Developing data-driven models for quantifying Cochlodinium polykrikoides using the Geostationary Ocean Color Imager (GOCI)

Harmful algal blooms have caused critical problems worldwide because they pose serious threats to human health and aquatic ecosystems. In particular, red tide blooms of Cochlodinium polykrikoides have caused serious damage to aquaculture in Korean coastal waters. In this study, multiple linear regression, regression tree (RT), and Random Forest models were applied to detect C. polykrikoides blooms in coastal waters. Five types of input data sets were implemented to test the performance of the models. The observed number of C. polykrikoides cells and reflectance data from Geostationary Ocean Color Imager images obtained in a 3-year period (2013–2015) were used to train and validate the models. The RT model demonstrated the best prediction performance when four bands and three-band ratio data were simultaneously used as input data. The results obtained via iterative model development with randomly chosen input data indicate that the recognition of patterns in the training data caused variations in the prediction performance. This work provides useful tools for reliable estimation of the number of C. polykrikoides cells using reasonable coastal water reflectance data sets. It is expected that administrators and decision-makers whose work is associated with coastal waters will be able to easily access and manipulate the RT model.     

Remote sensing chlorophyll a of optically complex waters (rias Baixas, NW Spain): Application of a regionally specific chlorophyll a algorithm for MERIS full resolution data during an upwelling cycle

This study takes advantage of a regionally specific algorithm and the characteristics of Medium Resolution Imaging Spectrometer (MERIS) in order to deliver more accurate, detailed chlorophyll a (chla) maps of optically complex coastal waters during an upwelling cycle. MERIS full resolution chla concentrations and in situ data were obtained on the Galician (NW Spain) shelf and in three adjacent rias (embayments), sites of extensive mussel culture that experience frequent harmful algal events. Regionally focused algorithms (Regional neural network for rias Baixas or NNRB) for the retrieval of chla in the Galician rias optically complex waters were tested in comparison to sea-truth data. The one that showed the best performance was applied to a series of six MERIS (FR) images during a summer upwelling cycle to test its performance. The best performance parameters were given for the NN trained with high-quality data using the most abundant cluster found in the rias after the application of fuzzy c-mean clustering techniques (FCM). July 2008 was characterized by three periods of different meteorological and oceanographic states. The main changes in chla concentration and distribution were clearly captured in the images. After a period of strong upwelling favorable winds a high biomass algal event was recorded in the study area. However, MERIS missed the high chlorophyll upwelled water that was detected below surface in the ria de Vigo by the chla profiles, proving the necessity of in situ observations. Relatively high biomass “patches” were mapped in detail inside the rias. There was a significant variation in the timing and the extent of the maximum chla areas. The maps confirmed that the complex spatial structure of the phytoplankton distribution in the rias Baixas is affected by the surface currents and winds on the adjacent continental shelf. This study showed that a regionally specific algorithm for an ocean color sensor with the characteristics of MERIS in combination with in situ data can be of great help in chla monitoring, detection and study of high biomass algal events in an area affected by coastal upwelling such as the rias Baixas.     

Detecting the red tide algal blooms from satellite ocean color observations in optically complex Northeast-Asia Coastal waters

Over the last few decades, the coastal regions throughout the world have experienced incidences of algal blooms, which are harmful or otherwise toxic because of their potential threat to humans as well as marine organisms, owing to accelerated eutrophication from human activities and certain oceanic processes. Previous studies have found that correct identification of these blooms remains a great challenge with the standard bio-optical algorithms applied to satellite ocean color data in optically complex coastal waters containing high concentrations of the interfered dissolved organic and particulate inorganic materials. Here a new method called the red tide index (RI) is presented which is capable of identifying potential areas of harmful algal blooms (HABs) from SeaWiFS ocean color measurements representing the typical Case-2 water environments off the Korean and Chinese coasts. The RI method employs the water-leaving radiances (L_w), collected from in-situ radiometric measurements of three SeaWiFS bands centered at 443 nm, 510 nm and 555 nm, to achieve derivation of indices that are then related to absorbing characteristics of harmful algae (i.e., L_w at 443 nm) from which a best fit with a cubic polynomial function with correlation coefficient of R² = 0.91 is obtained providing indices of higher ranges for HABs and lower and slightly reduced ranges for turbid and non-bloom waters. Similar indices derived from the use of remote sensing reflectance (R_rs), normalized water-leaving radiance (nL_w) and combination of both are found rather inadequate to characterize the variability of the encountered bloom. In order to quantify the HABs in terms of chlorophyll (Chl), an empirical relationship is established between the RI and in-situ Chl in surface waters from about 0.4-71 mg m^− 3, which yields a Red tide index Chlorophyll Algorithm (RCA) based on an exponential function with correlation coefficient R² = 0.92. The established methods were extensively tested and compared with the performances of standard Ocean Chlorophyll 4 (OC4) algorithm and Local Chlorophyll Algorithm (LCA) using SeaWiFS images collected from typical red tide waters of Korean South Sea (KSS), East China Sea (ECS), Yellow Sea (YS) and Bohai Sea (BS) during 1999-2002. The standard spectral ratio algorithms, the OC4 and LCA, yielded large errors in Chl retrievals for coastal areas, besides providing false information about the encountered HABs in KSS, ECS, YS and BS waters. On the contrary, the RI coupled with the standard spectral ratios yielded comprehensive information about various ranges of algal blooms, while RCA Chl showing a good agreement with in-situ data led to enhanced understanding of the spatial and temporal variability of the recent HAB occurrences in high scattering and absorbing waters off the Korean and Chinese coasts.     

Solving the ocean color problem using a genetic programming approach

The ocean color problem consists in evaluating ocean components concentrations (phytoplankton, sediment and yellow substance) from sunlight reflectance or luminance values at selected wavelengths in the visible band. The interest of this application increases with the availability of new satellite sensors. Moreover, monitoring phytoplankton concentrations is a key point for a wide set of problems ranging from greenhouse effect to industrial fishing and signaling toxic algae blooms. To our knowledge, it is the first attempt at this regression problem with genetic programming (GP). We show that GP outperforms traditional polynomial fits and rivals artificial neural nets in the case of open ocean waters. We improve previous works by also solving a range of coastal waters types, providing detailed results on estimation errors. To our knowledge, we are the firsts to publish numerical results regarding coastal waters. Experiments were conducted with a dynamic fitness GP algorithm in order to speed up computing time through a process of progressive learning.     

Neural network estimation of chlorophyll a from MERIS full resolution data for the coastal waters of Galician rias (NW Spain)

In typical Case 2 waters, accurate remote sensing retrieval of chlorophyll a (chla) is still a challenging task. In this study, focusing on the Galician rias (ΝW Spain), algorithms based on neural network (NN) techniques were developed for the retrieval of chla concentration in optically complex waters, using Medium Resolution Imaging Spectrometer (MERIS) data. There is considerable interest in the accurate estimation of chla for the Galician rias, because of the economic and social importance of the extensive culture of mussels, and the high frequency of harmful algal events. Fifteen MERIS full resolution (FR) cloud-free images paired with in situ chla data (for 2002-2004 and 2006-2008) were used for the development and validation of the NN. The scope of NN was established from the clusters obtained using fuzzy c-mean (FCM) clustering techniques applied to the satellite-derived data. Three different NNs were developed: one including the whole data set, and two others using only points belonging to one of the clusters. The input data for these latter two NNs was chosen depending on the quality level, defined on the basis of quality flags given to each data set. The fitting results were fairly good and proved the capability of the tool to predict chla concentrations in the study area. The best prediction was given for the NN trained with high-quality data using the most abundant cluster data set. The performance parameters in the validation set of this NN were R² = 0.86, mean percentage error (MPE) = − 0.14, root mean square error (RMSE) = 0.75 mg m^− 3, and relative RMSE = 66%. The NN developed in this study detected accurately the peaks of chla, in both training and validation sets. The performance of the Case-2-Regional (C2R) algorithm, routinely used for MERIS data, was also tested and compared with our best performing NN and the sea-truthing data. Results showed that this NN outperformed the C2R, giving much higher R² and lower RMSE values.This study showed that the combination of in situ data and NN technology improved the retrieval of chla in Case 2 waters, and could be used to obtain more accurate chla maps. A local-based algorithm for the chla retrieval from an ocean colour sensor with the characteristics of MERIS would be a great support in the quantitative monitoring and study of harmful algal events in the coastal waters of the Rias Baixas. The limitations and possible improvements of the developed chla algorithms are also discussed.     

Passive optical remote sensing of cyanobacteria and other intense phytoplankton blooms in coastal and inland waters

Increased frequency and extent of potentially harmful blooms in coastal and inland waters world-wide require the development of methods for operative and reliable monitoring of the blooms over vast coastal areas and a large number of lakes. Remote sensing could provide the tool. An overview of the literature in this field suggests that operative monitoring of the extent of some types of blooms (i.e. cyanobacteria) is relatively straightforward. Operative monitoring of inland waters is currently limited to larger lakes or using airborne and hand-held remote sensing instruments as there are no satellite sensors with sufficient spatial resolution to provide daily coverage. Extremely high spatial and vertical variability in biomass during blooms of some phytoplankton species and the strong effects of this on the remote sensing signal suggest that water sampling techniques and strategies have to be redesigned for highly stratified bloom conditions, especially if the samples are collected for algorithm development and validation of remote sensing data. Comparing spectral signatures of different bloom-forming species with the spectral resolution available in most satellites and taking into account variability in optical properties of different water bodies suggests that developing global algorithms for recognizing and quantitative mapping of (harmful) algal blooms is questionable. On the other hand some authors cited in the present paper have found particular cases where satellites with coarse spectral and spatial resolution can be used to recognize phytoplankton blooms even at species level. Thus, the algorithms and methods to be used depend on the optical complexity of the water to which they will be applied. The aim of this paper is to summarize different methods and algorithms available in an attempt to assist in selecting the most appropriate method for a particular site and problem under investigation.     

Remote sensing of emergent and submerged wetlands: an overview

To plan for wetland protection and responsible coastal development, scientists and managers need to monitor changes in the coastal zone, as the sea level continues to rise and the coastal population keeps expanding. Advances in sensor design and data analysis techniques are now making remote-sensing systems practical and cost-effective for monitoring natural and human-induced coastal changes. Multispectral and hyperspectral imagers, light detection and ranging (lidar), and radar systems are available for mapping coastal marshes, submerged aquatic vegetation, coral reefs, beach profiles, algal blooms, and concentrations of suspended particles and dissolved substances in coastal waters. Since coastal ecosystems have high spatial complexity and temporal variability, they should be observed with high spatial, spectral, and temporal resolutions. New satellites, carrying sensors with fine spatial (0.4–4 m) or spectral (200 narrow bands) resolution, are now more accurately detecting changes in coastal wetland extent, ecosystem health, biological productivity, and habitat quality. Using airborne lidars, one can produce topographic and bathymetric maps, even in moderately turbid coastal waters. Imaging radars are sensitive to soil moisture and inundation and can detect hydrologic features beneath the vegetation canopy. Combining these techniques and using time-series of images enables scientists to study the health of coastal ecosystems and accurately determine long-term trends and short-term changes.     

A wavelet-based autoregressive fuzzy model for forecasting algal blooms

This paper proposes fuzzy models for forecasting the complex behavior of algal blooms. The models are developed through the integration of autoregressive models, the Takagi-Sugeno fuzzy model, and discrete wavelet transform algorithms. The premise parts of the proposed models are determined using the subtractive clustering technique and the consequent parts are optimized using weighted least squares. To train and validate the proposed fuzzy models, a large number of data sets were collected from Daecheong reservoir in Geum River in the Republic of Korea. The data include both water quality and hydrological variables. Total nitrogen, total phosphorous, dissolved oxygen, chemical oxygen demand, biochemical oxygen demand, pH, air temperature, water temperature and outflow water were evaluated as input signals while chlorophyll-a was used as an output. It is demonstrated from the simulation that the proposed fuzzy models are effective in forecasting algal blooms.     

Detection of algal blooms over optically complex waters of the Arabian Gulf and Sea of Oman using MODIS fluorescence data

The Arabian Gulf and the Sea of Oman are two of the most complex and turbid ecosystems in the world where algal blooms frequently occur. The conventional blue/green band ratio shows low performance to detect these algal batches in this region due to the effect of the non-algal parameters, shallow water depth, and atmospheric aerosols. Thus, an attempt to use MODIS (Moderate Resolution Imaging Spectroradiometer) fluorescence for the detection of algal blooms in this region have been undertaken using in situ measurements (Chlorophyll a: Chl-a, coloured dissolved organic matters: CDOM, Secchi disk depth: SDD, and radiometric) collected in 2006, 2013, and 2014, and MODIS satellite images. MODIS fluorescence line height (FLH in W m^?2 µm^?1 sr^?1) data showed low correlation (coefficient of determination: R² ~0.46) with near-concurrent in situ Chl-a (mg m^?3). This disparity is caused by the effect of the suspended sediments (SDD), CDOM (<2 mg m^?3 or >2 mg m^?3), and bottom reflectance (water depth: WD) parameters, where an increase of 1% in their magnitudes can cause a respective change of 13.4%, ?0.8% or 6%, and 1.4% in the FLH. In this work, the positions of the FLH bands have been relocated to include 645 nm to reduce the effect of these parameters on Chl-a, which has improved the performance to R² of 0.76. This modified FLH (MFLH) model was found to perform well in the Arabian Gulf where the estimated bias, root-mean-square error (RMSE), and coefficient of determination are, respectively, 0.03, 1.06, and 0.76. High values of MFLH are indicating the areas of the algal blooms, while no overestimation was observed in the mixed pixel coastal areas. This result is explained by less sensitivity of this model to the non-algal particles, shallow water, and aerosols.     

Satellite remote sensing of surface oceanic fronts in coastal waters off west-central Florida

Two algorithms designed to detect deepwater oceanic features and arbitrary edge profiles were tuned to automatically delineate fronts in coastal waters off west-central Florida using satellite-derived sea surface temperature (SST), chlorophyll-a concentration (Chl), normalized water-leaving radiance (nL_w), and fluorescence line height (FLH) images during select periods in the spring and fall of 2004 and 2005. The dates correspond to recreational king mackerel, Scomberomorus cavalla, tournaments. A histogram-based algorithm was useful to detect coastal surface SST, nL_w, and FLH fronts, specifically. A gradient-based algorithm, with a smaller kernel box of 3 × 3 pixels, best identified nearshore (< 10 m depth) features in Chl images at the mouth of Tampa Bay, but was less effective for fronts farther offshore where gradients were weaker. Local winds and tide levels estimated from a coastal observing buoy, and bathymetric gradients were examined to help understand the factors that influenced front formation and stability. Periods of strong and variable winds led to front movement of up to 10 km per day or dissipation within 2-3 days in over 80% of the fronts detected in SST, Chl, nL_w, and FLH imagery. Short episodes of less variable wind velocities typically led to more stable and stationary fronts, within 3-5 km, for up to four days. The occurrence of fronts closely associated with the coastal bathymetry, namely at the 20 m and 30 m isobaths, was significantly higher in the fall SST imagery and in the spring Chl imagery. Fall SST fronts related to bathymetric gradients likely resulted from progressive cooling of the water with depth. Stronger Chl and nL_w443 gradients at the mouths of estuaries in the fall compared to the spring were attributed to increased precipitation and periods of stronger winds or tides. The FLH imagery was most useful in delineating coastal algal blooms. The automatic front detection techniques applied here can be an important tool for resource managers to track coastal oceanographic features daily, over synoptic spatial scales.     

Fuzzy expert systems and GIS for cholera health risk prediction in southern Africa

Cholera (Vibrio cholerae) is endemic in southern Africa and frequently breaks out in epidemics along the eastern seaboard. Extensive resources are directed at combating cholera yet it remains a significant problem. Limited resources could better be directed to prevent outbreaks if it were possible to assess the risk of an outbreak in space and time. The CSIR in South Africa is investigating technologies to predict health risk in line with national priorities. This paper describes an early warning GIS prototype tool aimed at identifying favourable preconditions for cholera outbreaks. These preconditions were defined using an expert system approach. The variables thus identified were input into a spatial fuzzy logic model that outputs risks. The model is based on the assumption that endemic reservoirs of cholera occur and that environmental conditions, especially algal blooms, trigger Vibrio growth in the natural environment. If the preconditions are met, the subsequent spread of cholera depends mainly on socio-economic factors such as human behaviour and access to safe water supply and sanitation. This paper focuses on the environmental preconditions. The methodology described relies on capturing expert knowledge and historic data that integrate climatic and biophysical parameters with epidemiological data to produce a fuzzy surface of cholera outbreak risk potential.     

Prediction of daily maximum ozone threshold exceedances by preprocessing and ensemble artificial intelligence techniques: Case study of Hong Kong

The objective of this study was to apply preprocessing and ensemble artificial intelligence classifiers to forecast daily maximum ozone threshold exceedances in the Hong Kong area. Preprocessing methods, including over-sampling, under-sampling, and the synthetic minority over-sampling technique, were employed to address the imbalance data problem. Ensemble algorithms are proposed to improve the classifier's accuracy. Moreover, a distance-based regional data set was generated to capture ozone transportation characteristics. The results show that a combination of preprocessing methods and ensemble algorithms can effectively forecast ozone threshold exceedances. Furthermore, this study advises on the relative importance of the different variables for ozone pollution prediction and confirms that regional data facilitate better forecasting. The results of this research can be promoted by the Hong Kong authorities for improving the existing forecasting tools. Moreover, the results can facilitate researchers' selection of the appropriate techniques in their future research.     

Spatiotemporal pattern validation of chlorophyll-a concentrations in Lake Okeechobee,Florida, using a comparative MODIS image mining approach

A comparative analysis was conducted using three types of data-mining models produced from Moderate Resolution Imaging Spectroradiometer (MODIS) Terra Surface Reflectance 1-day or 8-day composite images to estimate chlorophyll-a (chl-a) concentrations in Lake Okeechobee, Florida. To understand the pros and cons of these three models, a genetic programming (GP) model was compared to an artificial neural network (ANN) model and multiple linear regression (MLR) model with respect to two different data sets related to model formulation. The first data set included the MODIS Terra bands from 1 to 7; the second data set extended the first data set by adding environmental parameters such as Secchi disc depth (SDD), total suspended solids (TSS), wind speed, water level, rainfall and air temperature collected around the lake in 2003 and 2004. The GP algorithm, which has an advantage in machine learning allowing us to select the appropriate input parameters that significantly impact the prediction accuracy, outperformed the other two models based on four statistical indices. Specifically, the GP modelling outputs revealed interesting determinations of chl-a concentrations for MODIS bands 3, 5, 6 and 7, corresponding to wavelengths 459–479, 1230–1250, 1628–1652 and 2105–2155 nm, respectively. The number of training data points is limited; therefore, the inclusion of additional environmental variables cannot improve the prediction accuracy of the GP-derived chl-a concentrations.     

Remote sensing of phytoplankton pigment distribution in the United States northeast coast

Phytoplankton pigments constitute many more compounds than chlorophyll a that can be applied to study phytoplankton diversity, populations, and primary production. In this study, field measurements were applied to develop ocean color satellite algorithms of phytoplankton pigments from in-water radiometry measurements. The match-up comparisons showed that the satellite-derived pigments from our algorithms agree reasonably well (e.g. 30-55% of uncertainty for SeaWiFS and 37-50% for MODIS-Aqua) to field data, with better agreement (e.g. 30-38% of uncertainty for SeaWiFS and 39-44% for MODIS-Aqua) for pigments abundant in diatoms. The seasonal and spatial variations of satellite-derived phytoplankton biomarker pigments, such as fucoxanthin, which is abundant in diatoms, peridinin, which is found only in peridinin-containing dinoflagellates, and zeaxanthin, which is primarily from cyanobacteria in coastal waters, revealed that higher densities of diatoms are more likely to occur on the inner shelf and during winter-spring and obscure other abundant phytoplankton groups. However, relatively higher densities of other phytoplankton, such as dinoflagellates and cyanobacteria, are likely to occur on the mid- to outer-continental shelf and during summer. Seasonal variation of riverine discharge may play an important role in stimulating algal blooms, in particular diatoms, while higher abundances of cyanobacteria coincide with warmer water temperatures and lower nutrient concentrations.     

COMPARING ANNs AND GENETIC PROGRAMMING FOR VOICE QUALITY ASSESSMENT POST-TREATMENT

In the U.K., the rehabilitation of a patient's voice following treatment for cancer of the larynx is managed by Speech and Language Therapists (SALT), who listen to a patient's stylized speech and then use their experience and domain knowledge to make an assessment of the current quality of the patient's voice. This process is very subjective and time consuming, and could benefit from using AI techniques to provide objective, reproducible assessments of voice quality. A comparative study of voice quality assessment post-treatment using Artificial Neural Networks (ANN), the preferred AI technique in this application area, and Genetic Programming (GP) is described, using the same dataset, training, and verification procedures. The GP approach was found to give more accurate classifications of bad quality (immediately post-treatment) and good quality (recovered) voicings than the ANN, and in addition, gave indication of the most significant parameters in the input dataset.     

Development and application of a neural network based ocean colour algorithm in coastal waters

An algorithm for determining chlorophyll‐a concentrations in shallow, case II waters has been developed and applied to nearly six years of Sea‐viewing Wide Field‐of‐view Sensor (SeaWiFS) data in order to observe the general chlorophyll‐a patterns in a coastal estuarine environment. Due to the fact that the current empirical chlorophyll‐a algorithm (OC4) used to process SeaWiFS data breaks down in coastal waters, a neural network based algorithm was developed. The neural network in the study uses SeaWiFS remote sensing reflectance data paired with in situ chlorophyll‐a data in the Delaware Bay and its adjacent coastal zone (DBAC) from a number of different days and seasons in an effort to overcome the limitations of single day algorithms and simulated dataset algorithms. Although the neural network model (NN) in this study displayed some difficulty representing high chlorophyll‐a values, it showed significant improvement over the OC4 algorithm. The performance parameters of the NN were an r ² of 0.79, a root mean square (RMS) error of 3.69?mg m^?3 and a relative RMS error of 0.77. The NN was used to reprocess approximately six years of cloud free imagery of the DBAC from which the spatial and temporal variability of the chlorophyll‐a distributions in the DBAC were analysed. Time series of absolute chlorophyll‐a values for five stations along the central axis of the Delaware Bay were analysed using Fourier analysis techniques, from which chlorophyll‐a patterns were found to have a quasi‐annual period. Furthermore, the spatial distributions of the chlorophyll‐a patterns were analysed using a general climatology and monthly climatologies of normalized chlorophyll‐a values. The climatologies generally agreed with spatial distributions determined from historic ship‐based data. The study found that summer blooms in the mid‐estuary of the Delaware Bay may be more important than previously observed. This suggests that more frequent and synoptic measurements via satellite can reveal important new information about even well studied regions.     

Detection of algal blooms in European waters based on satellite chlorophyll data from MERIS and MODIS

A technique for algal-bloom detection in European waters is described, based on standard chlorophyll a concentration (Chl) data from two ocean-colour sensors, the Moderate Resolution Imaging Spectroradiometer (MODIS) and Medium Resolution Imaging Spectrometer (MERIS). Comparison of the two data sources shows good agreement in case 1 waters, whereas the difference is significant in coastal waters including turbid areas. A relationship between the water-leaving reflectance at 667 nm and Chl for case 1 waters was used to eliminate pixels where Chl retrieval is contaminated by backscatter from inorganic suspended matter. Daily Chl data are compared to a predefined threshold map to determine whether an algal bloom has occurred. In this study, a threshold map was defined as the 90th percentile of previous years' data to take account of regional differences in typical Chl levels, with separate maps for each sensor to take account of sensor-specific bias. The algal-bloom detection processing chain is described, and example results are presented.     

Seasonal and spatial distribution of chlorophyll-a concentrations and water conditions in the Gulf of Tonkin, South China Sea

The Gulf of Tonkin is a semi-closed gulf northwest of the South China Sea, experiencing reversal seasonal monsoon. Previous studies of water conditions have been conducted in the western waters of the gulf, but very few studies of the Chlorophyll-a (Chl-a) distribution have been carried out for the entire gulf. The present study investigates seasonal and spatial distributions of Chl-a and water conditions in the Gulf of Tonkin by analyzing Sea-viewing Wide Field-of-View Scanner (SeaWiFS) derived Chlorophyll-a (Chl-a), in situ measurements, sea surface temperatures (SST), and other oceanographic data obtained in 1999 and 2000. The results show seasonality of Chl-a and SST variations in the Gulf of Tonkin, and reveal phytoplankton blooming events in the center part of the gulf during the northeast monsoon season. In summer, Chl-a concentrations were relatively low (<0.3 mg m⁻³) and distributed uniformly throughout most of the area, with a belt of higher Chl-a concentrations along the coast, particularly the coast of Qiongzhou Peninsula; in winter, Chl-a concentration increased (0.5 mg m⁻³) in the entire gulf, and phytoplankton blooms offshore-ward from the northeast coast to the center of the gulf, while Chl-a concentrations reached high levels (0.8-1 mg m⁻³) in the center of the blooms. One peak of Chl-a concentrations was observed during the northeast monsoon season in the year. SST were high (27-29 °C) and distributed uniformly in summer, but lower with a large gradient from northeast (17 °C) to southwest (25 °C) in winter, while strong northeast winds (8-10 m/s) were parallel to the east coast of the gulf. Comparison of Chl-a values shows that SeaWiFS derived Chl-a concentrations match well with in situ measurements in most parts of the gulf in May 1999, but SeaWiFS derived Chl-a are higher than in situ data in river mouth waters. The seasonal variation of Chl-a concentrations and SST distribution were associated with the seasonally reversing monsoon; the winter phytoplankton blooms were related to vertical mixing and upwelling nutrients drawn by the northeast wind.     

The detection and mapping of algal blooms from space

Algal blooms change the colour of water through absorption by pigments and scattering by cells and associated detrital material. This paper gives a brief introduction, primarily using ocean colour imagery, to the use of satellite Earth observation measurements in detecting and mapping algal blooms. Two examples, using Sea-Viewing Wide Field-of-View Sensor (SeaWiFS) imagery, show blooms within northern European waters.     

Chlorophyll-a Predicting Model Based on Dynamic Neural Network

Algal blooms are one of the most prevalent global problems. Studying the Chlorophyll-a (Chl-a) predicting model helps to control algal blooms. Predicting the behavior of algae is difficult because of the complex physical, chemical, and biological processes involved. Artificial neural network (ANN) models have been determined to be useful and efficient, especially for such problems for which the characteristics of the processes are difficult to describe using numerical models. An indoor simulated environment is designed for algal cultivation to analyze the temporal change in the algae biomass of Taihu Lake during summer. A Chl-a prediction model based on a nonlinear autoregressive neural network with exogenous inputs (NARX) that can detect and consider within the time dependency is proposed. The NARX model is compared to a static neural network and a dynamic neural network: feedforward neural network (FNN) and Elman recurrent neural network (ERNN). The performance of the proposed NARX model was examined with experimental data collected over 3 months in 2010. The results showed that the NARX model outperformed the other ANN models and significantly enhance the accuracy of Chl-a prediction.