Enhancing rainfall measurement using microwave links at three scales

Abstract

The application of microwave signals in telecommunication has advanced in recent decades. Hence the CMLs (Commercial Microwave Links) have been spread all over the lands. Meanwhile, the decrease in power of the signal due to rainfall has raised the idea of measuring rainfall with microwave attenuation. This PhD research delves into the challenges of these techniques and investigates potential solutions to enhance the accuracy of measuring rainfall. The objectives of the study set to three levels of raindrops’ impact, signal characteristics, and catchment evaluations.

In the first stage, the cross-sections of different sizes of raindrops are calculated by solving Maxwell’s differential equations in COMSOL software. Rainfall is calculated from different Drop Size Distribution (DSD) models. Then, the correlation between the resulting attenuation and rainfall provided different coefficients for the power correlation between rainfall and attenuation. Regarding the variability in DSD with rainfall intensity, alterations in extinction cross-section play a role in establishing a fluid relationship between attenuation and rainfall. This necessitates adjusting constants for the correlation across different rainfall intensities.

The results support the idea of having varying coefficients for different rainfall intensities. At the second level, the signal characteristics were investigated to find a robust method for extracting attenuation from the received power. The study led to defining a reference signal level through signal denoising methods of moving average, Butterworth, and Chebyshev methods. The study showed the superior performance of the moving average in predicting rainfall, especially in 38 GHz links (H and V) indicated by NSE and KGE. Overall, the moving average also resulted in 2%, 16%, and 14% lower RMSE than the conventional method in microwave links 26H, 38H, and 38V, respectively. Butterworth and Chebyshev filters exhibited higher NSE and lower RMSE in 38 GHz links, with competitive accuracy in the 26 GHz link compared to conventional methods. The effect of wet antenna also decreased and even vanished in some cases.

At last, the catchment runoff was used to calibrate the relationship between rainfall and attenuation. In a study in 15 catchments in Melbourne, the suggested method established a relation between attenuation and the quick flow of 15 flow stations. Meanwhile, the coefficients of the rainfall-attenuation formula were calibrated to reach the highest accuracy of runoff prediction. The results indicate a more robust connection between attenuation and quick flow in contrast to conventional rain gauge data. The new suggested method achieved a maximum NSE of 0.78 and KGE of 0.86. Compared to rain gauges, the conventional CML model improved NSE by an average of 197%, while the new model showed an additional 50% enhancement. RMSE optimisation also demonstrated improvement, with a 15% enhancement using the conventional CML model and an additional 4% on average with the new model. Moreover, the revised flow calibration technique reduces the wet antenna effect by an average of 49%.