| 摘要: |
| 【目的】土壤水分是影响土壤水盐运动的关键因子。滨海盐渍土地面积广泛,为了更好的掌握和利用滨海地区的盐渍土资源,亟需建立高效的光谱模型以监测土壤水分状态。【方法】以黄河三角洲地区滨海盐渍土为研究对象,通过添加不同盐分浓度的溶液,模拟自然状态下滨海地区土壤水分的蒸发过程。同时采用近地面高光谱技术测量不同含水量的土壤光谱数据,进行了18种不同的光谱变换,分别建立了土壤光谱数据与土壤含水率之间的偏最小二乘回归模型(PLSR)。【结果】结果表明,平滑+归一化光谱变换(包含变量归一化、范围归一化、最大值归一化、面积归一化)后的模型均取得了比较好的效果,其中的平滑+变量归一化后的光谱模型可以直接用于盐渍土水分的反演(R2 = 0.7131,RMSE = 0.0950,RPD = 1.8237)。进一步组合不同的光谱变换方法之后建模,发现平滑+变量归一化+MSC变换后的模型精度得到了明显的提高(R2 = 0.8661,RMSE = 0.0628,RPD = 2.7643)。但是也有部分模型在多种光谱变换方法组合后,精度呈现了明显下降,说明光谱建模时需要选择适当的光谱变换方法。【结论】本次研究建立的高光谱预测模型稳定性强,预测精度高,所使用的水分数据数值宽泛,能够很好地适用于滨海盐渍土区不同土壤水分状态的遥感监测。 |
| 关键词: 土壤含水率;高光谱模型;光谱变换;滨海盐渍土;人工模拟 |
| DOI： |
| 分类号:S152；TP79 |
| 基金项目:国家自然科学基金(41601211)；土壤与农业国家重点实验室开放基金(Y20160007)；山东省重点研发计划(2017CXGC0303)；国家级大学生科技创新(201710435074) |
|
| Prediction model of soil moisture based on spectrum and simulated evaporation data of coastal saline soil |
|
ZHANG Xiao-Guang, KONG Fan-Chang
|
|
Qingdao Agricultural University
|
| Abstract: |
| 【Objective】Soil moisture is a key influence factor to the movement of soil water and salt. Area of saline soil in the coastal region is broad. In order to better grasping and using the potential land resources, soil moisture need to be monitoring by establishing efficient spectral models. 【Method】In this paper, coastal saline soil was selected in Yellow River delta area as the research object. Salt solutions with different salt concentration were added to the same type of soils separately, and then the process of soil water evaporation in natural condition was simulated. Spectral data with different water content and salt content were obtained by the surface hyperspectral remote sensing technology. 18 different spectral transformations were done separately to the processed spectral data. partial least squares regression(PLSR) models were established between the soil moisture content and soil spectral data base on the above processed spectral data. 【Result】Results showed that Among the all types of spectral transformation, models based on smoothing + normalization spectral transformation (contained normalization of variables, range normalization, maximum normalization, and area normalization) has achieved good results. Model based on smoothing + normalization of variables had the best accuracy(R2 = 0.7131,RMSE = 0.0950,ratio of standard deviation to root mean square error (RPD) = 1.8237). It indicated that this model could be directly applied to saline soil moisture inversion. Many combinations of spectral transformation were formed from the several spectral transformation methods. Found by combinations of spectral transformation, model established after the transformation of smoothing+ normalization of variables+ multiplicative scatter correction model got obvious improvement and achieved good prediction (R2 = 0.8661, RMSE = 0.8661, RPD = 2.7643). Accuracies of the models established after other combinations of spectral transformation presented significant decline. It indicated that suitable spectral transformation need to chose for modelling. 【Conclusion】The established model of this paper has the characters of strong stability and high prediction accuracy. The data values of soil moisture can be used to the model was distributed in a wide extent. So it could be applied to the monitoring of different soil moisture states by remote sensing in the coastal saline soil area. |
| Key words: Soil moisture content;Hyperspectral model;Spectral transformation;Coastal saline soil;Artificial simulation |
