**1. Introduction**

Accurate and continuous monitoring of rainfall is very important in many applications. While measurements equipment such as satellites, radar, and weather stations are commonly used for rainfall monitoring, other opportunistic sources for relevant data are being exploited as we are living in the era of big data [1,2]. Big data research is pushing the boundaries of these new technologies and analytic tools, and one such important technology for providing weather data is the use of existing physical measurements in wireless microwave signals, such as the signal level in commercial cellular communication networks for near-ground rainfall monitoring [3]. Microwave backhaul links are used for communications between cellular base stations, and can also be used for measuring the path-averaged rain rate. Utilizing microwave backhaul links for environmental monitoring has also been recently mentioned as one of the Internet of Things (IoT) applications [4]. The densely deployed microwave links all of the world have grea<sup>t</sup> potential to be used to complement existing monitoring systems.

In the telecommunications industry, to meet the ever increasing demand in consumer data traffic, many countries have already started the deployment of 5G networks. Microwave backhauling is widely used in many frequency bands above 6 GHz and will also remain an essential medium for transport of 5G, in addition to fiber for macro radio deployments. Forty percent of backhaul connections are expected to be based on microwave by 2023, as reported in [5].

The radio spectrum is a scarce resource that is governed by national and international regulations. Operation of 5G networks requires enormous transmission capacity and ultra-low transmission latency [6], which bring grea<sup>t</sup> challenges to microwave transmission links [5]. Lower frequencies allow signals to transmit over longer distances and penetrate buildings better. At higher frequencies, signals have limited coverage, however because of much wider bandwidths they can achieve high capacity. The millimeter-wave (mmWave) technology ranging from 30 to 300 GHz is the key to enabling fast speed and high capacity backhauling in future wireless networks [7].

Governments around the world have allowed operations in the millimeter bands for backhauling, often for little or no licensing fee. Traditional backhaul communications are typically used throughout the 6–42 GHz frequency bands, but they are becoming increasingly popular at various mmWave frequency bands in the 50 GHz, 60 GHz, E-band (71–76 and 81–86 GHz), and 92–95 GHz bands throughout the world [8,9]. Even higher frequencies may be of interest to support the evolution of mobile broadband backhauling beyond 2020, such as 92–114.5 (W-band) and 141–174.8 GHz (D-band) frequency ranges [10,11].

A typical cellular network covers a large area and includes conditions ranging from urban canyons to open rural land. Depend on population density and propagation characteristics, geographical land is classified into one of four categories: dense urban, urban, suburban, and rural. For 5G networks, backhaul links in di fferent frequency bands are adopted for di fferent environments to achieve high capacity [5]. For densely populated areas (categorized as "dense urban"), the E-band (70 or 80 GHz) is favorable for links over a few kilometers and o ffering high capacity in the 10 Gbps band. The W-band (92–114.5 GHz) and D-band (130–174.7 GHz) are currently under investigation and high millimeter frequency bands will be able to support 40 Gbps capacity over about a kilometer range. Microwave links for urban environments typically have short distances and high capacity demand. An E-band link is suitable in these scenarios. In suburban areas, the link length increases and capacity is lower compared to dense urban and urban areas. Traditional bands (e.g., 6 to 42 GHz), multi-carrier, and multi-band, (mid-band, 15–23 GHz) with E-band solutions can be deployed. The range is typically 8 km. For rural environments, the link length increases further, while end site capacity decreases. For these environments, the traditional microwave band is preferred. The range is typically around 15 km. In this article, examples of several latest microwave technologies, including E-band (71–76 and 81–86 GHz) links, line-of-sight multiple-input multiple-output (LOS-MIMO) backhaul links, and multi-band solutions, are investigated.

E-band can provide double the 5 GHz bandwidth, o ffering a 10 GHz aggregate spectrum (71–76 and 81–86 GHz), enabling Gbps data rates. An 1.4-km long E-band link, which was tested by Ericsson and Deutsche Telekom, has demonstrated a data transmission rate of 40 Gbps, with a round-trip latency performance of less than 100 ms [12]. This is about four times greater data throughput compared to current mmWave backhaul links. The outdoor small cell E-band backhaul links can be rapidly deployed everywhere, including street lamps, rooftops, and the sides of buildings. E-band is becoming an essential backhauling band with high global alignment, which is also expected to facilitate dense mmWave 5G deployments on street-level sites. However, signals in mmWave frequencies are known to su ffer from large propagation loss and rain attenuation is one of the main limiting factors [13,14]. As a result, the E-band links are used for high capacity transmission but at shorter distances compared to traditional bands, and can generally be applied to lengths of up to 3 km.

The emerging concept of carrier aggregation enables a much more e fficient use of diverse backhaul spectrum assets. As it is easier to obtain wider channels at higher frequencies, we can aggregate a low frequency carrier for availability and a high frequency carrier for capacity. A multi-band solution combining E-band with traditional bands can increase the transmission distance. The traditional band links are used to guarantee the availability of high-priority services and support transmission distances of 3 to 10 km. This will allow the use of E-band to provide transmission for 5G in much wider geographical areas. A commonly used combination is 18–42 GHz bands and E-band (70 or 80 GHz) for distances up to 5 km (dense urban and urban environments), and 6-15 and 18–42 GHz bands for longer transmission ranges (suburban and rural environments). In this article, we will study the impact of atmospheric conditions on E-band backhaul links in city environments, especially rain attenuation. A 38 GHz mmWave backhaul link deployed in the same region will also be studied for comparison analysis.

The di fferent forms of antenna technology refer to single or multiple inputs and outputs. When there are more than one antenna at the transmit side and receive side of the radio link, this is referred to as a multiple-input multiple-output (MIMO) system. MIMO can be used to provide improvements in both channel robustness and channel throughput compared to a single-input single-output (SISO) system, where there is a single antenna at the transmit side and receive side of the radio link [15,16]. MIMO has been widely used in wireless local area networks (WLANs), long-term evolution (LTE) mobile networks, and fifth generation cellular systems. MIMO with a spatial multiplexing scheme allows capacity to increase almost linearly with the number of antennas. Recently, MIMO technology has been applied to increase transmission rates in point-to-point backhaul links in mmWave bands for next-generation wireless backhaul networks [17–20]. This is referred to as a line-of-sight multi-input multi-output (LOS-MIMO) communication system. Most existing studies of the impact of rain on signal attenuation are for SISO microwave links. Signal attenuation in a MIMO backhaul link due to rain and other meteorological conditions are ye<sup>t</sup> to be studied.

Large signal attenuation can occur due to heavy rain and can severely a ffect the mmWave link quality. Modeling and measurements of mmWave attenuation due to rainfall for near-ground communication links have been addressed in recent studies and are considered very important topics [21–26]. A power law empirical mathematical model relating the rain rate and rain-induced signal attenuation is given by International Telecommunication Union (ITU) Recommendation P. 838-3 and other relevant papers [27–29]. This model is used in the design of reliable communication systems. Recently, it has been suggested that microwave links in cellular networks can be considered as passive weather monitoring sensors, and a power law model relating the rain and rain attenuation can be adopted for rainfall estimation [30–32]. This approach exploits the fact that the strength of electromagnetic signals is weakened by certain weather conditions, especially rain. It makes microwave linking a potential tool for monitoring rainfall conditions with high temporal and spatial resolution. There is significant potential to increase the number of observation points and improve the quality of weather services, including forecasting, now-casting, flood warnings, and hydrological measurements. The most powerful impact is expected in developing countries and regions where no other measurements currently exist. Making use of the existing commercial wireless networks is equivalent to deploying a very high density of weather monitoring sensors and forming wireless environmental sensor networks (WESN) [30,33] all over the world.

The major contributions in this paper are (1) studying and comparing the rain attenuation characteristics of latest mmWave backhaul links and (2) studying the performance of rain rate estimation based on SISO and MIMO links at di fferent frequencies, using existing measurements of the received signal level of the mmWave backhaul links.

This paper is organized as follows. Section 2 presents a brief summary of characteristics of mmWave propagation, the method of rain rate retrieval from the received signal level of mmWave links, and the setup of outdoor test links. Section 3 presents the experimental results on signal variation in sunny and rainy weather, wet antenna e ffects and the performance of rain rate retrieval studies using the latest mmWave backhaul test links. Then, the uncertainties in the experiment and the potentials of the proposed technology in supporting rainfall and flood monitoring in urban areas are discussed in Section 4. Finally, we summarize the work in Section 5.

#### **2. Materials and Methods**
