资源科学 ›› 2018, Vol. 40 ›› Issue (10): 2110-2117.doi: 10.18402/resci.2018.10.18

• • 上一篇    下一篇

基于PALSAR雷达数据的于田绿洲土壤盐渍化反演

再屯古丽·亚库普1,2(), 买买提·沙吾提1,2(), 阿卜杜萨拉木·阿布都加帕尔1,2, 张东1,2   

  1. 1. 新疆大学资源与环境科学学院,乌鲁木齐 830046
    2. 新疆大学绿洲生态教育部重点实验室,乌鲁木齐 830046
  • 收稿日期:2017-06-12 修回日期:2018-05-28 出版日期:2018-10-25 发布日期:2018-10-20
  • 作者简介:

    作者简介:再屯古丽·亚库普,女,新疆轮胎县人,硕士生,主要从事干旱区资源与环境遥感应用研究。E-mail: zaytungul1992@163.com

  • 基金资助:
    国家自然科学基金项目(41361016;41561089)

Soil salinity inversion in Yutian Oasis based on PALSAR radar data

Yakup ZAYTUNGUL1,2(), Sawut MAMAT1,2(), Abdujappar ABDUSALAM1,2, Dong ZHANG1,2   

  1. 1. College of Resources and Environmental Science, Xinjiang University, Urumqi 830046, China
    2. Ministry of Education Key Laboratories of Oasis Ecology, Xinjiang University, Urumqi 830046, China
  • Received:2017-06-12 Revised:2018-05-28 Online:2018-10-25 Published:2018-10-20

摘要:

土壤盐渍化是当今土地退化和荒漠化的主要形式之一,不仅严重制约农业和经济的发展,并且对生态环境和人类生存造成威胁。本研究以新疆于田绿洲为研究区,利用四极化PALSAR (Phased Array type L-band Synthetic Aperture Radar)数据后向散射系数,土壤含水量,土壤pH值以及土壤盐分实测值,采用多元线性回归模型、地理加权回归模型和BP神经网络模型,以土壤含盐量作为因变量建立了定量反演模型。从土壤盐分反演结果图可以看出,反演结果与地面实地考察结果基本一致。经过模型验证得知,3层BPANN模型的均方根误差RMSE=0.99,平均相对误差MRE=0.31,模型性能指数RPD=5.34,其模型预测能力优于前2种传统模型。本文建立的神经网络模型无需考虑复杂的介电常数,在一定程度上能够满足土壤盐渍化监测的需要,促进PALSAR数据在土壤盐渍化监测中的应用。

关键词: 土壤盐渍化, PALSAR-2雷达数据, 后向散射系数, 神经网络, 反演, 新疆于田绿洲

Abstract:

Soil salinization and/or desertification is one of the main forms of land degradation and environmental issues. Meanwhile, it causes the destruction of resources, hampers to development of agriculture and threats to the environment and human survival. Yutian Oasis was identified as a study area and soil salinity information was extracted from the PALSAR-2 ALOS-2 data, which exhibited a kind of fine four-polarization SLC (single look complex) format and were bought in 2015 with 5.1 (range resolution)×4.3 (azimuth resolution) ground resolutions. Considering the distribution of saline soil spatial variability, 68 points were designed as sampling points, Hand-held GPS (global position system) receiver was used to record the coordinates of sampling points and 0-10 cm topsoil samples were collected in the field. Soil total soluble salt content was measured in the lab. The four-polarization back-scatter coefficient values corresponding to the sampling points were extracted based on the previous results by the spatial analysis module of ArcGIS. Total salt content was taken as dependent variable and four-polarization PALSAR-2 data back-scatter coefficient values, soil moisture and pH values as independent variables. The multiple linear regression (MLR), geographically weighted regression (GWR) and back propagation artificial neural network (BP ANN) were adopted to establish the quantitative inversion models of soil salt content. Results illustrated that among the ANN (BP), MLR and GWR models employed in this contribution, the ANN (BP) model was identified as the most potential predictive model of soil salinity. Best predictive results were achieved using ANN (BP) with R2=0.84, RMSE=0.99, MRE=0.31 and RPD=5.34. The established ANN (BP) model in this paper can reduce the smoothing effect compared with the two traditional models and improve the accuracy and reliability of model predictions, which meets the needs of soil salinity monitoring to a certain extent. It can promote and develop the application of microwave remote sensing in the soil salinity monitoring.

Key words: soil salinity, PALSAR-2 data, back-scatter coefficient, artificial neural network, inversion, Yutian Xinjiang