| 基于随机森林方法反演墨西哥湾海表盐度 |
| |
| 引用本文： | 吴芳芳,傅智一,胡林舒,张丰,杜震洪,刘仁义.基于随机森林方法反演墨西哥湾海表盐度[J].海洋学报,2021,43(9):126-136. |
| |
| 作者姓名： | 吴芳芳 傅智一 胡林舒 张丰 杜震洪 刘仁义 |
| |
| 作者单位： | 1.浙江大学 浙江省资源与环境信息系统重点实验室，浙江 杭州 310028 |
| |
| 基金项目： | 国家重点研发计划重点专项（2018YFB0505000）；国家自然科学基金（41671391，41922043，41871287） |
| |
| 摘 要： | 盐度是表征物理和生物地球化学过程的重要参数之一，光学遥感可满足较高分辨率的监测需要并避免射频干扰问题，为沿海水域的海表盐度研究提供可行的途径。本文基于MODIS-Aqua的412 nm、443 nm、488 nm、555 nm和667 nm波段的遥感反射率(Rrs412、Rrs443、Rrs488、Rrs555、Rrs667)、海表温度以及实测的海表盐度数据构建随机森林模型，基于模型结果分析墨西哥湾海表盐度时空异质性及海表盐度与影响因子(海表温度和遥感反射率)之间的相关关系。研究结果表明:(1)随机森林模型能较准确地反演墨西哥湾海表盐度，其均方根误差为0.335，决定系数为0.931；(2)湾区海表盐度空间分布呈近岸?河口低、离岸高，环状向内增值的态势，其变化受河流流量、风力以及环流的影响；(3)海表温度与海表盐度存在较强的相关性，海表温度对海表盐度的反演影响显著；(4)海表温度、遥感反射率与海表盐度的相关性呈现空间异质性。
|
| 关 键 词： | 海表盐度 随机森林 时空异质性 墨西哥湾 |
| 收稿时间： | 2021-02-03 |
Retrieval of sea surface salinity in the Gulf of Mexico based on random forest method |
| |
| Institution: | 1.Zhejiang Provincial Key Laboratory of Geographic Information Science, Zhejiang University, Hangzhou 310028, China2.School of Earth Sciences, Zhejiang University, Hangzhou 310027, China |
| |
| Abstract: | Salinity is an important parameter to characterize physical and biogeo-chemical processes. Optical satellite images with high resolution can avoid radio frequency interference, and provide a feasible way to monitor sea surface salinity (SSS) in coastal regions. Using an extensive dataset of ship-based SSS and MODIS estimated remote sensing reflectance (Rrs) at 412 nm, 443 nm, 488 nm, 555 nm and 667 nm and sea surface temperature (SST) a random forest (RF) model has been utilized to retrieve the SSS. Based on the predicted SSS, we analyze the spatiotemporal heterogeneity of SSS in the Gulf of Mexico and contribution of each factor with correlations coefficient. The results show that: (1) the RF model can accurately estimate the SSS in the Gulf of Mexico (RMSE=0.335, R2=0.931); (2) the spatial distribution pattern of SSS shows a ring-shaped inward value increase, which is affected by river discharge, wind forcing and circulation; (3) there is a strong correlation between SSS and SST, and SST significantly impact in retrieving SSS; (4) the correlation between SST, Rrs and SSS appears spatial heterogeneity. |
| |
| Keywords: | |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《海洋学报》浏览原始摘要信息 |
| 点击此处可从《海洋学报》下载免费的PDF全文 |
|
