Two heights can describe the same point
A height is incomplete without the surface from which it is measured. Satellite positioning provides an ellipsoidal height, referenced to a smooth mathematical model of Earth. Mapped elevation can instead refer to a gravity-related surface associated with mean sea level. The two zero surfaces do not coincide. A difference between the numbers can therefore be a reference difference, rather than evidence that the receiver or map is wrong. Adding decimal places does not resolve a mismatch in what those numbers are describing.
Sources: geoid-basics
Gravity supplies the missing reference
GSI explains the geoid as the mean-sea-level reference extended conceptually beneath the land. Its height above the reference ellipsoid is the geoid height. In the basic relationship, subtracting geoid height from ellipsoidal height gives elevation. This is why the gravity field enters a problem that initially looks purely geometric. The model is not a second terrain map showing mountains beneath the visible mountains; it describes a reference surface used to interpret heights. The correct conversion depends on where the point is located.
Sources: geoid-basics
A model has a construction history
GSI distinguishes models derived from gravity from models fitted to geoid heights inferred through GNSS and leveling. Earlier Japanese models combined these approaches to agree with the established leveling-based height system. That fitting could also incorporate accumulated leveling errors and the effects of crustal deformation. The gravity-based JPGEO2024 model was built using improved gravity observations instead. Knowing this history explains why a model update is more than a cosmetic replacement of one grid: the relationship to the national height system can change.
Sources: geoid-model
Airborne measurements fill difficult gaps
Gravity observations are hard to distribute evenly across mountains and coastal areas using ground surveys alone. GSI's airborne program addressed those gaps with an aircraft carrying a relative gravimeter, plus GNSS and inertial instruments to locate measurements and describe aircraft motion. Its equipment used a stabilized sensor arrangement and vibration damping. Crossing survey lines supported quality evaluation. These observations contributed to the geoid model, showing why an accurate height reference depends on dedicated measurement engineering even when the eventual user only sees a conversion value.
Sources: geoid-airborne
Validation tests the conversion independently
GSI evaluated JPGEO2024 by comparing the model with geoid heights obtained from GNSS and leveling measurements made during the same period. It reports that validation in three regions met the model's three-centimeter target. This is evidence about a particular reference model and evaluation, not a promise that every receiver will deliver elevations accurate to three centimeters. The receiver's height estimate is another input. Keeping model performance separate from the complete measurement system prevents an attractive specification from being applied beyond its tested scope.
Sources: geoid-model
A datum change can alter the number
Japan updated its national vertical survey results to a geoid-based datum on April 1, 2025. GSI documents changes in published heights and notes that some remote islands retain separate origins requiring additional reference treatment. Consequently, two elevation datasets can disagree because they belong to different realizations of the height system. A useful comparison records the datum, geoid model and any applicable local convention. Resolving those references comes before attributing the difference to physical ground movement or treating one value as a corrected version of the other.
Sources: geoid-datum