On August 26, severe flash floods and debris flows struck the Himalayan region at the border of Nepal and Tibet, China. Following the collapse of glaciers and rock masses, massive amounts of ice, rocks, and mud surged into river channels, forming high-speed floods and debris flows that devastated villages, roads, bridges, and hydropower facilities along the way. As of August 28, the disaster had caused hundreds of deaths and nearly a thousand missing persons. The barrier lakes formed in the disaster area also posed a secondary flood risk, forcing the suspension of some rescue operations.
For such high-altitude disasters, the difficulty of early warning lies not only in detecting abnormal changes in glaciers, mountain bodies, and river water levels in advance, but also in transmitting this information out in a timely manner. When a disaster strikes, power supplies, roads, and terrestrial communication infrastructure may all be compromised. Meanwhile, monitoring nodes installed in mountainous areas face practical constraints such as being unattended, low temperatures, high humidity, insufficient coverage, and long-term power supply challenges. An IoT (Internet of Things) system that normally uploads data smoothly might face a network outage precisely when it is needed most.
This raises a more debatable technical question: if the public network itself can become part of the disaster, how should sensors scattered across mountains, river valleys, and near glaciers maintain connectivity? From LoRa, Wi-SUN, and NB-IoT/LTE-M to 4G/5G and satellite Non-Terrestrial Networks (NTN) entering IoT terminals, different communication technologies are forming a new disaster monitoring and early warning network.
What Exactly Does Disaster Early Warning Require?
The formation of a debris flow or flash flood is usually not a sudden event at a single moment. Rainfall, soil moisture content, groundwater pressure, mountain displacement, and river water levels often change continuously before a disaster occurs. If these variables can be monitored over the long term and abnormal trends identified, there is an opportunity to advance the early warning time from "discovery" after the disaster strikes to the risk accumulation stage.
The most fundamental category of data comes from the meteorological environment. Rain gauges can record cumulative rainfall and rainfall intensity per unit of time, while parameters such as temperature and humidity, air pressure, snow depth, and glacier temperature can be used to assess environmental changes in high-altitude regions. For debris flows and landslides, short-duration heavy rainfall is particularly critical, as large amounts of rainwater entering the soil directly alter soil moisture content and stability.
The second category of data comes from the mountain itself. Soil moisture sensors, pore water pressure gauges, inclinometers, crack meters, and vibration sensors can monitor changes both inside and on the surface of the mountain. If a specific area continuously shows rising moisture content, changing inclination, or abnormal vibrations, it may indicate declining mountain stability. For key hazard zones, high-precision GNSS positioning can also be integrated for long-term observation of millimeter- or even centimeter-level displacements.
The third category involves river and gully data. Once a debris flow forms, it ultimately propagates rapidly along gullies and river channels; therefore, water level, flow velocity, discharge, and water pressure are also crucial monitoring targets. Deploying automatic water level gauges and flow velocity sensors upstream can buy downstream areas a few minutes or even longer for evacuation.
When conditions permit, cameras, millimeter-wave radar, LiDAR, drones, and satellite remote sensing can also be incorporated to observe terrain and river channel changes over a broader area. However, the data volume generated by these devices is far higher than that of ordinary environmental sensors, placing greater demands on network bandwidth, power consumption, and edge computing capabilities.
From this, it is evident that natural disaster monitoring actually corresponds to two entirely different communication requirements. One involves small data packets for rainfall, inclination, and water levels, which may only be a few bytes per transmission, yet require nodes to remain online long-term with low power consumption. The other involves video, radar, and remote sensing data, which demand higher bandwidth and stronger data processing capabilities.
The real difficulty lies in the fact that these monitoring devices are often deployed precisely in valleys, river channels, glaciers, and uninhabited areas. They require long-term unattended operation, have limited power supply conditions, and cellular signals may not be stable. Therefore, the core issue facing the disaster IoT is not "how much data to transmit," but how to reliably transmit the most critical data under conditions of low power consumption, long distance, and extreme environments. This also determines that LoRa, NB-IoT, Wi-SUN, cellular networks, and satellite communications are each suited to play different roles.
No Single Communication Technology Can Handle All Natural Disasters
Deploying sensors on mountains and in river valleys is only the first step. Once the project truly enters the engineering phase, a more practical question quickly arises: through what network should this data actually be transmitted back?
Natural disaster monitoring has an obvious difference from ordinary urban IoT: devices are often located in sparsely populated, topographically complex, or even cellular-uncovered areas, and remain unattended for long periods. Most of the year, many nodes only need to upload a few dozen bytes of data such as temperature, rainfall, water level, and inclination. However, once an anomaly occurs, the information must be sent out promptly. Therefore, coverage distance, power consumption, infrastructure dependency, and network survivability during disasters are often more critical than peak transmission rates.
Let us first look at LoRa technology, which primarily addresses the issues of "long distance, low power consumption, and self-built networks."
LoRa operates in unlicensed frequency bands. A single gateway can connect a massive number of sensor nodes, while terminals remain in a low-power state and only upload data when needed. Typical application data provided by Semtech shows that LoRa links can achieve coverage distances of several kilometers or even up to 10 kilometers in suitable environments, and battery-powered nodes can operate for years. This model is particularly suitable for small-data devices such as rain gauges, soil moisture sensors, inclinometers, and water level gauges.
More importantly, it does not require carrier base stations near every monitoring point. Mountainous areas can first establish their own local LoRa monitoring networks, and then use a few gateways to uniformly backhaul data to the cloud. This means that even if there is no cellular signal in a valley, monitoring can continue as long as the sensors can connect to the gateway.
However, LoRa does not completely solve the communication reliability issue. It merely shifts the risk from a massive number of sensor nodes to the gateways. If the gateways rely on 4G for backhaul, and the disaster causes base station power outages or fiber optic interruptions, the data from front-end sensors may be trapped in the mountains even if they are still functioning normally.
Looking at carrier-based cellular communication technologies, NB-IoT and LTE-M address "how to directly connect a massive number of low-power devices via carrier networks."
Both are cellular LPWA technologies defined by 3GPP. NB-IoT leans more towards low-rate, small-data, and low-power fixed devices; LTE-M provides higher data capabilities and is more suitable for devices requiring mobility or positioning functions. The GSMA views them as complementary mobile IoT technologies, while reducing terminal power consumption for long-term connectivity through mechanisms such as PSM and eDRX.
The biggest advantage of this approach is simple deployment. Near rivers, roads, and villages with existing cellular network coverage, devices like water level gauges and rain stations can directly connect to the carrier network without the need to build an additional wide-area infrastructure.
The existing problems are also obvious: the reliability of cellular IoT ultimately still relies on base stations, power, and transmission networks.
This may not be an issue under normal circumstances, but once floods, landslides, or earthquakes simultaneously destroy power, fiber optics, and base stations, NB-IoT, LTE-M, and ordinary 4G/5G may all lose connectivity. Therefore, in disaster early warning systems, cellular networks are highly suitable as primary communication links, but not necessarily as the sole link.
In scenarios with a continuous distribution of a massive number of nodes, such as river valleys and roads, Wi-SUN provides another approach. It builds a wireless Mesh network based on IEEE 802.15.4-SUN, where nodes can forward data hop-by-hop, rather than all devices directly seeking the same base station. The Wi-SUN Alliance currently positions it primarily for smart utilities, smart cities, and large-scale IoT networks. For disaster monitoring, the value of this architecture lies in connecting a massive number of monitoring points deployed along rivers and mountains, while reducing reliance on a single cellular access point.
The technology that truly changes the communication boundaries in remote areas may come from satellite communications. 3GPP officially incorporated Non-Terrestrial Networks (NTN) into the standard system in Release 17, where IoT-NTN allows cellular IoT technologies like NB-IoT and eMTC to extend to satellite networks. In other words, in the future, some IoT terminals can directly send small amounts of critical data via satellites without terrestrial base station coverage.
This trend has already begun to reach the chip level. The Nordic nRF9151 simultaneously supports LTE-M, NB-IoT, NB-NTN, and GNSS; Sony ALT1350 integrates LTE-M/NB-IoT, unlicensed band communication, and satellite connectivity into a single IoT SoC (System on Chip); Qualcomm 212S directly targets off-grid fixed monitoring devices, with official application scenarios already including environmental monitoring and early fire detection.
Satellite communication is certainly not a panacea. Compared to terrestrial LPWAN, it still faces issues such as terminal cost, power consumption, antennas, sky visibility conditions, and network tariffs, and may also suffer from obstruction in canyons and dense forests. Therefore, a more realistic disaster IoT architecture is not about picking a "winner" among LoRa, NB-IoT, and satellites, but rather assigning them tasks at different levels.
A massive number of low-power sensors can achieve regional networking through LoRa or Wi-SUN; when there is cellular coverage, data can be uploaded via NB-IoT, LTE-M, or 4G/5G; and when the public network is interrupted or monitoring points are located in uninhabited areas, satellite NTN can take over as backup backhaul.
For natural disaster monitoring, the truly critical technical metrics are therefore changing: while the extent of network coverage is important, what matters more is whether the system has a second path to send out that critical early warning information after the terrestrial network fails.
A Truly Reliable Disaster IoT Should Be a "Multi-Layer Communication Network"
If the aforementioned communication technologies are placed into a single system, it becomes evident that natural disaster monitoring is actually unsuitable for relying on a single network. A truly reliable solution is closer to a multi-layer communication network jointly composed of sensor nodes, regional networks, edge gateways, wide-area backhaul, and early warning platforms.
The bottom layer consists of sensing nodes. Rain gauges, water level gauges, soil moisture sensors, inclinometers, crack meters, vibration sensors, and GNSS devices are deployed long-term near mountains, river valleys, and glaciers. The data volume generated by these devices is usually not large, but they are numerous, widely dispersed, and require continuous operation for years or even longer. Therefore, LoRa, Wi-SUN, or proprietary Sub-GHz protocols are more suitable for undertaking the connectivity tasks at this layer, aggregating a massive number of sensor nodes to regional gateways with lower power consumption.
The next layer up is the edge gateway. It should not merely be a simple data forwarder but also needs to possess certain local computing capabilities. For example, when rainfall continuously increases, soil moisture exceeds the threshold, and the mountain inclination begins to change abnormally, the gateway can perform multi-parameter judgments locally, without waiting for all raw data to be uploaded to the cloud for calculation.
This is particularly critical for disaster systems. Ordinary IoT can default to "the cloud is always online," but natural disasters can precisely destroy fiber optics, power, and cellular base stations simultaneously. If all early warning logic is placed in the cloud, once the wide-area network is interrupted, the system may still lose its final alarming capability even if the front-end sensors successfully capture the anomaly.
Therefore, above the regional gateways, at least two wide-area backhaul paths need to be established.
Under normal circumstances, NB-IoT, LTE-M, Cat.1, or 4G/5G can be prioritized because cellular networks have relatively low costs, broad coverage, and are convenient for connecting with existing cloud platforms and emergency management systems. However, in key geological disaster areas, satellite IoT or NTN can serve as a secondary link. When base stations fail or monitoring points are located in uninhabited areas, the system can still send location, risk levels, and critical sensor data via satellites.
Some critical nodes can also be further equipped with local alarms, broadcasting devices, or proprietary trunked communication terminals. This way, even if completely disconnected from the cloud, as long as local sensors and gateways are still working, nearby residents and rescue personnel can be directly notified.
The resulting system actually possesses three layers of redundancy: communication redundancy, energy redundancy, and computing redundancy.
At the communication level, multiple data paths are formed through LoRa, Wi-SUN, proprietary Mesh, cellular, and satellite networks; at the energy level, solar power, batteries, and backup energy storage can be adopted to reduce the impact of grid interruptions; at the computing level, some early warning models are pushed down to edge nodes, avoiding the cloud becoming the sole decision-making center.
This architecture implies that the focus of disaster IoT design has shifted from merely pursuing "coverage rate" to "network resilience." If one link is broken, data can still go out through another path; if the cloud platform is temporarily disconnected, local judgments can still be made; if some nodes are damaged, the network can still maintain the most basic monitoring and alarming capabilities.
For natural disasters, what truly matters is how many communication paths the system has left when roads are washed away, base stations lose power, and fiber optics are interrupted. Only by incorporating this "failure state" into the design can the disaster IoT truly possess early warning value.
Proprietary Communication Protocols: Building a "Temporary Communication Network" When the Public Network Fails
Besides LoRaWAN, NB-IoT, and Wi-SUN, there is another easily overlooked communication method at natural disaster sites—proprietary wireless protocols.
Proprietary protocols do not necessarily mean using special spectrums. Many solutions still operate in the Sub-1GHz or 2.4GHz bands, but developers can define the PHY, packet formats, node wake-up, channel hopping, retransmission mechanisms, and Mesh routing rules themselves. Compared to standardized protocols, its biggest feature is the ability to optimize the network around a specific type of task.
This flexibility highly matches the needs of disaster monitoring. For example, a mountain inclination monitoring node might normally upload data only once every 10 minutes or even longer. However, once an anomaly in inclination or vibration occurs, it needs to immediately increase sampling and communication frequencies, prioritizing the transmission of alarm information. Compared to running a full IP protocol stack, using a streamlined proprietary Sub-1GHz protocol can reduce communication overhead and lower power consumption, while allowing customization for alarm priority, node synchronization, and relay methods.
Chip manufacturers such as TI and Silicon Labs have already provided underlying development capabilities for this. The TI CC13xx series supports Proprietary RF mode, allowing developers to adjust the PHY, data rates, preamble, sync words, and packet structures themselves; Silicon Labs' EFR32FG series can also develop custom protocols and network topologies through the Flex SDK and RAIL underlying wireless interfaces. This means that the same type of Sub-GHz wireless chip can run standard protocols or be designed as a dedicated communication system for specific disaster scenarios.
Compared to ordinary star networks, proprietary Mesh and MANET (Mobile Ad Hoc Network) are particularly worth attention at disaster sites.
Star networks require nodes to maintain connections with a central gateway, whereas Mesh allows devices to forward data to each other. If a certain node, base station, or communication path is destroyed by floods or landslides, the network can re-establish paths through other nodes. This capability of "no center or weak center, self-organizing, and self-healing" gives proprietary Mesh significant value after existing communication infrastructure fails.
Rajant's Kinetic Mesh is a typical case. Its InstaMesh protocol allows nodes to establish multiple wireless connections and dynamically select paths. Once a link fails, it can switch to other nodes or frequency bands. This system has been used in scenarios such as mobile hospitals, disaster recovery centers, and temporary emergency command.
After Hurricane Maria severely hit Puerto Rico in 2017, causing massive damage to local cellular communication infrastructure, LCG Holdings subsequently tested the goTenna Pro Mesh communication system, verifying the feasibility of establishing redundant communication links at disaster sites without public networks. The AWS Disaster Response Team also collaborated with Doodle Labs to design an emergency communication architecture, connecting rescue personnel, vehicles, drones, and field terminals via Mesh Rider MANET, allowing video, location, and field data to continue flowing even after cellular or wired networks were interrupted.
These cases illustrate that proprietary communication can generally undertake two types of tasks in disaster scenarios: one is a long-term, low-power monitoring private network, connecting rainfall, water level, inclination, and displacement sensor nodes via Sub-1GHz; the other is a temporary emergency network after a disaster, connecting rescue personnel, drones, cameras, vehicles, and command centers via Mesh and MANET.
Among domestic manufacturers, Dayu Semiconductor provides a case extending from pre-disaster monitoring to post-disaster communication. At the recently held Shenzhen IoT Exhibition, the author saw at Dayu Semiconductor's booth that Dayu developed the FishHOC long-distance synchronous sensing network based on its self-developed SoC U1 and proprietary narrowband communication protocol FishLINK-X, targeting scenarios with insufficient public network coverage, a massive number of nodes, and long-term unattended operation. According to data disclosed by Dayu Semiconductor, FishHOC can achieve transmission distances of up to 50 kilometers at a 17dBm transmit power, while maintaining milliwatt-level communication power consumption and providing microsecond-level time synchronization capabilities.
FishHOC can also be deployed in star, chain, or strip configurations, with relay nodes handling data forwarding. When some relays fail, surrounding nodes can re-seek available relays to complete registration, thereby enhancing the network's survivability in complex environments. The test network published by Dayu covers approximately 16 km × 2 km, deploying a total of 16,000 nodes. The entire network completes one round of data collection in less than an hour, and broadcast commands are sent to all nodes in under 5 seconds.
Such capabilities are highly compatible with geological disaster monitoring. Early warning for debris flows, landslides, and earthquakes requires long-term collection of data such as rainfall, displacement, inclination, and vibration. Among these, monitoring tasks like vibration and seismic waves may also require different nodes to maintain a relatively precise time reference to analyze the sequence of signal occurrences at different locations. Public information shows that FishHOC has been applied to report displacement, rainfall, and other data at hidden geological disaster points, while also covering scenarios such as forest fire prevention, railway monitoring, photovoltaics, and oil exploration.
Dayu's technology layout has not stopped at low-speed sensing networks.
Targeting post-disaster personnel communication, its FishTALK digital trunked communication solution is also developed based on the U1. The company disclosed that it can achieve communication distances of up to 40 kilometers at a 50mW transmit power and supports multi-party voice. After disasters like earthquakes and debris flows cause cellular network interruptions, such communication systems operating independently of carrier base stations can be used for voice dispatch between rescue teams and temporary command posts.
For drones and on-site video, Dayu provides FishLINK broadband video transmission. Unlike FishHOC, which transmits small amounts of sensor data, FishLINK targets higher bandwidth services like video. The company disclosed a line-of-sight transmission distance of up to 100 kilometers, and it has been applied to scenarios such as forest fire prevention, smart firefighting, and geographic mapping. After a disaster, drones can fly over road-interrupted areas to scout landslides, barrier lakes, and affected villages, and then transmit the video back to the on-site command center via broadband links.
From this product portfolio, a noteworthy technical approach can be seen: disaster communication does not necessarily require a single network to handle all data, but can be layered according to data types—FishHOC handles low-speed sensing, FishTALK undertakes voice communication, and FishLINK resolves video and image backhaul.
This layered approach is actually consistent with the technical direction of the entire disaster IoT. Before a disaster, a massive number of low-power nodes monitor environmental changes long-term; when a disaster occurs, even if parts of the public network fail, regional private networks can still maintain the flow of critical data; entering the rescue phase, ad hoc networks, broadband communication, and even satellite networks take over personnel dispatch and image backhaul.
Of course, the cost of proprietary protocols is also very clear. Lower standardization means it is difficult to directly interconnect devices from different manufacturers, long-term system maintenance relies more heavily on original suppliers, and spectrum planning, network security, device management, and subsequent upgrades need to be resolved independently. For public disaster prevention systems with broad coverage and operational cycles that may reach ten years or even longer, LoRaWAN, Wi-SUN, and 3GPP cellular standards are still more conducive to forming a scaled ecosystem. Therefore, proprietary communication protocols are better understood as a supplementary layer in the disaster communication system, rather than a replacement for standard networks.
Chip Suppliers Behind the Disaster IoT
From the perspective of the chip industry chain, there is no "universal chip" in the natural disaster IoT that can solve all connectivity issues. Low-power sensor nodes on mountains, Mesh networks in river valleys, regional gateways, cellular backhaul, and satellite backups have vastly different requirements for communication chips. Current major participants can be roughly divided into four routes: LoRa/Sub-GHz, Wi-SUN/proprietary protocols, cellular LPWA, and satellite NTN.
At the bottom layer of sensor nodes, Semtech remains the core chip supplier in the LoRa ecosystem. The SX1262 supports LoRa and FSK, with a maximum transmit power reaching +22dBm and a minimum receive sensitivity of -148dBm; the LR1121 further covers Sub-GHz, 2.4GHz, and satellite S/L Band, giving the same RF platform greater network expansion space.
ST, on the other hand, chose to integrate the MCU (Microcontroller Unit) with Sub-GHz RF. The STM32WL not only supports LoRaWAN but can also run Wi-SUN, mioty, and proprietary protocols. This "MCU + wireless" SoC (System on Chip) architecture helps reduce the number of chips and power consumption in sensor nodes.
If there is a need to build Mesh or custom networks, TI and Silicon Labs provide greater protocol development space. The TI CC1312R simultaneously supports Wi-SUN, IEEE 802.15.4, and Proprietary RF, with long-range mode receive sensitivity reaching -121dBm; Silicon Labs EFR32FG25 targets Sub-GHz and Wi-SUN networks, while also supporting proprietary protocols, making it suitable for a massive number of nodes forming multi-hop networks. Domestic manufacturer Dayu Semiconductor further extends this approach to geological disaster scenarios, using the U1 chip and FishLINK-X proprietary protocol as the foundation, undertaking low-power long-distance sensor data transmission through FishHOC.
More obvious changes are happening at the wide-area backhaul layer. The Nordic nRF9151 has already integrated LTE-M, NB-IoT, NB-NTN, and GNSS simultaneously, allowing a single terminal to form more connection choices between terrestrial cellular and satellite networks. Sony ALT1350 further integrates LTE-M/NB-IoT, unlicensed band communication, NTN, positioning, and Sensor Hub into the same SoC.
Satellite chips are also beginning to be redesigned specifically for IoT. MediaTek MT6825 supports 3GPP Release 17 IoT-NTN, with the official target applications directly including remote infrastructure monitoring; Qualcomm 212S has a clearer positioning, targeting off-grid fixed IoT devices, with typical scenarios including power grid monitoring, early fire detection, and soil and environmental management.
In contrast, Cat.1 bis chips like ASR Microelectronics ASR1606 and UNISOC 8910DM do not pursue extreme low power consumption for sensor nodes; their value is more reflected in regional gateways, DTUs, and terminals that need to transmit more data. Both chips target existing 4G networks. ASR1606 adopts a 22nm process and integrates baseband, memory, and PMIC (Power Management IC), while 8910DM also integrates capabilities like Bluetooth and Wi-Fi Scan.
The author believes that the truly valuable solution in the future is highly unlikely to bet on just one technology, but rather to recombine these chips according to sensing, networking, backhaul, and emergency communication, forming a disaster IoT that can continue to work after partial infrastructure failure.
The True Disaster IoT is About Staying Online When the Network is at Its Worst
The recent debris flow reminds us once again that what natural disaster early warning truly tests is not how much data can be collected when network conditions are good, but whether the system can continue to work after roads are interrupted, base stations lose power, fiber optics are damaged, or even the public network is disconnected.
This is also the biggest difference between disaster IoT and ordinary IoT. It requires not just sensors and connectivity, but a communication system capable of tolerating partial failures: front-end nodes use LoRa, Wi-SUN, or proprietary Sub-GHz networks for long-term data collection, regional gateways handle local judgments, cellular networks are responsible for daily backhaul, satellite NTN provides the last line of wide-area backup, and when necessary, proprietary Mesh and emergency communication systems support on-site rescue.
From a chip perspective, this also means that future disaster communication will not be "dominated" by a single protocol. What is truly valuable is forming complementarity and redundancy among different communication technologies.
Ultimately, measuring whether a disaster IoT is reliable does not depend on how many nodes are online on a sunny day, but on whether that most critical early warning information can still be sent out when the mountain begins to move, river water levels rise rapidly, and base stations have already lost power.