中文摘要：

土壤属性的快速、精确测定是实现现代精细农业的基础。本研究分析了江苏省北部滨海土壤的属性特征以及碳酸钙的可见-近红外反射光谱特征,探讨利用可见-近红外光谱估算滨海土壤碳酸钙含量的可行性,比较不同光谱反射率数据集、不同预处理方法以及不同建模方法定量反演的优劣。结果表明：（1）苏北滨海土壤有机质含量较低、碳酸钙含量较高,其光谱曲线在2 340 nm处有较明显的碳酸钙吸收特征;（2）滨海土壤碳酸钙含量与土壤的可见-近红外波段反射率呈正相关,且碳酸钙含量高低对于土壤的近红外波段反射率的影响高于可见光波段;（3）可见-近红外反射光谱可用于估算滨海土壤碳酸钙含量。就建模结果而言,381～2 459 nm波段反射光谱数据集、log（1/R）预处理、偏最小二乘回归三者结合的效果比较理想。

英文摘要：

【Objective】Rapid and accurate measurement of soil properties is fundamental to modern precision agriculture. CaCO_3 is a major component of the carbonate in soil and has some important effects on a series of physical,chemical and biological properties of soil,such as soil p H,characteristics of soil colloids,soil nutrition and soil heavy metals absorption capacity. The traditional method for measuring soil CaCO_3 content is mainly based on chemical analysis,which is often rather costly,time-consuming and destructive. Therefore,the method is far from efficient to meet the requirement of modern precision agriculture. The technology of soil reflection spectroscopy can be used to make up the shortages of the traditional method,and provide a new approach for the study of pedology. This paper is oriented to explore feasibility of using visible-near infrared reflection spectra to estimate CaCO_3 content in coastal soil in North Jiangsu,to evaluate impacts of reflection spectra data sets,pre-processing methods on accuracy of the estimation of CaCO_3 content in coastal soil and to compare different modeling methods on estimating CaCO_3 content in coastal soil. 【Method】A total of 142 coastal soil samples were collected from North Jiangsu and analyzed for soil spectra with a portable Field Spec 3. Soil properties of the soil samples were also determined with chemical methods and characterized,and the characteristics of visible-near infrared reflection spectra of the CaCO_3 in the soil were analyzed. On such a basis,reflection spectra data sets of various spectral bands（e.g. seven sensitive bands,355～2 495 nm,355～780 nm,781～2 495 nm and 381～2 459 nm）,reflectance（R）and its pre-processing methods（e.g. multiplicative signal correction（MSC）,first derivative（FD）,standard normal variate（SNV）,log（1/R）and 1/log R）,and three regressions methods（e.g. partial least squares regression method（PLSR）,principal component regression method（PCR）and multiple stepwise linear regression metho