基于近红外光谱的土壤氮含量模型及生物炭对土壤光谱的影响

采集添加生物炭的土壤（标记为ABS）和不添加生物炭的土壤（标记为CS），获取其近红外光谱，通过预处理算法和偏最小二乘法（partial least squares,PLS）建立两种土壤氮含量预测模型。试验结果显示，CS和ABS分别经过Baseline和Smoothing预处理的预测模型效果最好，定向系数（determination coefficient,R2）分别为0.913和0.753，预测均方根误差（root mean square error of prediction,RMSEP）分别为0.093和0.753，利用近红外光谱可对两种土壤氮含量建模预测。研究了生物炭对土壤光谱及建模的影响，结果表明，添加生物炭会改变土壤成分含量，使近红外光谱和建模不同于普通土壤，而联合建模可减小差异的影响，取得较好的预测效果，联合建模结果显示，经过Smoothing预处理的预测效果最好，R2为0.907,RMSEP为0.086。

Soil Nitrogen Content Model Based on Near Infrared Spectroscopy and Effect of Biochar on Soil Spectra

Soil with biochar（labeled as ABS）and soil without biochar（labeled as CS）were collected and near-infrared spectra were obtained, two kinds of soil nitrogen content prediction models were established through pretreatment algorithm and partial least squares regression（PLS）, results showed that prediction models of CS and ABS pretreated with Baseline and Smoothing were the best, R2 was 0.913 and 0.753 respectively, RMSEP was 0.093 and 0.753 respectively, near infrared spectroscopy could be used to model and predict nitrogen content of two kinds of soil．Then, effects of biochar on soil spectra and modeling were studied, results showed that addition of biochar would change content of soil components, making near-infrared spectroscopy and modeling different from ordinary soil, and joint modeling could reduce impact of differences and achieve better prediction results, joint modeling results showed that prediction effect after smoothing pretreatment was the best, R2 was 0.907, RMSEP was 0.086．