Mapping of Sediment Grain Size Using Remote Sensing Data and Kriging Algorithm

详细信息    查看全文 | 下载全文 | 推荐本文 |

• 作者：Dong-Ho JANGNo Wook PARKSu Joong NAM
• 会议时间：2009-12-01
• 作者单位：Dong-Ho JANG(Department of Geography,Kongju National University,Kongju-si,Chungchungnam-do 314-701,Korea)No Wook PARK(Department of Geoinformatic Engineering,Inha University,253 Yonghyun-dong,Nam-gu,Incheon 402-751,Korea)Su Joong NAM(Department of International Trade and Commerce,Kongju National University,Kongju-si,Chungchungnam-do 314-701,Korea)
• 母体文献：第4届中日韩地理学家学术研讨会暨第12次全国青年地理工作者学术研讨会论文集
• 会议名称：第4届中日韩地理学家学术研讨会暨第12次全国青年地理工作者学术研讨会
• 会议地点：广州
• 主办单位：中国地理学会
• 语种：chi

摘要

This paper has demonstrated the efficiency of IKONOS imagery and multivariate geostatistical kriging for the high-resolution mapping of sediment grain size.The presented technique used high-resolution remote-sensing imagery as local means for multi-Gaussian kriging.Unlike traditional calibration approaches such as multiple linear regression,GAM was employed as a nonlinear calibration method.Experimental results from a case study of Baramarae beach,Korea,showed that accounting for high spatial resolution of IKONOS imagery via SKlm with MR or GAM,despite its poor spectral resolution,greatly improved the prediction accuracy.The integration of IKONOS imagery could reflect the detailed local variation of grain size distribution in the study area,as compared with traditional univariate multi-Gaussian kriging based on only sparsely sampled ground data.SKlm with GAM was the best prediction model in terms of prediction error based on cross validation.GAM,which was able to fit nonlinear relationships between ground-based grain size values and input IKONOS bands,showed a superior predictive capability to that of MR as the local mean estimator.From a methodological viewpoint,GAM,as applied in this study,has great potential as a local mean estimator in geostatistical analysis.Within the geostatistical indicator framework (Journel 1983),original variability of interest is first coded into a set of indicator variables with two values (i.e.,0 and 1).GAM can be used as the local mean estimator when it is used as a binary dependent variable as in the indicator approach.It also has an advantage in dealing with simultaneously categorical data as well as continuous data.The potential benefits of geostatistical integration of remote sensing imagery for any environmental thematic mapping can be enhanced when the following issues are considered in future work.In this study,there was a one-year mismatch in timing between the acquisition date of IKONOS imagery and field survey.Although this time inconsistency did not affect the performance of geostatistical analysis,a recent simultaneous acquisition of remote sensing imagery and field survey data is needed to reflect the dynamic characteristics of a sediment environment in the study area since the time considered (i.e.,the year 2001).Due to the short exposure time of tidal flats,it may be difficult to obtain both remote sensing imagery and field observation data at the same time.The second Korea Multi-purpose Satellite (KOMPSAT-2),which was launched in July 2006 and is Korea’s own satellite,can be a useful high-resolution information source for monitoring the national peninsula coastal area.High spectral resolution data,as well as high spatial resolution,should also be considered in order to fully derive the quantitative relationships between field data and each spectral band.Because the study area is small,the use of lower spatial resolution data with more spectral bands such as ASTER is not feasible.Airborne hyperspectral data can be a potential alternative.Finally,the consideration on additional environmental factors such as surface moisture content and topography is also required to fully reveal intertidal sedimentary characteristics.