GNSS interferometric reflectometry (GNSS-IR) can be considered as another remote sensing technique for continuous and local monitoring of soil moisture content, which can be performed in various weather conditions such as rainy and cloudy conditions, as well as different lighting conditions such as day and night. In GNSS-IR Changes in soil moisture result in changes in the signal-to-noise ratio (SNR) component of reflected signals. By analyzing reflected signals, useful information about the reflector can be obtained. SNR is highly dependent on soil moisture. As the soil moisture content increases, the dielectric constant of the soil increases, which causes the reflected signals to have higher amplitudes and higher SNR. Conversely, as the soil moisture content decreases, the reflected signals have lower amplitudes and lower SNR. Therefore, analyzing the SNR of the reflected signals can provide useful information about the soil moisture content.
In this study, data from station P038 in the New Mexico region is used, where multipath signals are used to estimate soil moisture changes over a four-year period from 2017 to 2020. This research has four main steps. In the first step, appropriate satellite tracks with elevation angle between 5 to 30 degrees are selected. SNR data are generated from RINEX files. Then, the initial reflection height is estimated for each path. The phase is obtained for each satellite on each day through several stages. Next, SNR metrics are calculated, and finally, vegetation cover effects are mitigated and removed and the result is converted to volumetric water content.
According to the estimates, the volumetric water content in 2017 was 8.88, which increased to 11.74 in 2018, then slightly decreased to 10.88 in 2019 and finally increased to 12.49 in 2020. In this article, the effectiveness of the LSTM neural network model in predicting the time series of volumetric soil moisture obtained from GNSS-IR signals is evaluated. This prediction will help farmers to prepare their irrigation schedules more efficiently.
The LSTM neural network can maintain its content over a long period of time and essentially remember previous information. Gates are also vectors with values between zero and one that determine how old information should be progressed and new information should be added. Generally, one means to pass and zero means to discard information. The input gate specifies which parts of the input data and to what extent they should be added to the memory content. The forget gate determines which parts of the memory content should be removed. The output gate also determines which part of the hidden state content should contain the memory content.
The model is trained using 80% of observations. By updating the network status with observed values instead of predicted values, the RMS error decreased from 0.09 to 0.05, and the predictions became more accurate. Investigations have shown that performing GNSS observations produces more homogeneous reflective effects around the antenna. Therefore, in order to increase the accuracy and quality of the results, it is suggested to use GNSS interferometric reflectometry instead of just GPS interferometric reflectometry.
