中文摘要：

土壤磷素为植物提供营养元素,是评价土壤质量的重要参数之一。传统的土壤全磷含量的测定方法不能实现对荒漠土壤养分有效监测,而运用遥感手段能够弥补传统手段的不足。有学者开展了通过近红外光谱来估算土壤全磷含量的研究,但由于土壤磷素近红外吸收系数小、吸收峰不明显等原因,使得土壤磷素估算的模型精度欠佳。为解决荒漠土壤全磷含量近红外光谱估算存在的不足,提高荒漠土壤全磷含量估算的精度,对准噶尔盆地东部荒漠土壤进行采样、化验分析和发射率光谱测量、处理,分析土壤热红外发射率特征,建立多种荒漠土壤全磷含量热红外发射率估算模型。结果表明：在土壤全磷含量高于0.200g·kg-1的条件下,在8.00~13μm波长范围内,热红外发射率随全磷含量的增加而增加,9.00~9.60μm波段范围内土壤热红外发射率对全磷含量最敏感;多元逐步回归建立的估算模型的估算效果差,不能用于荒漠土壤全磷含量热红外发射率的估算,经过偏最小二乘回归建立的估算模型效果优于多元逐步回归建立的模型;偏最小二乘回归建立的连续去除一阶导数模型最优,校正和验证的R2分别达到了0.97和0.82,校正和验证的RMSE仅有0.010 6和0.015 7,RPD为2.62,模型能够极好的对土壤全磷含量进行估算。该研究的成果为荒漠土壤全磷含量定量遥感估算提供有效支撑,通过有效监测荒漠土壤全磷含量的时空动态变化,为区域生态环境的修复提供依据。

英文摘要：

Soil phosphorus provides nutrient elements for plants,is one of important parameters for evaluating soil quality.The traditional method for soil total phosphorus content （STPC）measurement is not effective and time-consuming.However,remote sensing （RS）enables us to determine STPC in a fast and efficient way.Studies on the estimation of STPC in near-infrared spec-troscopy have been developed by scholars,but model accuracy is still poor due to the low absorption coefficient and unclear ab-sorption peak of soil phosphorus in near-infrared.In order to solve the deficiency which thermal-infrared emissivity estimate des-ert soil total phosphorus content,and could improve precision of estimation deserts soil total phosphorus.In this paper,charac-teristics of soil thermal-infrared emissivity are analyzed on the basis of laboratory processing and spectral measurement of deserts soil samples from the eastern Junggar Basin.Furthermore,thermal-infrared emissivity based RS models for STPC estimation are established and accuracy assessed.Results show that：when STPC is higher than 0. 200 g·kg-1 ,the thermal-infrared emissivity increases with the increase of STPC on the wavelength between 8. 00 μm and 13 μm,and the emissivity is more sensitive to STPC on the wavelength between 9. 00 and 9. 6μm;the estimate mode based on multiple stepwise regression was could not to estimate deserts soil total phosphorus content from thermal-infrared emissivity because the estimation effects of them were poor. The estimation accuracy of model based on partial least squares regression is higher than the model based on multiple stepwise re-gression.However,the accuracy of second-order differential estimation model based on partial least square regression is higher than based on multiple stepwise regression;The first differential of continuous remove estimation model based on partial least squares regression is the best model with R2 of correction and verification are up to 0. 97 and 0. 82 respectively,and RMSE of correction and verific