基于不同地表曲面模型预测土壤有机碳含量

Local terrain attributes,which are derived directly from the digital elevation model,have been widely applied in digital soil mapping.This study aimed to evaluate the mapping accuracy of soil organic carbon (SOC) concentration in 2 zones of the Heihe River in China,by combining prediction methods with local terrain attributes derived from different polynomial models.The prediction accuracy was used as a benchmark for those who may be more concerned with how accurately the variability of soil properties is modeled in practice,rather than how morphometric variables and their geomorphologic interpretations are understood and calculated.In this study,2 neighborhood types (square and circular) and 6 representative algorithms (Evans-Young,Horn,Zevenbergen-Thorne,Shary,Shi,and Florinsky algorithms) were applied.In general,35 combinations of first-and second-order derivatives were produced as candidate predictors for soil mapping using two mapping methods (i.e.,kriging with an external drift and geographically weighted regression).The results showed that appropriate local terrain attribute algorithms could better capture the spatial variation of SOC concentration in a region where soil properties are strongly influenced by the topography.Among the different combinations of first-and second-order derivatives used,there was a best combination with a more accurate estimate.For different prediction methods,the relative improvement in the two zones varied between 0.30％ and 9.68％.The SOC maps resulting from the higher-order algorithms (Zevenbergen-Thorne and Florinsky) yielded less interpolation errors.Therefore,it was concluded that the performance of predictive methods,which incorporated auxiliary variables,could be improved by attempting different terrain analysis algorithms.     

Application of the Tweedie distribution to zero-catch data in CPUE analysis

We focus on the zero-catch problem of CPUE (catch per unit effort) standardization. Because the traditional CPUE model with a log-normal error structure cannot be applied in this case, three methods have often been utilized as follows:

(1) Ad hoc method adds a small constant value to all response variables.

(2) Catch model with a Poisson or negative-binomial (NB) error structure.

(3) Delta-type two-step method such as the delta-normal model (after estimating the ratio of zero-catch using a logit or probit model, a model such as CPUE log-normal or Catch-Poisson is applied to CPUE without zero-data).

However, there are some statistical problems with each of these methods.In this paper, we carried out the CPUE standardization mainly using the Tweedie distribution model based on the actual by-catch data (silky shark, Carcharhimus falciformis, in the North Pacific Ocean caught by Japanese training vessels) including many observations with zero-catch (>2/3rd) and tuna fishery data as a target (yellowfin tuna, Thunnus albacares, in the Indian Ocean caught by Japanese commercial vessels) where the ratio of zero-catch is not so high (<1/3rd). The Tweedie model is an extension of compound Poisson model derived from the stochastic process where the weight of the counted objects (i.e., number of fish) has a gamma distribution and has an advantage of handling the zero-catch data in a unified way.We also compared four candidate models, the Catch-NB model, ad hoc method, Delta-lognormal model (delta-type two-step method) and Tweedie distribution, through CPUE analyses of actual fishery data in terms of the statistical performance. Square error and Pearson's correlation coefficient were calculated based on the observed CPUE and the corresponding predicted CPUE using the n-fold cross-validation.As a result, the differences in the trend of CPUE between years and model performance between the ad hoc method and Tweedie model were found to be not so large in the example of yellowfin tuna (target species). However, the statistical performance of Tweedie distribution is rather better than Delta-lognormal model, the Catch-NB distribution and ad hoc method in the example of silky shark (by-catch species). Standardized CPUE year trend of ad hoc method was found to be quite different from that of the Tweedie distribution and other two models. Model performance of the Tweedie distribution is good judging from the 5-fold cross-validation using the fishery data if including many zero-catch data such as by-catch species.     

蔬菜地土壤重金属含量空间插值精度分析

蔬菜地土壤重金属含量影响环境安全和人类健康。采用地统计分析和GIS相结合的方法,对研究区土壤重金属镉、汞、铬、砷、铅进行空间插值和精度分析,根据交叉检验参数和相对误差指标比较普通克里格法和协同克里格法的空间预测效果。结果表明：研究区土壤重金属存在中强等变异性,在重金属间显著相关性条件下,普通克里格法对汞、镉、砷含量预测精度较高;协同克里格法对铬和铅含量预测精度较高。建议根据土壤重金属空间变异程度选择合适的空间插值方法。     

距离倒数和克里克方法在土壤养分制图中的比较研究

空间插值方法以及模型参数的选择是制约插值精度的重要因素。为此,使用普通克里克、普通对数克里克和距离倒数加权方法对三峡库区王家沟小流域内的土壤养分(有机质、全氮、全磷)进行插值,研究了权重指数、邻域样本数目和变异函数模型对插值结果的影响,并用交叉验证的方法对这几种插值方法进行了比较。结果显示,距离倒数加权对于土壤有机质和全氮制图具有最高的精度,对数克里克对于全磷的模拟则是最优的。     

安徽省热量资源空间分布模型及插值方法研究

【目的】对安徽省热量资源进行栅格化处理,解决无测站地区的热量资源估算问题,为区域农业气候资源评估提供依据。【方法】利用安徽省1971~2000年气象站地面资料,在地理信息系统技术(GIS)支持下,采用多元回归结合空间插值方法,建立安徽省热量资源要素空间分布模型,并比较反距离权重法(IDW)、克里格和样条函数3种空间插值方法的精度。【结果】基于GIS平台,建立了安徽省250 m×250 m分辨率的温度和积温栅格化数字地图。对空间插值方法的交叉验证结果表明,3种空间插值方法的平均绝对误差、平均误差平方的平方根排序均为IDW<克里格法<样条函数法,IDW法效果最优。【结论】应用GIS空间内插技术,构建高分辨率的栅格化热量资源要素集,可有效提高农业气候资源分析和利用的精细化程度。     

基于土壤传递函数的土壤持水特性间接估计及其空间变异性研究

A total of 107 soil samples were taken from the city of Qingdao,Shandong Province,China.Soil water retention data at 2.5,6,10,33,100,300,and 1 500 kPa matric potentials were measured using a pressure membrane apparatus.Multiple linear regression (MLR) was used to develop pedotransfer functions (PTFs) for single point estimation and van Genuchten parameter estimation based on readily measurable soil properties,i.e.,MLR-based point (MLRP) PTF and MLR-based parametric (MLRV) PTF.The double cross-validation method was used to evaluate the accuracy of PTF estimates and the stability of the PTFs developed in this study.The performance of MLRP and MLRV PTFs in estimating water contents at matric potentials of 10,33,and 1 500 kPa was compared with that of two existing PTFs,the Rawls PTF and the Vereecken PTF.In addition,geostatistical analyses were conducted to assess the capabilities of these PTFs in describing the spatial variability of soil water retention characteristics.Results showed that among all PTFs only the Vereecken PTF failed to accurately estimate water retention characteristics.Although the MLRP PTF can be used to predict retention characteristics through traditional statistical analyses,it failed to describe the spatial variability of soil water retention characteristics.Although the MLRV and Rawls PTFs failed to describe the spatial variability of water contents at a matric potential of 10 kPa,they can be used to quantify the spatial variability of water contents at matric potentials of 33 and 1 500 kPa.     

Models for estimation of tree volume in the miombo woodlands of Tanzania

Volume of trees is an important parameter in forest management, but only volume models with limited geographical and tree size coverage have previously been developed for Tanzanian miombo woodlands. This study developed models for estimating total, merchantable stem and branches volume applicable for the entire miombo woodlands of Tanzania. We used data from 158 destructively sampled trees, including 55 tree species collected from wide geographical and biophysical ranges. We developed general and site-specific models with diameter at breast height only as the independent variable, together with models with both diameter at breast height and tree height. Leave-one-out cross-validation was used to evaluate the models. The total tree volume models that included diameter at breast height and tree height had appropriate predictive capabilities with relative root mean square errors (RMSE_r) ranging from 30.5% to 47.6%. The RMSE_r for total tree volume models with diameter at breast height only ranged from 39.9% to 48.0%. The site-specific models had slightly lower RMSEr values relative to the general models. The relative mean prediction error of the general total tree volume model with diameter at breast height and tree height was lower (0.6%) than those of the previously developed models (?30.7% to 31.2%). Based on the evaluations, we recommend the general total tree models to be applied over a wide range of geographical and biophysical conditions in Tanzania.     

一种基于改进时间卷积网络的生猪价格预测方法

针对传统的生猪价格预测方法存在预测精度不够高，容易陷入局部最小值等问题，为更加精准地预测生猪价格，采用随机森林回归(RFR)、极限梯度回升(XGBoost)、轻型梯度提升机(LightGBM)3种机器学习模型和改进网络结构的时间卷积网络(TCN)模型方法，以经过Z-Score标准化预处理的西南地区某省2011—2020年每周生猪价格数据为样本，对生猪价格预测进行研究。结果表明:TCN模型预测结果的均方误差(MSE)为0.340 606，平均绝对误差(MAE)为0.288 424，决定系数(R²)为0.995 683，均优于其他3种机器学习模型；与3种机器学习模型中效果最好的极限梯度回升(XGBoost)预测结果比较，3个指标分别提升了26%、8%和0.15%。改进网络结构的时间卷积网络模型可以更加精准地预测生猪价格。     

A Novel Method for Flatness Pattern Recognition via Least Squares Support Vector Regression

To adapt to the new requirement of the developing flatness control theory and technology, cubic patterns were introduced on the basis of the traditional linear, quadratic and quartic flatness basic pat...