1. bookVolumen 7 (2022): Edición 1 (January 2022)
Detalles de la revista
License
Formato
Revista
eISSN
2444-8656
Primera edición
01 Jan 2016
Calendario de la edición
2 veces al año
Idiomas
Inglés
access type Acceso abierto

Red tide monitoring method in coastal waters of Hebei Province based on decision tree classification

Publicado en línea: 24 Jun 2022
Volumen & Edición: Volumen 7 (2022) - Edición 1 (January 2022)
Páginas: 43 - 60
Recibido: 29 Oct 2021
Aceptado: 15 May 2022
Detalles de la revista
License
Formato
Revista
eISSN
2444-8656
Primera edición
01 Jan 2016
Calendario de la edición
2 veces al año
Idiomas
Inglés
Abstract

According to the water characteristics of the coastal waters of Hebei Province, this paper selects the data of the Marine Environmental Quality Bulletin of Hebei Province from 2009 to 2018 published on the website of the Department of Natural Resources of Hebei Province, and proposes a red tide monitoring method based on decision tree classification for the pre-processed MODIS 1B image data. The most important thing in the construction of decision tree is the determination of threshold, and this process is finally determined according to the value of Entropy. In this paper, the newly constructed red tide monitoring method is used to extract the occurrence area of red tide and count the red tide area. Finally, the decision tree classification method is compared with other typical red tide monitoring methods. The experimental results show that the red tide occurrence area and statistical area extracted by the red tide monitoring method based on decision tree classification are closer to the data displayed in the Ocean Bulletin, which demonstrates that this method is suitable for red tide monitoring in the coastal waters of Hebei Province.

Keywords

Introduction

Hebei Province is adjacent to the Bohai Sea, and the sea area under its jurisdiction is about 7,200 km2. Three of the 11 prefecture level cities in the province are coastal cities. With the development of economy, red tides occur frequently in the coastal waters of Hebei Province in recent years. Through consulting the Bulletin of Marine Environmental Quality of Hebei Province [1], it is found that there are five times of super large red tide information with an area of more than 1000km2: a red tide with an area of about 1000km2 occurred in the coastal waters of Qinhuangdao from late June to early September 2009, and a red tide with an area of about 3350km2 was found on June 24, 2010, from June 8 to August 20, 2012, The maximum area of water colour anomaly area monitored from June 8 to August 20, 2012 is about 3400km2, the maximum area of red tide found from June 20 to August 31, 2013 is about 1450km2, and the maximum area of red tide found in the offshore area of Qinhuangdao from May 15 to August 7, 2014 is about 2000km2. In addition to these extra-large red tides, there are many other types of red tides that have been occurring almost every year, including large red tides with an area of hundreds of kilometres, medium red tides with an area of 50–100 km and small red tides with an area of less than 50 km. Red tides endanger marine life, marine fishery production and human health and safety, and their occurrence additionally affects tourism, aquatic resources and fishery income in Hebei Province. The means to take scientific and effective methods to monitor red tides in time is of great significance. Hebei Normal University of Science and Technology is located in the beautiful coastal city of Qinhuangdao. In recent years, based on the need for transformation and development of theoretical knowledge about algal behaviour into practical application in attenuating the effect of red tides, the University attaches great importance to the integration of teaching and scientific research with local industries and local economy. For improving the marine environment and promoting the development of marine economy in Hebei Province, the University encourages teachers to actively carry out marine related research and focus on the methods of monitoring marine red tide, which will contribute to the marine development of Hebei Province.

MODIS Image

MODIS medium resolution imaging spectrometer mounted on Terra and Aqua satellites can obtain 36 bands of image data by scanning the global surface [2]. The resolution of image data in different bands is 250 m, 500 m or 1,000 m. MODIS images in various bands cover observation information about atmosphere, land and ocean [3], and MODIS images can be downloaded free of charge on NASA websites or some domestic websites, and this ready availability provides convenient conditions for using MODIS images for red tide monitoring [4] and real-time reflection of red tide changes [5]. Many scholars have applied MODIS remote sensing images to different sea areas and water bodies, and put forward different characteristic red tide monitoring methods. According to the water characteristics of the coastal waters of Hebei Province and the pre-processed MODIS 1B image data, a red tide monitoring method based on decision tree classification is proposed in this paper. Using this method, the occurrence region of red tide is extracted, and the red tide area is counted. Finally, this method is compared with other typical red tide monitoring methods that are based on MODIS remote sensing images. The red tide data used in this paper are from the Marine Environmental Quality Bulletin of Hebei Province from 2009 to 2018 published on the website of Hebei Provincial Department of Natural Resources. The remote sensing images used are MODIS Aqua 1B products downloaded from NASA website.

Image selection

Since the data quality of MODIS images is greatly affected by time and space, it is necessary to select high-quality images to facilitate the later red tide extraction. According to the Marine Environmental Quality Bulletin of Hebei Province, the red tide with the largest area of 2,000 km2 occurred in the coastal waters of Qinhuangdao from 15 May to 7 August, 2014. After comparison, the MODIS images of the Bohai Bay waters on 26 May 2014 were selected. The present study provides a detailed description of the red tide monitoring method based on decision tree classification, together with its experimental verification. There are two main reasons for selecting the images corresponding to 26 May 2014. First, the MODIS 1B image data quality of that day is relatively good, and the area of this red tide monitored by GOCI image is close to the maximum on that day [6]. Therefore, the experiment using the image of 26 May 2014 can be compared with the maximum area of red tide published in the Bulletin of Marine Environmental Quality of Hebei Province. Second, for the adaptability analysis of common red tide monitoring methods in Qinhuangdao sea area in document [7], the MODIS image of the same day is also used. The same scene image is used for the experiment, and thus it becomes convenient to compare the experimental results with the extraction results of other red tide monitoring methods.

Figure 1 shows the original MODIS image obtained on 26 May 2014, and Figure 2 shows the pre-processed image. The pre-processing operations in this paper mainly include the following steps: geometric correction, radiometric calibration, solar zenith angle correction and target area clipping. The decision tree classification method is described below based on the pre-processed image.

Fig. 1

Initial MODIS image

Fig. 2

Pre-processed image

Red tide monitoring method based on decision tree classification

The main principle of decision tree classification [8, 9] is to obtain the classification rules through certain methods, and then divide the remote sensing images level by level according to the classification rules. According to the relevant properties of decision tree, the decision tree classification method can be used to extract red tide information and achieve the purpose of monitoring red tides.

Image enhancement

In order to make it easier to distinguish the red tide occurrence region, the values of bands 1, 4 and 3 of the pre-processed image are used for true colour synthesis in this paper, and the obtained true colour image is shown in Figure 3. The dark blue ocean and the dark brown land can be distinguished from Figure 3. However, the colours of algae and phytoplankton in the ocean are somewhat similar to the surface of the sea, and thus these cannot be well distinguished. In order to make the colour of the image closer to human visual habits, the stretching method of the true colour image is adjusted from the original default linear stretching to histogram equalisation stretching, and the resultant enhanced image is shown in Figure 4. As can be seen from Figure 4, the ocean is blue, the soil is yellowish brown, the land vegetation is green or dark green and the algae and phytoplankton on the water body are light green, indicating that the overall tone level of the stretched image is relatively clear and in line with human visual habits. Therefore, the true colour image in Figure 4 is used as a reference diagram for interpreting red tide information.

Fig. 3

True colour image

Fig. 4

Enhanced image

Red tide monitoring model based on decision tree classification

Since the establishment of decision rules is the key to determine the classification quality of decision tree [10], only the establishment of scientific decision rules can better extract red tide. The red tide monitoring decision rules of MODIS Image are analysed as below.

First, according to the characteristics of remote sensing reflectance of MODIS Image, the specific method to distinguish land, red tide water and clean water in true colour image is determined. A=R3R5R3+R5 A = {{{R_3} - {R_5}} \over {{R_3} + {R_5}}} Eq. (1) is used to distinguish water and land (i.e. extracting water), where R3 and R5 are the remote sensing reflectance of band 3 and band 5, respectively. The formula is a new method obtained by improving the normalised difference water body index method. The improved formula can further strengthen the ratio of water body and reduce the water body index of background objects, which is conducive to the segmentation of threshold. R=R4R1R3R1 R = {{{R_4} - {R_1}} \over {{R_3} - {R_1}}} The method of distinguishing a red tide water body from a clean water body (i.e. judging whether the water colour is abnormal or not) adopts the multi-band difference ratio method [13], that is, the band difference ratio is established with bands 1, 3 and 4. Eq. (2) is shown below, where R1, R3 and R4 are the remote sensing reflectance of bands 1, 3 and 4, respectively.

Then, in order to determine the critical values of A and R in the classification process, 12 point data are selected in the land, water colour anomaly area and clean water area of remote sensing image, respectively, and the values of each point are calculated by Eqs (1) and (2). The calculation results are shown in Table 1. It can be clearly seen from Table 1 that among the 12 samples, the A values of land points are all less than 0, while the A values of abnormal water and clean water points are all greater than 0. Therefore, based on whether the A value is greater than 0, it becomes possible to distinguish water and land. For the distinction between red tide water and clean water, the value range of R can be determined by in-depth analysis of water colour anomaly area and clean water area. After repeated experimental verification, the threshold of R is finally determined as 0.45.

Sample calculation results

Sample point Land Abnormal water area Clean water area
A R A R A R
1 –0.490186 0.431966 0.208190 0.520284 0.157197 0.343478
2 –0.515848 0.436719 0.197401 0.511940 0.224858 0.336290
3 –0.488859 0.328717 0.195482 0.518240 0.179280 0.350442
4 –0.456237 0.349821 0.176524 0.524408 0.245110 0.403250
5 –0.341247 0.420170 0.167894 0.473622 0.159053 0.331198
6 –0.451948 0.454620 0.180254 0.535965 0.376934 0.249489
7 –0.392420 0.459112 0.189532 0.505973 0.342486 0.267309
8 –0.398071 0.518295 0.197775 0.519787 0.318415 0.143954
9 –0.477858 0.515863 0.218227 0.510321 0.230179 0.421184
10 –0.502500 0.504848 0.183310 0.522109 0.248343 0.393807
11 –0.502628 0.430932 0.172382 0.467220 0.141973 0.334235
12 –0.461805 0.554743 0.169851 0.525808 0.149018 0.372128

According to the above methods and threshold analysis, a red tide monitoring model based on decision tree classification can be established, as shown in Figure 5.

Fig. 5

Red tide monitoring model based on decision tree classification

According to the red tide monitoring model based on the decision tree shown in Figure 5, the classification rules for extracting land, clean water and red tide water can be summarised as follows: (1) Land: A < 0. (2) Clean water area: A ≥ 0 and R ≥ 0.45. (3) Red tide water body area: A ≥ 0 and R < 0.45.

Experiment

In order to evaluate the actual effect of the red tide monitoring model based on decision tree, this section applies the model to the MODIS 1B image on May 26 for experimental verification.

First, the method of extracting water body is tested. Figure 6 shows the band grey map of the pre-processed image. According to the decision tree classification rules, the threshold of Eq. (1) is set to 0 to obtain the image after extracting water body, as shown in Figure 7. It can be seen from Figure 7 that the Bohai Bay is well identified as a whole, which lays a foundation for determining the Qinhuangdao sea area.

Fig. 6

Band grayscale

Fig. 7

Water body extraction

Then, according to the method of judging whether the water colour is abnormal, the result shown in Figure 8 is obtained after processing Figure 7. The red part in Figure 8 represents the red tide water body and the blue part represents the clean water body. Based on comparing with the true colour image in Figure 4, the sea area of Qinhuangdao is marked with an ellipse. Using ENVI area statistics tool, it is calculated that the area of red tide water in Qinhuangdao sea area is about 1,943 km2, which is consistent with the red tide occurrence area and the maximum red tide area of 2,000 km2 in Qinhuangdao sea area shown in Hebei Marine Environmental Quality Bulletin in 2014. Therefore, it can be determined that the decision tree classification method can extract red tide information.

Fig. 8

Inversion results of decision tree classification

In order to verify whether the decision tree classification method provides a performance that is better than those of other methods, MODIS images corresponding to the same scene were also tested using the typical red tide monitoring methods (i.e. chlorophyll method [11, 12], single band ratio method [13, 14] and multi-band difference ratio method [15]).

The inversion results of Chla concentration obtained by the empirical model of the OC3M algorithm is shown in Figure 9. The colour of the image is relatively single and close, and thus it is difficult to distinguish the water body and the land part. Therefore, the stretching range of this image is adjusted from the original default Linear (linear stretching) to Equalisation (histogram balanced stretching). The stretching principle is to stretch the narrower brightness range in the original image to the full radiance range (0–255). After linear stretching, the grey distribution range of each band is widened, and the contrast of the image is also improved. The stretching formula is shown in Eq. (3), and the results of the stretching are shown in Fig. 10. From the enhanced inversion results, the land and the ocean can be clearly distinguished; further, it is possible to find the location of the Qinhuangdao waters, and the location of the red tide can also be clearly displayed. xbb1b2b1=xaa1a2a1b2b1a2a1(xaa1)(xa[ a1,a2 ],xb[ b1,b2 ]) {{{x_b} - {b_1}} \over {{b_2} - {b_1}}} = {{{x_a} - {a_1}} \over {{a_2} - {a_1}}} \Rightarrow {{{b_2} - {b_1}} \over {{a_2} - {a_1}}}\left( {{x_a} - {a_1}} \right)\left( {{x_a} \in \left[ {{a_1},{a_2}} \right],{x_b} \in \left[ {{b_1},{b_2}} \right]} \right)

Fig. 9

Inversion results of Chla concentration

Fig. 10

Inversion of image enhancement results

Figure 11 is the synthetic pseudo-colour image, which is the inversion result of Chla concentration assigned to MODIS by the three RGB channels of the colour map. Using the values of the 1 and 2 bands, it becomes possible to clearly distinguish the dark brown ocean and the purple land. The overall tone level of the image is relatively clear, and it is easy to distinguish the area where the red tide occurs. Through repeated experiments, it is concluded that when the concentration of chlorophyll a is greater than 2 mg/m3 (this value is the measured data of the year), the seawater surface is abnormal. The results are shown in Figure 12, in which the abnormal area (the red and orange part) is exactly the sea area of Qinhuangdao, and the concentration changes are also shown in the same figure.

Fig. 11

Pseud colour synthesis

Fig. 12

Image enhancement

The method used for the extraction of red tide water based on single band is shown below: R4R3>Cr {{{R_4}} \over {{R_3}}} > Cr where R3 is the reflectivity of MODIS channel 3, R4 is the reflectivity of MODIS channel 4 and Cr is a constant. The ratio of Eq. (4) mainly reflects the concentration of chlorophyll a in the water body. Figure 13 is the result obtained by Eq. (4), and the value of Cr has not been determined at this time. The results obtained through the mask are shown in Figure 14. The DN value (pixel value) of the Qinhuangdao sea area (the circled area in Figure 14) is analysed, and it is found that the DN value of the red tide occurrence area is perceptibly larger than that of the surrounding sea area, with all values being above 0.9 [54]. Using the red tide area extracted based on Cr > 0.9, we ascertain that the red part in the Qinhuangdao area circled in Figure 14 is the desired area.

Fig. 13

Single band ratio method

Fig. 14

Red tide extraction results

The red tide monitoring method based on the multi-band difference ratio is to use Eq. (2) to monitor the red tide information in the Qinhuangdao sea area, and the effect is shown in Figure 15. Through the analysis of the DN value of the Qinhuangdao sea area, the results are obtained after repeated experiments. When R < 0.45, it is a non-red tide water body, and when R > 0.45, it is a red tide water body. The results are shown in Figure 16.

Fig. 15

Multi-band difference ratio method

Fig. 16

Red tide extraction results

The areas of red tide water bodies extracted based on the four methods mentioned above are shown in Table 2.

Red tide area extracted by various methods

Red tide monitoring method Red tide area
Chlorophyll method 1,548 km2
Single band ratio method 2,780 km2
Multi-band difference ratio method 1,866 km2
Decision tree classification 1,943 km2

The comparative analysis of the experimental results shows that the red tide areas extracted by chlorophyll method, multi-band difference ratio method and decision tree classification method are basically the same. The red tide area extracted by chlorophyll method is about 1,548 km2, the red tide area extracted by single band ratio method is about 2,780 km2, the red tide area extracted by multi-band difference ratio method is about 1,866 km2 and the red tide area extracted by decision tree classification method is 1,943 km2. Compared with the 2,000 km2 red tide area shown in the Ocean Bulletin, the error of the decision tree classification method is smallest. Among the four methods mentioned above, the red tide area extracted by single band ratio method far exceeds the actual maximum area, which is similar to the verification results available in the literature [6]. The measurement error is the largest, indicating that this method extracts more false red tide information and is not suitable for red tide monitoring in the coastal waters of Hebei Province. It is normal that the red tide area extracted by the other three methods is less than the actual maximum area, because the red tide is developing and changing, and the red tide on that day does not reach the maximum area. Through comparative analysis, it can be seen that the red tide area extracted by the decision tree classification method is the closest to the data shown in the Hebei Provincial Marine Environmental Quality Bulletin, indicating that the decision tree classification method has better red tide monitoring effect and higher accuracy than other methods. In order to test the adaptability of the decision tree classification method to other types of red tides, some large, medium and small red tides were selected from the red tides from 2009 to 2018. In order to assess the red tide extraction area more intuitively, the red tide monitoring effect map and red tide area based on the decision tree classification method from 20 June 2009 to 12 July 2009 are given below. The comparison of its true colour map and decision tree classification is shown in Figures 17 to 28.

Fig. 17

True colour synthesis on 20 June 2009

Fig. 18

Inversion results of decision tree model on 20 June 2009

Fig. 19

True colour synthesis on 22 June 2009

Fig. 20

Inversion results of decision tree model on 22 June 2009

Fig. 21

True colour synthesis on 24 June 2009

Fig. 22

Inversion results of decision tree model on 24 June 2009

Fig. 23

True colour synthesis on 26 June 2009

Fig. 24

Inversion results of decision tree model on 26 June 2009

Fig. 25

True colour synthesis on 3 July 2009

Fig. 26

Inversion results of decision tree model on 3 July 2009

Fig. 27

True colour synthesis on 12 July 2009

Fig. 28

Inversion results of decision tree model on 12 July 2009

Based on analysing the results of the six groups of image data, it can be inferred that there is no large area of abnormal colour in the Qinhuangdao sea area in Figure 17. At the same time, similar results are also obtained in the decision tree model result diagram in Figure 18, with only sporadic red tide areas. In Figure 19, it can be seen that there is a large area of abnormal colour in the Qinhuangdao waters. Meanwhile, it can be seen that there is a large red tide area in the result of the decision tree model in Figure 20. Until 24 June 2009, the red tide expanded very fast, and the affected area was also very large. In Figure 23–26, it can be seen that the area of red tide gradually shrinks from 26 June to 3 July, 2009. In Figures 27 and 28, the red tide area is shown to be significantly smaller.

In order to facilitate the sorting out of the occurrence process of red tides, the red tide area on June 20, 22, 24 and 26 and July 3 and 12, 2009 is now displayed in the form of a curve graph. The red tide area statistical chart is shown in Figure 29.

Fig. 29

Statistical chart of red tide area

Figure 29 indicates the changes in the area of red tides clearly. The early stage of red tide outbreaks commenced on 20 June 2009, and large-scale outbreaks began on 22 June 2009. The outbreak of red tides in Qinhuangdao waters was the most serious on 24 June 2009, and the area of red tides began to shrink gradually between 25 June and 3 July.

Through the verification, it was found that the decision tree classification method has a good effect on the detection of red tides above 100 km2, and the monitoring area is highly consistent with the publicly displayed data. The monitoring effect of red tides below 100 km2 is relatively poor, and especially, the error of the extracted area is large, which is related to the resolution limit of the MODIS images themselves, and the error of ENVI area statistics tool. The maximum resolution of each band of MODIS Image is 250 m. The larger the area, the easier it is to distinguish the red tide. The boundary and area statistics are relatively accurate. It is difficult to define the red tide boundary using an area that is too small. Of course, the accuracy of area statistics cannot be guaranteed.

In general, the red tide monitoring method based on decision tree classification can extract the red tide occurrence area more accurately, and this method is more suitable than other methods for red tide monitoring in the coastal waters of Hebei Province.

Conclusions

The red tide monitoring method based on decision tree classification uses the reflectance of the 3rd and 5th bands of MODIS Image to extract the water body. This method avoids the error caused by manual mask and can accurately separate the water body and land. The red tide monitoring method based on decision tree classification provides a new scheme for red tide monitoring using MODIS remote sensing images. However, due to the complex inducing conditions of red tides, and the occurrence time of red tides generally lasting for several days to several months, the use of a single MODIS Image can only reflect the current situation of red tides, and cannot enable an understanding of red tide dynamics. In the future, MODIS Image will be used to continuously monitor the development process of red tides, and to constantly explore the inducing causes and influencing factors of red tides. Moreover, the red tide monitoring method based on multi-source remote sensing images will continue to be studied, and the red tide monitoring method based on the combination of multi-source remote sensing images and decision tree classification method will be studied.

Fig. 1

Initial MODIS image
Initial MODIS image

Fig. 2

Pre-processed image
Pre-processed image

Fig. 3

True colour image
True colour image

Fig. 4

Enhanced image
Enhanced image

Fig. 5

Red tide monitoring model based on decision tree classification
Red tide monitoring model based on decision tree classification

Fig. 6

Band grayscale
Band grayscale

Fig. 7

Water body extraction
Water body extraction

Fig. 8

Inversion results of decision tree classification
Inversion results of decision tree classification

Fig. 9

Inversion results of Chla concentration
Inversion results of Chla concentration

Fig. 10

Inversion of image enhancement results
Inversion of image enhancement results

Fig. 11

Pseud colour synthesis
Pseud colour synthesis

Fig. 12

Image enhancement
Image enhancement

Fig. 13

Single band ratio method
Single band ratio method

Fig. 14

Red tide extraction results
Red tide extraction results

Fig. 15

Multi-band difference ratio method
Multi-band difference ratio method

Fig. 16

Red tide extraction results
Red tide extraction results

Fig. 17

True colour synthesis on 20 June 2009
True colour synthesis on 20 June 2009

Fig. 18

Inversion results of decision tree model on 20 June 2009
Inversion results of decision tree model on 20 June 2009

Fig. 19

True colour synthesis on 22 June 2009
True colour synthesis on 22 June 2009

Fig. 20

Inversion results of decision tree model on 22 June 2009
Inversion results of decision tree model on 22 June 2009

Fig. 21

True colour synthesis on 24 June 2009
True colour synthesis on 24 June 2009

Fig. 22

Inversion results of decision tree model on 24 June 2009
Inversion results of decision tree model on 24 June 2009

Fig. 23

True colour synthesis on 26 June 2009
True colour synthesis on 26 June 2009

Fig. 24

Inversion results of decision tree model on 26 June 2009
Inversion results of decision tree model on 26 June 2009

Fig. 25

True colour synthesis on 3 July 2009
True colour synthesis on 3 July 2009

Fig. 26

Inversion results of decision tree model on 3 July 2009
Inversion results of decision tree model on 3 July 2009

Fig. 27

True colour synthesis on 12 July 2009
True colour synthesis on 12 July 2009

Fig. 28

Inversion results of decision tree model on 12 July 2009
Inversion results of decision tree model on 12 July 2009

Fig. 29

Statistical chart of red tide area
Statistical chart of red tide area

Red tide area extracted by various methods

Red tide monitoring method Red tide area
Chlorophyll method 1,548 km2
Single band ratio method 2,780 km2
Multi-band difference ratio method 1,866 km2
Decision tree classification 1,943 km2

Sample calculation results

Sample point Land Abnormal water area Clean water area
A R A R A R
1 –0.490186 0.431966 0.208190 0.520284 0.157197 0.343478
2 –0.515848 0.436719 0.197401 0.511940 0.224858 0.336290
3 –0.488859 0.328717 0.195482 0.518240 0.179280 0.350442
4 –0.456237 0.349821 0.176524 0.524408 0.245110 0.403250
5 –0.341247 0.420170 0.167894 0.473622 0.159053 0.331198
6 –0.451948 0.454620 0.180254 0.535965 0.376934 0.249489
7 –0.392420 0.459112 0.189532 0.505973 0.342486 0.267309
8 –0.398071 0.518295 0.197775 0.519787 0.318415 0.143954
9 –0.477858 0.515863 0.218227 0.510321 0.230179 0.421184
10 –0.502500 0.504848 0.183310 0.522109 0.248343 0.393807
11 –0.502628 0.430932 0.172382 0.467220 0.141973 0.334235
12 –0.461805 0.554743 0.169851 0.525808 0.149018 0.372128

Hebei Ocean Bureau. Hebei Marine Environmental Quality Bulletin [EB/OL]. Available from: http://zrzy.hebei.gov.cn/heb/gk/tjxx/. 2009–2018. Hebei Ocean Bureau Hebei Marine Environmental Quality Bulletin [EB/OL] Available from: http://zrzy.hebei.gov.cn/heb/gk/tjxx/. 2009–2018. Search in Google Scholar

Huang WY, Xiao QM, Lou XL. Progress in Remote Sensing Technology and Application of Red Tide Satellite at Home and Abroad. Sensing Technology and Application. 2002; 17(1): 32–36. HuangWY XiaoQM LouXL Progress in Remote Sensing Technology and Application of Red Tide Satellite at Home and Abroad Sensing Technology and Application 2002 17 1 32 36 Search in Google Scholar

MODIS User Manual. MODIS User Manual Search in Google Scholar

Liang J. Application of Remote Sensing Technology in Red Tide Monitoring in Coastal Waters. Technology Utilization. 2019; 26(5): 156–158. LiangJ Application of Remote Sensing Technology in Red Tide Monitoring in Coastal Waters Technology Utilization 2019 26 5 156 158 Search in Google Scholar

Zhang T, Su FZ, Yang XM, et al. Method and application of extracting red tide information from MODIS remote sensing data - a case study of Zhujiangkou. Journal of Earth Information Science. 2009; 11(2): 243–249. ZhangT SuFZ YangXM Method and application of extracting red tide information from MODIS remote sensing data - a case study of Zhujiangkou Journal of Earth Information Science 2009 11 2 243 249 Search in Google Scholar

Jiang DJ, Wang K, Xia Y. Comparative study on remote sensing monitoring methods of red tide in Bohai Sea. Marine Environmental Science. 2020; 39(3): 460–467. JiangDJ WangK XiaY Comparative study on remote sensing monitoring methods of red tide in Bohai Sea Marine Environmental Science 2020 39 3 460 467 Search in Google Scholar

He Z, Xiao SZ. Study on red tide remote sensing monitoring in the bohai sea in 2014. Mar. Bull. 2017; 19: 37–51. HeZ XiaoSZ Study on red tide remote sensing monitoring in the bohai sea in 2014 Mar. Bull. 2017 19 37 51 Search in Google Scholar

Chen JL, Wang Ch. Remote Sensing Image Classification Based on ENVI. Beijing Surveying and Mapping. 2018; 32(8): 933–937. ChenJL WangCh Remote Sensing Image Classification Based on ENVI Beijing Surveying and Mapping 2018 32 8 933 937 Search in Google Scholar

Cao HD, Hong YT, Zhang W, et al. Remote Sensing Image Decision Tree Classification Based on ENVI. Beijing Surveying and Mapping. 2017; 2: 67–71. CaoHD HongYT ZhangW Remote Sensing Image Decision Tree Classification Based on ENVI Beijing Surveying and Mapping 2017 2 67 71 Search in Google Scholar

Chen LP, Huang SW. Research on decision tree classification method of remote sensing image based on rules. Journal of Anhui University of Technology. 2014; 34(4): 46–50. ChenLP HuangSW Research on decision tree classification method of remote sensing image based on rules Journal of Anhui University of Technology 2014 34 4 46 50 Search in Google Scholar

Fan XC, Chen SB, Zhou C, et al. Ocean chlorophyll inversion algorithm based on MODIS data. Science, Technology and Engineering. 2017; 17(10): 7150. FanXC ChenSB ZhouC Ocean chlorophyll inversion algorithm based on MODIS data Science, Technology and Engineering 2017 17 10 7150 Search in Google Scholar

Zhanga XG, Xub YS. Study on Dynamic Model of Marine Ecosystem Based on Chlorophyll Concentration Distribution. Geoscience and Remote Sensing. 2021; 4(1): 5131. ZhangaXG XubYS Study on Dynamic Model of Marine Ecosystem Based on Chlorophyll Concentration Distribution Geoscience and Remote Sensing 2021 4 1 5131 Search in Google Scholar

Wu YW. Application of Remote Sensing Image in Red Tide Monitoring in Qinhuangdao Sea Area. Qinhuangdao: Northeastern University; 2013. WuYW Application of Remote Sensing Image in Red Tide Monitoring in Qinhuangdao Sea Area Qinhuangdao Northeastern University 2013 Search in Google Scholar

Prayogo LM, Basith A. Shallow Water Depth Mapping Using Single Band and Band Ratio on High-Resolution Imagery. Seminar Nasional Geomatika. 2020; p. 1–9. PrayogoLM BasithA Shallow Water Depth Mapping Using Single Band and Band Ratio on High-Resolution Imagery Seminar Nasional Geomatika 2020 1 9 10.24895/SNG.2020.0-0.1179 Search in Google Scholar

Bian MM, Song JL, Wang L, et al. Red tide monitoring method in Qinhuangdao sea area based on multi-band difference ratio method. Hebei Fisheries. 2020; 1: 3187. BianMM SongJL WangL Red tide monitoring method in Qinhuangdao sea area based on multi-band difference ratio method Hebei Fisheries 2020 1 3187 Search in Google Scholar

Artículos recomendados de Trend MD

Planifique su conferencia remota con Sciendo