Deep Learning for an Improved Prediction of Rainfall Retrievals From Commercial Microwave Links

Abstract

Commercial microwave links (CMLs) have proven useful for providing rainfall information close to the ground surface. However, large uncertainties are associated with these retrievals, partly due to challenges in the type of data collection and processing. In particular, the most common case is when only minimum and maximum received signal levels (RSLs) over a given time interval (hereafter 15 min) are stored by mobile network operators. The average attenuation and the corresponding rainfall rate are then calculated based on a weighted average method using the minimum and maximum attenuation. In this study, an alternative to using a constant weighted average method is explored, based on a machine learning model trained to produce actual attenuation from minimum/maximum values. A rainfall retrieval deep learning model was designed based on a long short‐term memory (LSTM) model architecture and trained with disdrometer data in a form that is comparable to the data provided by mobile network operators. A first evaluation used only disdrometer data to mimic both attenuation from a CML and corresponding rainfall rates. For the test data set, the relative bias was reduced from 5.99% to 2.84% and the coefficient of determination (R2) increased from 0.86 to 0.97. The second evaluation used this disdrometer‐trained LSTM to retrieve rainfall rates from an actual CML located nearby the disdrometer. A significant improvement in the overall rainfall estimation compared to existing microwave link attenuation models was observed. The relative bias reduced from 7.39% to −1.14% and the R2 improved from 0.71 to 0.82.

Wildfire Smoke Particulate Matter Concentration Measurements Using Radio Links From Cellular Communication Networks

“…Because of the influence of south‐westerlies, Melbourne was less exposed but experienced a series of major smoke events lasting in the order of 12–48 h for each event during January 2020. Accordingly, this study focused on the area of Greater Melbourne, where access to CML data has been made possible through another experiment focusing on rainfall measurements (Pudashine et al., 2020).…”

“…A Vaisala WXT520 automatic weather station dedicated to the study of rainfall using radio links (Pudashine et al., 2020) was installed at Mt Waverley Reservoir (Figure 2) to measure air temperature (T, °C) and humidity (RH, %). These observations were used as reference for the variable time‐series presented in this study.…”

“…Rain gauges operated by the BoM and the local water agency (Melbourne Water) were used to detect periods of rainfall for each link following the same procedure as described in Pudashine et al. (2020). A total of 168 gauges operated by Melbourne Water and 258 gauges operated by the BoM with a tipping resolution of 0.2 mm and temporal resolution of 6 min were used.…”

“…Because of the influence of south-westerlies, Melbourne was less exposed but experienced a series of major smoke events lasting in the order of 12-48 h for each event during January 2020. Accordingly, this study focused on the area of Greater Melbourne, where access to CML data has been made possible through another experiment focusing on rainfall measurements (Pudashine et al, 2020). Figure 1 shows in true visible colors an area comprising Greater Melbourne (at the center of each subpanel) and the Australian Alps to the East, where smoke plume columns can be seen in subpanel (a).…”

Earth-to-Earth Microwave Rain Attenuation Measurements: A Survey On the Recent Literature

“…Another interesting research which used a deep learning approach relevant to recurrent neural networks (RNN) is reported in [66]. The key-points are (1) the rainfall estimation by taking into consideration only the minimum and maximum attenuation data in order to produce a constant weighted average method for calculating the actual attenuation and (2) that the proposed RNN model was developed by using a disdrometer's (OTT PARSIVEL1 located 20 km from Melbourne) data with an interval of 30 s that they are comparable to those data acquired from network operators.…”