Technical ArticleSpectral Analysis and Estimations of Soil Salt and Organic Matter ContentsLiu, Shibin1; Li, Yi1; He, Chansheng2,3Author Information 1College of Water Resources and Architecture Engineering, Northwest Agriculture and Forestry University, Yangling, China. 2Department of Geography, Western Michigan University, Kalamazoo, MI. 3Key Laboratory of West China’s Environment System, Lanzhou University, Lanzhou, China. Address for correspondence: Yi Li, PhD, College of Water Resources and Architecture Engineering, Northwest Agriculture and Forestry University, Yangling, Shaanxi 712100, China; Email: [email protected] Financial Disclosures/Conflicts of Interest: This work was supported by the Key Project of National Natural Science Foundation of China (no. 91125010) and China 111 project (No. B12007). The authors declare no conflict of interest. Received October 29, 2012. Accepted for publication April 2, 2013. Soil Science: March 2013 - Volume 178 - Issue 3 - p 138-146 doi: 10.1097/SS.0b013e318295ba8f Buy Metrics Abstract Levels of soil salt content (SC) and organic matter content (SOMC) are key factors for assessing soil fertility. Field soils with high salt content hinder crop growth severely. Our objectives are to establish models for SC and SOMC and evaluate the models for prediction accuracy. Spectral reflectance of soil samples was obtained under controlled laboratory conditions using a portable spectrometer. A total of 211 samples were divided into a training set and a validating set for modeling soil properties. Derived from the originally observed reflectance (λ) data, six quantitative λ-related indices were applied to establish models for SC and SOMC by stepwise linear regression analysis using the training data. Adjusted coefficient of determination for each model was used for evaluating model stability. The established models were evaluated for the accuracy of predicting SC and SOMC using the validation data. Results showed that among the established models, the model relating λ to SC (the root mean square error [RMSE] = 2.99, the correlation coefficient [r] = 0.73) had a small difference with the model relating continuum-removed reflectance (CR) to SC (RMSE = 2.94, r = 0.76). Considering the convenience of utilization for the model and the statistical analysis of model outputs, the model relating λ to SC was selected as the best model for prediction. With the smallest RMSE (0.84) and the largest r (0.96) values among the six models, the model relating reciprocal values of λ to SOMC was selected as the best model for prediction. This work supplies a feasible method to estimate SC and SOMC. It has potential for remote sensing applications to determine soil properties. © 2013 Lippincott Williams & Wilkins, Inc.