基于GIS的西北太平洋柔鱼资源空间插值及不确定性分析

Spatial interpolation and uncertainty of neon flying squid(Ommastrephes bartramii) resources in the Northwest Pacific Ocean

空间插值方法在渔业资源领域应用广泛,但插值中涉及的半变异函数选择和插值结果评价并未得到充分阐释。以2009年和2010年西北太平洋柔鱼资源为例,探讨普通克里金插值方法在渔业资源中的应用,重点分析最优半变异函数的判别和插值结果的空间不确定性。分析认为,GS+的决定系数和残差项一定程度上可以识别最优函数(模型);从交叉检验衍生的拟合优度和一致率,同样可以用于最优函数选择,也能分析插值结果的空间不确定性。研究表明,2009年最优半变异函数为球形模型,2010年为指数模型。此外,通过图对一致性和Kappa系数,对各种模型的插值结果进行了两两比较,结果显示2009年球形模型和指数模型的结果最相似,2010年球形模型和高斯模型最相似。

Spatial interpolation is widely used in the field of fisheries resources. In most previous studies the selection of semivariance and the uncertainty assessment of results were not sufficiently investigated. Here, we discuss the application of spatial interpolation to fisheries resources using neon flying squid （Ommastrephes bartramii）resources in the northwest Pacific Ocean in 2009 and 2010. Our focus is the selection of semivariance and the spatial uncertainty assessment of results based on ordinary Kriging. The optimal semi-variance is often estimated using determination coefficient and residual error in GS ＋ software. We propose a method using goodness-of-fit and map agreement derived from cross validation of ArcGIS to select an optimal semivariance model. Moreover, spatial uncertainty was assessed using cross validation of ArcGIS. We found that spatial patterns in O. bartramii resources were different across various methods （semi-variances） , and optimal semi-variances were spherical and exponential models in 2009 and 2010, respectively. Additionally, maps agreement and kappa were used to compare two maps produced from spatial interpolation. The spherical and Gaussian models were not suitable for the spatial interpolation in 2010. The analysis shows that the results generated spherical and exponential models similar to each other in 2009, while the results generated spherical and Gaussian models similar to each other in 2010. The effects of spatial scale on the spatial interpolation of marine fisheries resources are discussed. Fine scales are able to represent spatial patterns and characters of the original dataset; the spatial scale 0.5°× 0.5° is suitable for the investigation of spatial patterns in fisheries resources.