Studies in Surveying and Mapping Science (SSMS) Volume 3, 2015
www.as-se.org/ssms
Neural Network Retrieval Model of Chlorophyll-a Concentration in Bohai Bay Huilin Ye*1, Jingqin Mu2, Guoqing Yao3 College of Information Engineering, China University of Geosciences(Beijing),Beijing,China;
1,3
College of Computer Science, Tangshan Normal University, Tangshan Hebei,China
2
hlye@cugb.edu.cn; 2mujingqin@163.com; 3gqyao@cugb.edu.cn
*1
Abstract Based on in-situ measurements of chlorophyll-a concentration and ETM images of Bohai bay, a neural network retrieval model is built up in this paper. According to the spectral characteristics of Bohai bay coastal water, analysis is conducted to decide the correlation between the reflectance of the first 4 bands of ETM and chlorophyll-a concentration. The BP neural network model is constructed, consisting of 4 input nodes (the reflectance of the first 4 bands of ETM), 8 hidden nodes and 1 output node, by comparing the performance of the network with different numbers of hidden nodes. The result indicates that the BP neural network model has a high accuracy with R2 achieved 0.956 and the BP neural network model provides a reliable basis and method for monitoring chlorophyll-a concentration in water by remote sensing technology quickly and accurately. Keywords Bohai Bay; Chlorophyll-a Concentration; BP Neural Network Model
Introduction Chlorophyll-a concentration in water is intimately linked with the amounts and types of algae, so chlorophyll-a concentration is a composite indicator to reflect the phytoplankton biomass in water, and also is a vital parameter to evaluate the quality of sea water, organic pollution levels and fishery resources. Detecting, analyzing and simulating chlorophyll-a concentration and its space-time distribution characteristics accurately have important research significances in grasping the nearshore marine fishery resources, monitoring the red tides and evaluating eutrophication condition. Traditional determination method of chlorophyll-a, mainly depending on underway sampling by survey vessels and measuring in the lab, is costly and time-consuming. And sampling point by point dispersively cannot meet the demands of measuring continuously and synchronously for a long range and a long time. Also it is difficult to characterize the distribution of water quality indicators in space and time change. Using remote sensing to estimate the chlorophyll-a concentration in sea can make up for the defects of common methods. Since the 1970s, kinds of analysis methods based on remote sensing data have been proposed by scholars at home and abroad. Chen, She, Lv, Li, Li and Wu have respectively built regression models for retrieving chlorophyll-a concentration using remote sensing images and in-situ measurements while Zhan, Wu, Zhao, Wang, Shen and He have established retrieval models of chlorophyll-a concentration in lakes, reservoirs or even seas by neural network technology. According to the theoretical foundation and the amount of measured data, the methods of retrieving water quality parameters by remote sensing data mainly fall into 3 classes: statistical methods, semi-empirical and semi-analytical methods and theoretical methods. Study Area The Bohai sea is the largest semi-enclosed inland sea in China, located in 37°7´-41°0´ north latitude, 117°35´-121°10´ east longitude, with a total area of 77000 km2 and the average depth of 18m. It is mainly composed of Liaodong bay, Bohai bay and Laizhou bay. Bohai bay is a shallow sea basin surrounded by lands, mostly silt and soft clay. The seabed terrain inclines towards the center of the bay. Plenty of suspended sand is injected while Luanhe river, Haihe river, the Yellow River and Xiaoqinghe river feed into the bay. The sediment mostly is silt and silty sand silt. Due to the increasing economic activities in the Bohai Rim in recent years, the water ecosystem of the Bohai sea has been seriously affected by accepting a large number of terrigenous pollutant every year. Because of its semi31