summer and spring. The

Z_{e u}

retrieved by the two algorithms were clearly different in the autumn and winter, with the average

Z_{e u}_K d

being approximately 1.5 times that of

Z_{e u}_I O P

, but they were similar in the spring and summer, suggesting that the influence of different

Z_{e u}

on the VGPM also mainly occurred in the autumn and winter.

P_{o p t}^{B}

obtained from different parameterization schemes was largely different in the summer and autumn, with the average

P_{o p t}^{B}_B

being approximately three times that of

P_{o p t}^{B}_Y

in the summer and 1.5 times in the autumn, whereas the difference between the two types of

P_{o p t}^{B}

in the spring and winter was relatively slight. Consequently, the influence of different parameterization schemes of

P_{o p t}^{B}

on the VGPM mainly occurred in the summer and autumn. In addition, combined with Figure 2, it can be seen that the low value of primary productivity of Experiments 1–4 in August may be related to the minimum value of

P_{o p t}^{B}_Y

in August, and the peak value of Experiments 5–8 from June to August may be related to the high level of

P_{o p t}^{B}_B

and small difference between the two sources of

C_{o p t}

and

Z_{e u}

from June to August. In summary, the different sources of

C_{o p t}

and

Z_{e u}

had a significant impact on the VGPM results in the autumn and winter, whereas the different parameterization schemes of

P_{o p t}^{B}

had a significant impact on the results in the summer.

As shown in Figure 8d–f, the patterns of the interannual variability of

C_{o p t}

Z_{e u},

and

P_{o p t}^{B}

from different sources and parameterization schemes were similar, but the magnitudes had large differences. The annual average of

C_{o p t}_E S A

was approximately 2.2 times that of

C_{o p t}_M O D I S

Z_{e u}_K d

was approximately 1.3 times that of

Z_{e u}_I O P

, and

P_{o p t}^{B}_B

was approximately 1.4 times that of

P_{o p t}^{B}_Y

. Among the three variables, only

C_{o p t}

(both

C_{o p t}_M O D I S

and

C_{o p t}_E S A

) showed decreasing trends, suggesting that the interannual downward trend of primary productivity in all eight experiments could have been caused by the decline in

C_{o p t}

The spatial distributions of

C_{o p t}

Z_{e u}

, and

P_{o p t}^{B}

from different sources and parameterization schemes (Figure 9) showed that there was little difference between

C_{o p t}_M O D I S

and

C_{o p t}_E S A

in the deep water areas of the central Yellow Sea, but there were significant differences in the shallow waters, such as the Bohai Sea, coastal waters of Jiangsu Province, and coastal waters of the Korean Peninsula. The magnitude of

C_{o p t}_E S A

was higher than that of

C_{o p t}_M O D I S

by approximately 4–6

mg / m^{3}

in the shallow waters. The spatial patterns of

Z_{e u}_I O P

and

Z_{e u}_K d

were similar; however,

Z_{e u}_K d

was 0–10 m higher than

Z_{e u}_I O P

. The spatial patterns of

P_{o p t}^{B}_Y

and

P_{o p t}^{B}_B

were relatively uniform, but

P_{o p t}^{B}_B

was approximately 1.5 times higher than

P_{o p t}^{B}_Y

in the entire study area. According to the previous results (Figure 7), the uncertainties in the primary productivity values in the shallow water areas were stronger than those in the central deep water areas. Because of the differences in

Z_{e u}

and

P_{o p t}^{B},

the different sources or parameterization schemes were relatively uniform in space (Figure 9f,i), and we could conclude that the different sources of

C_{o p t}

were the main reason for the spatial variability of uncertainty in the VGPM.

4.2. Comparison with Previous Studies

The monthly average primary productivity of the eight experiments was compared with the measured primary productivity values at 36 stations in the Yellow Sea obtained by Choi et al. in September 1992 [31], and their mean biases were calculated (Figure 10). The primary productivity of the eight experiments was significantly higher than that of the observed results. The mean biases between Experiments 1–4 and the observed results were relatively small, approximately 300

mgC / (m^{2} \cdot d)

, whereas those between Experiments 5–8 and the observed results were more than 700

mgC / (m^{2} \cdot d)

. This suggests that the VGPM estimates of the primary productivity in the YBS, adopting the parameterization scheme of

P_{o p t}^{B}_Y

, was significantly closer to the observed values. Although the monthly average results in the comparison and the observed values represent the instantaneous observation results, the inconsistency in time increases the artificial bias. However, this artificial bias could be the same for all eight experiments. Thus, the comparison is still instructive and suggests the importance of using a local

P_{o p t}^{B}

parameterization scheme in VGPM.

Additionally, we compared the primary productivity results from this study with those obtained from three alternative models: CAFE (Carbon, Absorption, and Fluorescence Euphotic-resolving [32]), CbPM (Carbon-Based Productivity Model [33]), and Eppley-VGPM (VGPM with the “Eppley” function [20]). These models were sourced from the Oregon State University website (http://orca.science.oregonstate.edu/index.php, accessed on 23 July 2023). The comparison revealed that the biases associated with the three alternative models were smaller than those of the VGPM using the parameterization

P_{o p t}^{B}_B

. Specifically, the biases of the CAFE and Eppley-VGPMs were relatively minor and comparable to those of the VGPM with the parameterization

P_{o p t}^{B}_Y

, indicating an improvement in these two models’ ability to estimate primary productivity in shelf seas. However, it is important to note that the biases for all models exceeded 300 mgC/(m²·d), representing 50% of the observed value. Consequently, caution is warranted when utilizing satellite estimates of primary productivity in shelf seas.

Compared to previous studies of the VGPM in the YBS, Tan and Shi, Jia et al., Yang, and Ding and Chen studied the spatiotemporal distribution of the primary productivity in the YBS using the VGPM [17,23,25,26]. The chlorophyll data products from MODIS were used in all of these studies. The parameterization scheme of

P_{o p t}^{B}_B

was used in these studies, except for the study by Tan and Shi, which used a parameterization scheme containing the chlorophyll concentration. Moreover, Tan and Shi and Ding and Chen used the empirical formula of

Z_{e u}

related to the chlorophyll concentration proposed by Morel and Berthon [34] in their studies, while Jia et al. and Yang used

Z_{e u}_K d

. Although the above studies used different sources of data and parameterization schemes, they all showed that the seasonal variations in primary productivity in the YBS had a bimodal distribution, which is consistent with the results of Experiments 1–4 in our study. In addition, this study found that the overall primary productivity values of the YBS showed declining trends. The decline in the primary productivity was consistent with the decrease in chlorophyll concentration in the YBS found by Wang et al., which could be related to the decrease in nutrient content in the YBS [35].

The results presented above indicate that primary productivity estimates derived from the VGPM are sensitive to the choice of input data sources and parameterizations. This sensitivity highlights the potential uncertainties associated with both process-based and empirical methods for estimating marine primary productivity. In recent years, process-informed machine learning approaches have been increasingly applied to marine environmental and ecological parameters, such as nutrients and chlorophyll, demonstrating significant improvements in data accuracy [36,37]. The application of machine learning techniques holds considerable promise for enhancing the accuracy of satellite-based estimates of marine primary productivity in the future.

5. Conclusions

This study assessed the sensitivity caused by the selection of data sources and parameterization schemes in the application of the VGPM for estimating the marine primary productivity in the YBS. The results showed that the VGPM using different data sources and parameterization schemes had significant differences in magnitude, seasonal and interannual variations, and spatial distributions for the primary productivity. The sensitivity of the VGPM was approximately 40% of the annual mean primary productivity. The minimum sensitivity was approximately 35% in the spring, whereas the maximum was approximately 70% in the winter and approximately 60% in other seasons. Spatially, the sensitivity of the nearshore shallow water areas (water depth < 40 m) was more than 60%, whereas that of the deep water areas was less than 30%. All VGPM results showed a decreasing trend in primary productivity in the YBS from 2003 to 2020.

C_{o p t}

data sources and

P_{o p t}^{B}

parameterization schemes are important for determining the magnitude of annual mean primary productivity. The seasonal variation in primary productivity was mainly dominated by the

P_{o p t}^{B}

parameterization schemes, whereas the spatial distributions was mostly affected by the data sources of

C_{o p t}

. In addition, the interannual decreasing trend was mainly related to the decreasing trend in

C_{o p t}

. VGPMs with the parameterized scheme

P_{o p t}^{B}_Y

were the most similar with respect to the other primary productivity models (e.g., CAFE and Eppley-VGPM), while those with the parameterized schemes

P_{o p t}^{B}_B

are the most different. Above all, this study suggests that because of the diversity of the data sources and parameterization schemes, the application of the VGPM in shelf seas, especially in coastal waters, could have strong uncertainty, which should be considered during the application of the VGPM in shelf seas.

Author Contributions

Conceptualization, Q.F.; Data curation, X.Z. and J.S.; Formal analysis, X.Z., J.S. and X.Y.; Funding acquisition, L.L. and Q.F.; Methodology, X.Y. and X.Z.; Visualization, Q.F.; Writing—original draft, X.Z., X.Y. and J.S.; Writing—review and editing, L.L. All authors have read and agreed to the published version of the manuscript.

Funding

The study was financially supported by the Open Research Fund of Shandong Key Laboratory of Marine Ecological Environment and Disaster Prevention and Mitigation (Grant No. 202305) and Open Research Fund of State Key Laboratory of Estuarine and Coastal Research (Grant No. SKLEC-KF202404).

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

These satellite data used in this study can be downloaded from the U.S. National Aeronautics and Space Administration (NASA) website (http://oceancolor.gsfc.nasa.gov), the European Space Agency (ESA) (https://www.oceancolour.org/), and the Oregon State University website (http://orca.science.oregonstate.edu/index.php).

Acknowledgments

The authors thank the National Oceanic and Atmospheric Administration, the European Space Agency, and the Oregon State University for their support regarding the satellite data.

Conflicts of Interest

Author Jianan Sun was employed by the company Ningbo Yonghuanyuan Environmental Engineering and Technology Co., Ltd. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

References

Field, C.B.; Behrenfeld, M.J.; Randerson, J.T.; Falkowski, P. Primary production of the biosphere: Integrating terrestrial and oceanic components. Science 1998, 281, 237–240. [Google Scholar] [CrossRef] [PubMed]
Behrenfeld, M.J.; Randerson, J.T.; McClain, C.R.; Feldman, G.C.; Los, S.O.; Tucker, C.J.; Pollack, N.H. Biospheric primary production during an ENSO transition. Science 2001, 291, 2594–2597. [Google Scholar] [CrossRef] [PubMed]
Thomas, H.; Bozec, Y.; Elkalay, K.; De Baar, H.J. Enhanced open ocean storage of CO₂ from shelf sea pumping. Science 2004, 304, 1005–1008. [Google Scholar] [CrossRef] [PubMed]
Laruelle, G.G.; Cai, W.J.; Hu, X.; Gruber, N.; Mackenzie, F.T.; Regnier, P. Continental shelves as a variable but increasing global sink for atmospheric carbon dioxide. Nat. Commun. 2018, 9, 454. [Google Scholar] [CrossRef] [PubMed]
Nielsen, E.S. The use of radio-active carbon (C14) for measuring organic production in the sea. ICES J. Mar. Sci. 1952, 18, 117–140. [Google Scholar] [CrossRef]
Wang, J. Primary Productivity of Phytoplankton and Black-and-white Bottle Oxygen Measurement. Freshw. Fish. 1980, 3, 26–30, (In Chinese with English abstract). [Google Scholar]
Fei, Z.L.; Zhu, M.Y. Determination of Marine Primary Productivity. Bohai Seas 1984, 1, 86–90, (In Chinese with English abstract). [Google Scholar]
Smith, R.C.; Eppley, R.W.; Baker, K.S. Correlation of primary production as measured aboard ship in southern California coastal waters and as estimated from satellite chlorophyll images. Mar. Biol. 1982, 66, 281–288. [Google Scholar] [CrossRef]
Eppley, R.W.; Stewart, E.; Abbott, M.R.; Heyman, U. Estimating ocean primary production from satellite chlorophyll. Introduction to regional differences and statistics for the Southern California Bight. J. Plankton Res. 1985, 7, 57–70. [Google Scholar] [CrossRef]
Platt, T. Primary production of the ocean water column as a function of surface light intensity: Algorithms for remote sensing. Deep Sea Res. Part A Oceanogr. Res. Pap. 1986, 33, 149–163. [Google Scholar]
Morel, A. Light and marine photosynthesis: A spectral model with geochemical and climatological implications. Prog. Oceanogr. 1991, 26, 263–306. [Google Scholar] [CrossRef]
Sathyendranath, S.; Longhurst, A.; Caverhill, C.M.; Platt, T. Regionally and seasonally differentiated primary production in the North Atlantic. Deep Sea Res. Part I Oceanogr. Res. Pap. 1995, 42, 1773–1802. [Google Scholar] [CrossRef]
Antoine, D.; André, J.M.; Morel, A. Oceanic primary production: 2. Estimation at global scale from satellite (coastal zone color scanner) chlorophyll. Glob. Biogeochem. Cycles 1996, 10, 57–69. [Google Scholar] [CrossRef]
Behrenfeld, M.J.; Falkowski, P.G. Photosynthetic rates derived from satellite-based chlorophyll concentration. Limnol. Oceanogr. 1997, 42, 1–20. [Google Scholar] [CrossRef]
Lomas, M.W.; Moran, S.B.; Casey, J.R.; Bell, D.W.; Tiahlo, M.; Whitefield, J.; Cokelet, E.D. Spatial and seasonal variability of primary production on the Eastern Bering Sea shelf. Deep Sea Res. Part II Top. Stud. Oceanogr. 2012, 65, 126–140. [Google Scholar] [CrossRef]
Vase, V.K.; Ajay, N.; Kumar, R.; Ramanathan, S.; Jayaraman, J.; Rohit, P. Temporal dynamics of primary productivity in the north-eastern Arabian Sea: An evaluation of ocean color models. Arab. J. Geosci. 2021, 14, 1338. [Google Scholar] [CrossRef]
Tan, S.; Shi, G. Satellite-derived primary productivity and its spatial and temporal variability in the China seas. J. Geogr. Sci. 2006, 16, 447–457. [Google Scholar] [CrossRef]
Wang, Y.; Liu, D.; Tang, D.L. Application of a generalized additive model (GAM) for estimating chlorophyll-a concentration from MODIS data in the Bohai and Yellow Seas, China. Int. J. Remote Sens. 2017, 38, 639–661. [Google Scholar] [CrossRef]
Shang, S.; Lee, Z.; Wei, G. Characterization of MODIS-derived euphotic zone depth: Results for the China Sea. Remote Sens. Environ. 2011, 115, 180–186. [Google Scholar] [CrossRef]
Carr, M.E.; Friedrichs, M.A.; Schmeltz, M.; Aita, M.N.; Antoine, D.; Arrigo, K.R.; Yamanaka, Y. A comparison of global estimates of marine primary production from ocean color. Deep Sea Res. Part II Top. Stud. Oceanogr. 2006, 53, 741–770. [Google Scholar] [CrossRef]
Yoon, J.E.; Park, J.; Yoo, S. Comparison of primary productivity algorithms for Korean waters. Ocean Sci. J. 2012, 47, 473–487. [Google Scholar] [CrossRef]
Jiao, N.; Liang, Y.; Zhang, Y.; Liu, J.; Zhang, Y.; Zhang, R.; Zhang, S. Carbon pools and fluxes in the China Seas and adjacent oceans. Sci. China Earth Sci. 2018, 61, 1535–1563. [Google Scholar] [CrossRef]
Yang, X.G. Remote Sensing Estimation of Chlorophyll and Primary Productivity in the Yellow Sea. Doctoral Dissertation, Graduate School of Chinese Academy of Sciences (Institute of Oceanography), Qingdao, China, 2013. (In Chinese with English abstract). [Google Scholar]
Cong, P.F. Remote Sensing Inversion of Marine Chlorophyll and Estimation of Marine Primary Productivity. Doctoral Dissertation, Institute of Remote Sensing Application, Chinese Academy of Sciences, Beijing, China, 2006. (In Chinese with English abstract). [Google Scholar]
Jia, Y.H.; Sun, Y.; Niu, B.Y. Estimation of marine primary productivity in the Yellow Sea based on RS and GIS. J. Huaihai Inst. Technol. 2010, 19, 87–91, (In Chinese with English abstract). [Google Scholar]
Ding, Q.X.; Chen, W.Z. Temporal and spatial variation of net primary productivity in China offshore based on VGPM. Ocean Dev. Manag. 2016, 8, 31–35, (In Chinese with English abstract). [Google Scholar]
Li, X.X.; Yuan, J.G.; Liu, X.J.; Li, Z.Y. Temporal and Spatial variation of net primary productivity in Bohai Sea based on MODIS data. J. Ecol. Environ. 2017, 26, 785–793, (In Chinese with English abstract). [Google Scholar]
Lee, Z.; Weidemann, A.; Kindle, J.; Arnone, R.; Carder, K.L.; Davis, C. Euphotic zone depth: Its derivation and implication to ocean-color remote sensing. J. Geophys. Res. Ocean. 2007, 112, C03009. [Google Scholar] [CrossRef]
Li, G.S.; Liang, Q.; Li, B.L. Seasonal changes of the euphotic depth in the East China Sea and its dynamic mechanism. Prog. Nat. Sci. 2003, 13, 90–94, (In Chinese with English abstract). [Google Scholar] [CrossRef]
Huang, N.E.; Shen, Z.; Long, S.R.; Wu, M.C.; Shih, H.H.; Zheng, Q.; Liu, H.H. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis. Proceedings of the Royal Society of London. Ser. A Math. Phys. Eng. Sci. 1998, 454, 903–995. [Google Scholar] [CrossRef]
Choi, J.K.; Noh, J.H.; Shin, K.S.; Hong, K.H. The early autumn distribution of chlorophyll-a and primary productivity in the Yellow Sea, 1992. Yellow Sea 1995, 1, 68–80. [Google Scholar]
Silsbe, G.M.; Behrenfeld, M.J.; Halsey, K.H.; Milligan, A.J.; Westberry, T.K. The CAFE model: A net production model for global ocean phytoplankton. Glob. Biogeochem. Cycles 2016, 30, 1756–1777. [Google Scholar] [CrossRef]
Behrenfeld, M.J.; Boss, E.; Siegel, D.A.; Shea, D.M. Carbon-based ocean productivity and phytoplankton physiology from space. Glob. Biogeochem. Cycles 2005, 19, GB1006. [Google Scholar] [CrossRef]
Morel, A.; Berthon, J.F. Surface pigments, algal biomass profiles, and potential production of the euphotic layer: Relationships reinvestigated in view of remote-sensing applications. Limnol. Oceanogr. 1989, 34, 1545–1562. [Google Scholar] [CrossRef]
Wang, Y.; Tian, X.; Gao, Z. Evolution of satellite derived chlorophyll-a trends in the Bohai and Yellow Seas during 2002–2018: Comparison between linear and nonlinear trends. Estuar. Coast. Shelf Sci. 2021, 259, 107449. [Google Scholar] [CrossRef]
Liu, H.; Lin, L.; Wang, Y.; Du, L.; Wang, S.; Zhou, P.; Yu, Y.; Gong, X.; Lu, X. Reconstruction of Monthly Surface Nutrient Concentrations in the Yellow and Bohai Seas from 2003–2019 Using Machine Learning. Remote Sens. 2022, 14, 5021. [Google Scholar] [CrossRef]
Hadjal, M.; Medina-Lopez, E.; Ren, J.; Gallego, A.; McKee, D. An Artificial Neural Network Algorithm to Retrieve Chlorophyll a for Northwest European Shelf Seas from Top of Atmosphere Ocean Colour Reflectance. Remote Sens. 2022, 14, 3353. [Google Scholar] [CrossRef]

Figure 1. The bathymetry of the YBS. The gray dashed line is the 40-m isobath.

Figure 2. Climatological monthly variation in the primary productivity of the eight experiments in the YBS from 2003 to 2020.

Figure 3. (a) Climatological monthly variation in the average primary productivity of the eight experiments. (b) The monthly variation in the coefficients of variation (CV).

Figure 4. Interannual variations of the primary productivity of the 8 experiments in the YBS from 2003 to 2020.

Figure 5. The residuals of the EMD analysis on the monthly mean primary productivity for the 8 experiments in the YBS from 2003 to 2020 (subfigures (a–h) correspond to Experiments 1–8, respectively).

Figure 6. The spatial distribution of the climatological mean primary productivity of the 8 experiments in the YBS.

Figure 7. The spatial distribution of the standard deviation (a) and the CV (b) of the primary productivity for the 8 experiments in the YBS.

Figure 8. (a–c) Climatological monthly means of

C_{o p t}

Z_{e u}

, and

P_{o p t}^{B}

of different sources and parameterization schemes, respectively. (d–f) Interannual variations from 2003 to 2020 of

C_{o p t}

Z_{e u}

, and

P_{o p t}^{B}

of different sources or parameterization schemes, respectively.

Figure 8. (a–c) Climatological monthly means of

C_{o p t}

Z_{e u}

, and

P_{o p t}^{B}

of different sources and parameterization schemes, respectively. (d–f) Interannual variations from 2003 to 2020 of

C_{o p t}

Z_{e u}

, and

P_{o p t}^{B}

of different sources or parameterization schemes, respectively.

Figure 9. The spatial distribution of the climatological mean

C_{o p t}

(a,b),

Z_{e u}

(d,e), and

P_{o p t}^{B}

(g,h), of the different data sources or parameterization schemes and their differences (c,f,i), respectively.

Figure 9. The spatial distribution of the climatological mean

C_{o p t}

(a,b),

Z_{e u}

(d,e), and

P_{o p t}^{B}

(g,h), of the different data sources or parameterization schemes and their differences (c,f,i), respectively.

Figure 10. Mean bias between the primary productivity of the VGPM (8 Exps.) and three alternative models (CAFE, CbPM, and Eppley-VGPM) and the observed values from Choi et al. in 1992 [31].

Table 1. The eight experiments of the VGPM estimates with different input variables combinations.

Exp.	Maximum Photosynthetic Rate	Euphotic Depth	Chlorophyll Concentration
1	$P_{o p t}^{B}_Y$	$Z_{e u}_I O P$	$C_{o p t}_M O D I S$
2	$P_{o p t}^{B}_Y$	$Z_{e u}_K d$	$C_{o p t}_M O D I S$
3	$P_{o p t}^{B}_Y$	$Z_{e u}_I O P$	$C_{o p t}_E S A$
4	$P_{o p t}^{B}_Y$	$Z_{e u}_K d$	$C_{o p t}_E S A$
5	$P_{o p t}^{B}_B$	$Z_{e u}_I O P$	$C_{o p t}_M O D I S$
6	$P_{o p t}^{B}_B$	$Z_{e u}_K d$	$C_{o p t}_M O D I S$
7	$P_{o p t}^{B}_B$	$Z_{e u}_I O P$	$C_{o p t}_E S A$
8	$P_{o p t}^{B}_B$	$Z_{e u}_K d$	$C_{o p t}_E S A$

Table 2. Statistical results of the climatological mean primary productivity of the eight experiments in the YBS from 2003 to 2020.

Experiments	Annual Mean (mgC/(m²∙d))	Monthly Max (mgC/(m²∙d))	Monthly Min (mgC/(m²∙d))
1	505.7	840.2	222.1
2	652.9	993.4	420.2
3	740.4	1297.7	471.9
4	1272.4	1748.4	528.4
5	903.2	1854.7	187.1
6	1103.0	1968.3	369.6
7	1167.3	2042.4	433.3
8	1778.3	2389.0	1079.3
Mean	1015.4 ± 308.5	1641.8 ± 507.9	464.0 ± 257.3

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

© 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

Share and Cite

MDPI and ACS Style

Zhao, X.; Sun, J.; Fu, Q.; Yan, X.; Lin, L. Sensitivity Assessment on Satellite Remote Sensing Estimates of Primary Productivity in Shelf Seas. J. Mar. Sci. Eng. 2024, 12, 2146. https://doi.org/10.3390/jmse12122146

AMA Style

Zhao X, Sun J, Fu Q, Yan X, Lin L. Sensitivity Assessment on Satellite Remote Sensing Estimates of Primary Productivity in Shelf Seas. Journal of Marine Science and Engineering. 2024; 12(12):2146. https://doi.org/10.3390/jmse12122146

Chicago/Turabian Style

Zhao, Xiaolong, Jianan Sun, Qingjun Fu, Xiao Yan, and Lei Lin. 2024. "Sensitivity Assessment on Satellite Remote Sensing Estimates of Primary Productivity in Shelf Seas" Journal of Marine Science and Engineering 12, no. 12: 2146. https://doi.org/10.3390/jmse12122146

APA Style

Zhao, X., Sun, J., Fu, Q., Yan, X., & Lin, L. (2024). Sensitivity Assessment on Satellite Remote Sensing Estimates of Primary Productivity in Shelf Seas. Journal of Marine Science and Engineering, 12(12), 2146. https://doi.org/10.3390/jmse12122146

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Menu

Sensitivity Assessment on Satellite Remote Sensing Estimates of Primary Productivity in Shelf Seas

Abstract

1. Introduction

2. Materials and Methods

2.1. Study Area

2.2. VGPM

2.3. Data Sources

2.4. Experimental Design

2.5. Data Analysis

3. Results

3.1. Seasonal Variations

3.2. Interannual Variations

3.3. Spatial Distribution

4. Discussion

4.1. Influences of $C_{o p t}$ , $Z_{e u}$ , and $P_{o p t}^{B}$ on the VGPM Results

4.2. Comparison with Previous Studies

5. Conclusions

Author Contributions

Funding

Institutional Review Board Statement

Informed Consent Statement

Data Availability Statement

Acknowledgments

Conflicts of Interest

References

Share and Cite

Article Metrics

Article Access Statistics

Further Information

Guidelines

MDPI Initiatives

Follow MDPI

Article Menu

Sensitivity Assessment on Satellite Remote Sensing Estimates of Primary Productivity in Shelf Seas

Abstract

1. Introduction

2. Materials and Methods

2.1. Study Area

2.2. VGPM

2.3. Data Sources

2.4. Experimental Design

2.5. Data Analysis

3. Results

3.1. Seasonal Variations

3.2. Interannual Variations

3.3. Spatial Distribution

4. Discussion

4.1. Influences of C o p t , Z e u , and P o p t B on the VGPM Results

4.2. Comparison with Previous Studies

5. Conclusions

Author Contributions

Funding

Institutional Review Board Statement

Informed Consent Statement

Data Availability Statement

Acknowledgments

Conflicts of Interest

References

Share and Cite

Article Metrics

Article Access Statistics

Further Information

Guidelines

MDPI Initiatives

Follow MDPI

4.1. Influences of $C_{o p t}$ , $Z_{e u}$ , and $P_{o p t}^{B}$ on the VGPM Results