Rainfall rate climatology mapping for millimetre-wave propagation loss estimation

Deriving the 0.01% exceedance rainfall rates that ITU-R P.837 demands is impossible from sparse gauge networks alone. GPM IMERG's 0.1°, 30-minute global precipitation record makes it tractable, with caveats.

Sensors

GPM IMERG (NASA/JAXA): Primary source. The Integrated Multi-satellitE Retrievals for GPM product combines passive microwave, infrared and gauge inputs at 0.1° (~11 km) spatial resolution and 30-minute temporal resolution globally between 60°N and 60°S. The Final Run carries a ~3.5-month latency for gauge correction; the Late Run is available within ~18 hours. Archive runs from June 2000 (via the TRMM-era V06 back-extension) to present.
TRMM archive (NASA/JAXA): The Tropical Rainfall Measuring Mission (1997–2015) provides the foundational long-record dataset for the tropics, with the TRMM 3B42 product at 0.25°, 3-hourly resolution. Useful for extending the climatological record before GPM launch in 2014, particularly for deriving stable exceedance statistics in equatorial and sub-tropical regions.
Meteosat SEVIRI (EUMETSAT): The Spinning Enhanced Visible and InfraRed Imager on Meteosat Second Generation provides 15-minute full-disk imagery at 3 km resolution in the thermal infrared, used to detect cold cloud-top temperatures as a proxy for convective intensity. Rapid Scan Service reduces the repeat cycle to 5 minutes over Europe and parts of Africa, helping resolve fast-moving convective cells that IMERG's 30-minute compositing can smear.
MSG Rapid Scan (EUMETSAT): Operated in tandem with SEVIRI, the Rapid Scan Service covers roughly 20°W to 45°E at 5-minute intervals. For link designers in Europe, the Middle East and East Africa, this provides the sub-30-minute convective tracking needed to validate whether IMERG is capturing the spatial extent of intense cells adequately.
TRMM Precipitation Radar / GPM DPR (NASA/JAXA): The Dual-frequency Precipitation Radar on GPM Core Observatory measures vertical rain profiles at Ku-band (13.6 GHz) and Ka-band (35.5 GHz) with a 5 km footprint and 250 m vertical resolution. Its near-surface rain rate retrievals are the closest spaceborne analogue to a gauge and serve as the calibration backbone for IMERG passive-microwave retrievals.

Why the worst 53 minutes per year determine your link budget

ITU-R P.530 and P.837 ask a specific question: what rain rate, in millimetres per hour, is exceeded for only 0.01% of an average year? That works out to roughly 53 minutes. For a 26 GHz fixed wireless link, the difference between a rain rate of 40 mm/h and 80 mm/h at that exceedance level can mean 15 to 20 dB of additional specific attenuation, which is the difference between a link that survives a tropical storm and one that drops every time a heavy shower passes. Getting that single number right is the entire point of rainfall climatology mapping.

The ITU-R P.837 model was historically populated from the Global Telecommunication System's rain-gauge network, which is dense in Western Europe and North America and extremely sparse across sub-Saharan Africa, the Pacific island groups and large parts of Southeast Asia. In those regions, the model's uncertainty is not a rounding error; it can exceed 50% of the exceedance rain rate itself. Satellite-derived precipitation climatologies exist precisely to fill that gap.

What rain does to a radio wave above 10 GHz

Water droplets interact with radio waves through two scattering regimes. At frequencies below roughly 30 GHz, where the wavelength is large relative to most drop diameters, Rayleigh scattering dominates: attenuation scales approximately with the sixth power of drop diameter and the square of frequency. Above 30 GHz, drop diameters approach the wavelength and Mie scattering becomes significant, introducing resonance effects and making the relationship between rain rate and attenuation less predictable from rain rate alone. The specific attenuation coefficients published in ITU-R P.838 encode this physics into power-law relationships of the form k·R^α, where R is rain rate in mm/h and k and α are frequency-dependent empirical constants.

At 26 GHz (a common 5G backhaul band), specific attenuation at 50 mm/h is approximately 8 dB/km. At 73 GHz (E-band), the same rain rate produces roughly 20 dB/km. At 140 GHz, which is attracting interest for short-range ultra-dense backhaul, it exceeds 40 dB/km. These are not theoretical curiosities; they set hard upper bounds on link distance for a given availability target. A link designer who underestimates the local 0.01% rain rate by 30% will build a link that fails more than three times as often as the service level agreement permits.

GPM IMERG as a climatological tool: what it does well and where it struggles

IMERG's 0.1° grid and 30-minute temporal sampling are well-matched to the spatial scale of stratiform rain systems but are a poor fit for isolated convective cells, which can have cores narrower than 5 km and intensities that peak and decay within 15 minutes. The product's passive-microwave retrievals estimate rain rate from brightness temperature depression at frequencies including 10, 19, 37 and 89 GHz, calibrated against the GPM Core Observatory's DPR. In flat terrain with predominantly stratiform rainfall, validation studies have found IMERG Final Run to be a reliable estimator of monthly and annual totals. At the sub-hourly, extreme-tail end of the distribution, it systematically underestimates peak rates by amounts that vary by climate regime but can reach 30 to 50% in areas dominated by convective systems.

Complex terrain compounds the problem. Orographic enhancement, blocked passive-microwave views and beam-filling errors all degrade retrievals in mountainous regions. Where a rain-gauge network exists, even a sparse one, gauge-adjusted IMERG (the Final Run product) is preferable to the uncorrected Early or Late runs. Where no gauges exist, the honest approach is to treat IMERG-derived exceedance rates as lower bounds and apply a terrain-aware correction factor informed by published regional validation studies before using them in a link budget.

From a precipitation archive to a rain-fade margin map

The analytical pipeline has four stages. First, the IMERG half-hourly time series is accumulated into a multi-year record (GPM's archive back-extended through the TRMM era gives roughly 25 years from 2000). Second, for each grid cell, the empirical complementary cumulative distribution function of rain rate is fitted, and the 0.01% exceedance value is extracted. Third, that rain rate is converted to specific attenuation using the ITU-R P.838 k and α coefficients for the frequency of interest. Fourth, the specific attenuation is integrated over the planned link path length, with an effective path length reduction factor applied per ITU-R P.530 to account for the fact that intense rain rarely fills an entire long path simultaneously.

The output is a gridded map of required rain-fade margin in decibels, at the spatial resolution of the input precipitation data. For a network planner designing a 26 GHz backhaul grid across, say, a large island group with no rain-gauge infrastructure, this replaces a guess with a defensible climatological estimate. It does not replace a proper link budget tool; it provides the single most uncertain input to that tool.

Honest limits and how to work around them

The 0.1° grid cell, roughly 11 km on a side, is larger than most microwave link paths. Assigning a single exceedance rain rate to a 5 km link is reasonable; assigning the same value to a 500 m E-band hop in a dense urban grid is not meaningfully worse than any alternative short of a co-located gauge. The 30-minute temporal resolution means that rain rates above roughly 100 mm/h, which can occur in less than 15 minutes during severe convection, are almost certainly underrepresented in the exceedance statistics. For tropical markets where extreme convective rain is the design driver, a correction factor derived from regional disdrometer or radar studies is advisable.

Archive depth matters for statistical stability. The 0.01% exceedance level corresponds to roughly 53 minutes per year. Estimating it reliably requires many years of record; a 10-year sample carries meaningful sampling uncertainty, and a 25-year sample is substantially better. The TRMM back-extension to 2000 is therefore worth using despite the lower spatial resolution of the TRMM 3B42 product. Satellize processes the combined TRMM and GPM archive for clients who need exceedance maps in data-sparse regions, drawing on the same open datasets that underpin ITU-R's own model updates. The Tonga crop-estimation programme demonstrated that open satellite archives, properly processed, can substitute for ground infrastructure that simply does not exist.

Integrating SEVIRI for convective event characterisation

IMERG's passive-microwave retrievals are geostationary-infrared-filled between overpasses, which introduces a secondary uncertainty in convective regions. Meteosat SEVIRI's 15-minute full-disk thermal infrared imagery, or the 5-minute Rapid Scan product, can be used to characterise the spatial structure and movement speed of convective systems independently of IMERG. Cold cloud-top temperatures below 235 K are a widely used proxy for deep convection; tracking the fraction of time a given grid cell spends under such cloud tops provides a qualitative cross-check on whether IMERG's exceedance statistics are plausible.

This is not a replacement for IMERG. Infrared brightness temperature is a poor quantitative predictor of surface rain rate; the relationship breaks down entirely for warm-topped convection and drizzle. Its value is diagnostic: if a region shows frequent deep convective signatures in SEVIRI but IMERG reports moderate exceedance rates, that discrepancy is a flag to apply conservative correction factors before finalising the link budget. Conversely, if SEVIRI shows predominantly stratiform cloud structures, IMERG's estimates are more likely to be reliable.

Typical figures

Spatial resolution (IMERG)	0.1° (~11 km at equator)
Temporal resolution (IMERG)	30 minutes
Archive depth	June 2000 to present (TRMM back-extension + GPM era, ~25 years)
IMERG Final Run latency	~3.5 months (gauge-corrected); Late Run ~18 hours (uncorrected)
SEVIRI full-disk repeat (standard)	15 minutes; 5 minutes on Rapid Scan Service
SEVIRI spatial resolution	3 km at nadir (thermal infrared channels)
Coverage	IMERG: 60°N to 60°S global; SEVIRI: Europe, Africa, Middle East full disk
Exceedance level targeted	0.01% of year (~53 min/yr), per ITU-R P.837 requirement
Frequency applicability	10 GHz to 300 GHz (ITU-R P.838 specific attenuation coefficients)
Deliverable grid formats	GeoTIFF, NetCDF, CSV per link path; compatible with standard link budget tools

Analytics Satellize can run

0.01% exceedance rain rate grid	Empirical complementary CDF fitted to 25-year IMERG/TRMM half-hourly time series per grid cell	GeoTIFF raster at 0.1° resolution, clipped to area of interest, with uncertainty band layer
Specific attenuation map by frequency	ITU-R P.838 power-law k·R^α applied cell-by-cell for client-specified frequency or frequency set	GeoTIFF stack (one band per frequency) suitable for import into radio planning software
Path-integrated rain-fade margin	ITU-R P.530 effective path length reduction with site-diversity factor; integrated along planned link geometries	Per-link CSV report with required fade margin in dB, confidence interval and dominant uncertainty source flagged
Terrain-aware correction factor layer	Orographic enhancement index derived from DEM slope and aspect combined with IMERG bias statistics from published regional validation literature	Multiplicative correction raster applied to exceedance grid; documented with methodology note
Convective fraction climatology (SEVIRI)	Cold cloud-top temperature threshold analysis (235 K) on SEVIRI thermal infrared archive; fraction of time per grid cell below threshold	GeoTIFF convective fraction map used as a qualitative IMERG reliability indicator; flagged cells where correction is recommended
Link availability exceedance summary	Monte Carlo propagation of rain rate uncertainty through ITU-R P.530 model to produce probability distribution of annual outage minutes	Per-link PDF report with outage probability curves at 0.001%, 0.01% and 0.1% exceedance levels

Sources

Who does the work

We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.