Reflection Model for Soil Moisture Measurement Using Near-infrared Reflection Sensor

Surface soil moisture is a significant parameter in environmental systems. A new sensor capable of estimating surface soil moisture from reflection data is presented, called near-infrared reflection sensor. Relative reflectance method is used for prediction model development. The present study investigated the reflection variations in four soil samples with a wide range of soil properties. The results showed that quadratic models were constructed between relative reflectance and soil moisture with R-2 of 0.902, 0.865, 0.955, and 0.953. The R-2 of all combined soils model is 0.886. Compared with individual quadratic models for each soil sample, all combined soils quadratic model generated prediction accuracy with values of root mean square error (RMSE = 2.43%, 3.34%, 5.21% and 2.99%). Soil moisture estimation is largely improved when the quadratic model are developed individually on each soil type except for the Soil 4, compared with all combined soils model. Individual quadratic model yielded performances very similar to the individual linear relationship. Therefore, it is feasible to construct a single quadratic model to minimize that factor effecting on reflection. The most important meaning of this study is that surface soil moisture can be rapidly and accurately measured by near-infrared reflection sensor. In the future, the prediction models can therefore provide quick assessment of surface soil moisture directly in the field.