Abstract: A formula relating the farm land area estimate derived from detailed land survey with that reported in official statistics has been deduced through incentive analysis of the metering caliber and the coverage of the data in both kinds of estimation. It reveals that the difference of the two kinds of farmland area estimates was mainly caused by the difference in the caliber of their measurement units and the difference in the coverage of the data applied in both kinds of estimation. Moreover, the ratio between the farmland area estimate derived from detailed land survey and that reported in statistics increases with the increasing proportion of the area converted to non-agricultural uses and decreases with the growing proportion of the newly reclaimed farm land. The difference of the two kinds of farmland area estimates in 1994 can mostly be interpreted by the difference in the caliber of their measurement unit, the so-called "accustomed mu", which is different from city to city, and the difference in the coverage of the data used in both kinds of estimation, when artificial omitting or exaggeration of farm land area exists only in very few cities. The estimates of farmland area reported in statistics in 1994 and those derived from detailed land survey can be matched with each other for most of the cities after unification of their measuring caliber and data coverage. According to the farm land area estimates for 1996 derived from the detailed land survey, the actual statistically calibrated area of farm land of the 21 cities in Guangdong Province was calculated and the deviation of the estimates reported in statistics in 1996 was obtained through comparison with them. It is considered that the errors of farm-land area estimates reported in statistics for the 21 prefecture-grade cities was comparatively small in 1994, and the phenomenon of overestimating or exaggerating farmland area in statistics existed only in very few cities before 1994. However, some fast-developed cities intentionally exaggerated their farmland area in statistics so as to conceal their rapid rates of farmland conversion to non-agricultural uses, knowing that the farmland area estimates derived from detailed land survey were generally much more than the estimates reported in statistics when the preliminary result of the detailed land survey was obtained, which artificially added some error to the farmland area estimate in statistics of those cities in 1996. Even though, the deviation of farmland area estimate in statistics is still less than 5% for 10 of the total 21 prefecture-grade cities, and 2 of them less than 0.3%. It can be concluded that the man-made exaggeration or overestimation of farm land area in statistics do exist, but such phenomenon generally appeared in the Pearl River Delta Economic Zone where the economy developed comparatively fast, and is not apparent in the cities in mountainous region. In order to protect the farmland resources effectively, and stop the cities from exaggerating farmland area in statistics, it is suggested to monitor the farmland change for the 21 cities with the means of remote sensing and the land use be controlled based on the in-situ verified remote sensing monitoring data.