| 基于国产高分一号卫星数据的区域土壤盐渍化信息提取与建模 |
| |
| 引用本文： | 曹雷,丁建丽,玉米提·哈力克,苏雯,宁娟,缪琛,李焕.基于国产高分一号卫星数据的区域土壤盐渍化信息提取与建模[J].土壤学报,2016(6):1399-1409. |
| |
| 作者姓名： | 曹雷 丁建丽 玉米提·哈力克 苏雯 宁娟 缪琛 李焕 |
| |
| 作者单位： | 1. 新疆大学资源与环境科学学院，乌鲁木齐 830046; 新疆大学绿洲生态教育部重点实验室，乌鲁木齐 830046; 德国埃希施塔特-因戈尔施塔特大学数学与地理学院，埃希施塔特 85071;2. 新疆大学资源与环境科学学院，乌鲁木齐 830046; 新疆大学绿洲生态教育部重点实验室，乌鲁木齐 830046 |
| |
| 基金项目： | 新疆维吾尔自治区科技支疆项目(201504051064)，高分辨率对地观测重大专项(民用部分)(95-Y40B02-9001-13/15-03-01)，国家留学基金委创新型人才培养国际合作项目(201505990312)资助 Supported by the Science and Technology Supporting Project of Xinjiang Uyghur Autonomous Region(201504051064)，the High Resolution Earth Observation Projects(Civil Part)(95-Y40B02-9001-13/1-03-01)，the Innovative Elite Training in International Cooperation Projects of China Scholarship Council(201505990312) |
| |
| 摘 要： | 土壤盐渍化是干旱半干旱地区土地退化的主要原因之一,给当地生态系统和社会经济的可持续发展带来了严重的威胁,而对盐渍化空间分布信息的提取是治理盐渍化的基础。因此,选取生态脆弱区渭—库绿洲为研究区,利用2014年7月19日GF-1多光谱影像数据,提取光谱指数及波段信息,结合实际采样点的土壤表层电导率数据(0~10 cm),采用偏最小二乘回归模型(partial least squares regression,PLSR)对土壤盐渍化进行模拟,并对研究区盐渍化分布进行模拟和评估。结果表明:实测土壤表层电导率与光谱指数相关性较好;利用PLSR对渭—库绿洲土壤表层盐渍信息建模,对土壤盐渍化信息提取效果较好,精度较高;充分利用了GF-1影像包含的信息,提高了高分辨率遥感影像盐渍化信息提取的精度;非盐渍化和轻度盐渍化面积分别占总面积的42.88%和17.16%,绿洲中部偏东及东南区域,盐渍化现象稍弱,可成为今后绿洲扩张的重点方向;而中度盐渍化、重度盐渍化和盐土面积分别占总面积的29.51%、8.57%和1.88%,绿洲北部/西部及西南方向的重度盐渍化区域紧挨绿洲区域,已严重威胁了绿洲经济的健康发展,亟待治理。
|
| 关 键 词： | 土壤盐渍化 高分一号 光谱指数 偏最小二乘回归法(PLSR) 渭-库绿洲 |
Extraction and Modeling of Regional Soil Salinization Based on Data from GF-1 Satellite |
| |
| Abstract: | AbstractObjective]Soil salinization,being one of the main causes of land degradation in arid and semi-arid regions,poses a great threat to sustainable development of the local social economy and ecological system.Method]How to extract the information of spatial distribution of soil salinization is the foundation for management of soil salinization. Therefore,the Weigan-Kuqa Oasis,an area rather fragile in ecology,was selected as an object in this study,using the GF-1 satellite multispectral images of the date of July 19,2014 as the main data source. A total of 16 spectral indices i.e. Normalized difference vegetation index(NDVI),soil adjusted vegetation index(SAVI),normalized differential salinity index (NDSI),salinity index(SI-T),brightness index(BI),salinity index(SI),salinity index 1 (SI1),salinity index 2(SI2),salinity index 3(SI3),salinity index(S1),salinity index(S2), salinity index(S3),salinity index(S5),salinity index(S6),intensity index 1(Int1),intensity index 2(Int2),and four bands,i.e. band 1(B1),band 2(B2),band 3(B3)and band 4(B4), were chosen for analysis. The images in pretreatment were computed by band in line with the spectral index formulas with the aid of software ENVI4.8. Hence,gray scale maps of different spectral indices were derived and pixel values of the 36 sampling points corresponding to the gray scale maps were extracted. The data of electrical conductivities in the surface soil layers(0~10 cm)of those sampling sites during 22~28 July 2014 were also collected for analysis of Pearson correlation with the pixel values using software SPSS 19.0. Thus sensitivities of different spectral indices to the data of soil salinization were figured out. PLSR models were built and validated for relationships of the mathematical formulas for five different electrical conductivities(i.e. measured conductivity,reciprocal of measured conductivity,logarithm of measured conductivity,MSR of measured conductivity and reciprocal of the logarithm of measured conductivity)with spectral indices. Measured conductivities of 24 samples were used for modeling and the remaining 12 samples for validation with the aid of the Unscrambler X10.3 software.Result]Results show:1)the measured surface soil conductivities are closely related to spectral indices,and moderately to SAVI,NDVI,NDSI, SI,SI1,S5,B3 and B4,with all correlations being significant at the 0.01 level;2)based on the GF-1 satellite images PLSR was used for modeling of surface soil salinization in the Weigan-Kuqa Oasis. The model based on reciprocal of electrical conductivity is better than all the others withR2=0.69 and RMSE=0.58 dS m-1. For the validation modelR2 is 0.78 and RMSE 0.53 dS m-1. In the images of satellite vertical projection salinized and non salinized patches in the land cover can be clearly distiquished from each other,with little confusing or mistaking information. Characteristic texture of salinization relative to degree is distinct, showing a clear layered structure,easy to distinquish and making the visual interpretation of the images more consistent with the actual degree of soil salinization. Consequently the effect of information extraction of soil salinization is quite good and high in precision;3)this study made full use of the information contained in GF-1 images,thus improving precision of the extraction of soil salinization information from GF-1 images. Non-salinized soil and slightly salinized soil in the oasis accounts for 42.88% and 17.16%,respectively,of total in area. Soil salinization is quite mild in the middle by east and southeast of the oasis,which suggests that the oasis may expand towards that direction in future. Moderately salinized soil,severely salinized soil and salinized soil in the oasis accounts for 29.51%,8.57% and 1.88%,respectively,of the total in area. The severely salinized soil is distributed closely on the north/west and southwest of the oasis and has already posed a serious threat to healthy development of the oasis economy and calls urgently for management.Conclusion]The study on use of GF-1 images to evaluate soil salinization may provide a scientific basis for prediction of salinization in future and assessment of the current situation and for the government in decision-making for management of soil salinization. |
| |
| Keywords: | Soil salinization GF-1 Spectral indices PLSR Weigan-Kuqa oasis |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
| 点击此处可从《土壤学报》浏览原始摘要信息 |
|
点击此处可从《土壤学报》下载全文 |
|
