First identify what the colors represent
A radar image can display reflectivity, an estimated rain rate or an accumulated amount. Those quantities answer different questions. Reflectivity describes the return from targets inside a sampled volume of air, commonly expressed on the logarithmic dBZ scale. Rain rate describes how quickly water is falling, while accumulation describes an amount over an interval. A color is only meaningful with its legend. Reading a reflectivity scale as if it were millimeters per hour silently skips the calculation between the measurement and the estimate.
Sources: radar-measurement
A few large drops change the signal
The conversion is difficult because radar and a rain gauge weight raindrops differently. Under the usual small-particle scattering approximation, reflectivity depends on drop diameter raised to the sixth power. Rainfall depends on drop volume and falling speed as well as the number of drops. Consequently, different mixtures of small and large drops can produce the same radar reflectivity but different rain rates. Empirical reflectivity–rainfall relationships capture useful patterns, yet there is no single conversion that perfectly represents every distribution of drops.
Sources: radar-zr
The beam samples air above the surface
A radar does not collect water at every colored point on its map. It sends radio waves outward and receives echoes from precipitation aloft. Mountains can block the path, while Earth's curvature makes low-altitude precipitation harder to observe at long distances. JMA describes these limitations as important considerations in Japan's mountainous terrain. This geometry means that a map location and a measurement at ground level can refer to different parts of a weather system, even when their horizontal coordinates appear to match.
Sources: radar-jma
Strong echoes can mislead in either direction
Melting snow can produce an unusually reflective layer, called a bright band, that causes rainfall to be overestimated. Water coating the melting particles helps explain why their echoes are stronger than those above or below. The opposite problem occurs when heavy precipitation weakens the radio signal along its path. Water on the radar's protective cover can also attenuate the signal. A weaker return may therefore reflect a compromised measurement path rather than a proportional reduction in precipitation at the distant target.
Sources: radar-jma
Two polarizations provide more clues
Dual-polarization radar compares horizontally and vertically polarized returns. These observations contain information about particle shape, size and consistency that a single reflectivity measurement cannot supply. They help distinguish rain, snow and hail, identify the melting layer, and reject some non-weather echoes. The additional measurements support better precipitation estimates, but they remain inputs to interpretation. A classification product and a rainfall product are calculated outputs, rather than photographs of individual drops or direct collections of the water reaching the surface.
Sources: radar-dual
Read the processing history as well as the image
The useful distinction is between a return, a rate and a total. Producing accumulated rainfall requires estimating rates and accounting for the time between scans. Gauge comparisons can then reveal and help adjust systematic differences. When comparing two maps, their displayed variable and accumulation interval matter as much as their palette. A mismatch need not mean one radar is broken: the products may embody different conversions or processing stages. Keeping those stages visible makes the map more informative without assigning it false precision.
Sources: radar-measurement