摘要
讨论了质地差异对土壤含水量高光谱预测模型的影响。选取巢湖流域60个土壤样本,通过水分梯度实验与光谱测量,获得有效土样高光谱数据382条。根据土样质地组成及质地分异度差异,共设计6个样本组,运用偏最小二乘回归方法(PLSR)建立各样本组土壤含水量高光谱预测模型。结果显示,6个样本组土壤含水量PLSR光谱预测模型均有较好的预测力,建模集R2在0.872~0.944,预测集R2在0.823~0.927,RPD均高于2.0,模型稳定性好、拟合程度高,具有极好的预测能力,但在模型预测精度上各样本组间有一定差异。总体上,样本组土样质地分异度小、土样粉粒含量高有益于提高模型预测精度。相关性分析显示,土壤含水量与土壤光谱反射率相关系数绝对值在0.6~0.9间,质地与土壤光谱反射率相关系数绝对值在0~0.25之间,土壤含水量与土壤光谱反射率高相关性相对削弱了质地对土壤含水量高光谱预测建模的影响,使模型保持较高的稳定性和预测力。
The effect of soil texture difference on hyperspectral prediction model for soil moisture content was discussed.Based on 60 soil samples in chaohu basin,382 effective hyperspectral data of soil samples with different moisture content were obtained by moisture gradient experiment and spectral measurement.According to the difference of soil texture composition and heterogeneity,six soil sample groups were designed,and the hyperspectral prediction models of soil moisture content were established by partial least square regression(PLSR).The results showed that the soil moisture content PLSR hyperspectral prediction models of 6 sample groups had good prediction ability.The calibration set R2 of the sample groups was 0.872-0.944,the prediction set R2 was 0.823-0.927,and the RPD was higher than 2.0.The models had good stability,high fitting degree and excellent prediction ability,but there were some differences in prediction accuracy between the models of sample groups.On the whole,the small texture heterogeneity and high silt content of soil sample were beneficial to improve the model prediction accuracy of sample groups.The correlation analysis showed that the absolute values of correlation coefficient between soil moisture content and soil spectral reflectance were 0.6-0.9,and the absolute values of correlation coefficient between texture and soil spectral reflectance were 0-0.25.The high correlation between soil moisture content and soil spectral reflectance weakened the influence of soil texture on spectral prediction model of soil moisture content,and kept the model stable and predictive.
作者
任申萍
吕成文
陈东来
邓浩然
REN Shen-ping;LYU Cheng-wen;CHEN Dong-lai;DENG Hao-ran(School of Geography and Tourism,Anhui Normal University,Wuhu 241003,China)
出处
《安徽师范大学学报（自然科学版）》
CAS
2020年第6期554-557,572,共5页
Journal of Anhui Normal University(Natural Science)
基金
国家自然科学基金项目(41371229).
关键词
土壤质地
土壤含水量
高光谱预测模型
soil texture
soil moisture content
hyperspectral prediction model
