Abstract: Based on the earthquake data during 1980mdash;2010 in North China region, we use the grids with ltimes;l unit size to cover the area and investigate the variation of non-empty box number of earthquakes. When grid size l is small enough, the number of non-empty boxes tends to be equal to earthquake frequency. When grid size l is large enough, the number of non-empty boxes will be close to 1. This study shows that the influence of grid size on result stability is very weak when l isge;0.5deg;. The variation of non-empty boxes is related to the expanded (enhanced) or constricted (quiescence) patterns of earthquake spatial distribution. The numbers of non-empty boxes with different l roughly show a consistent changing tendency, potentially reflect the stress perturbation or undulation during a short time period indirectly. Small earthquakes are much more clustered, the difference between results with different threshold magnitude gradually tend to be small when threshold magnitude increases gradually. The non-empty grid numbers show a normal distribution, therefore the normal distribution range for non-empty grid numbers can be calculated with the given confident probability and the abnormal could be detected if the data is outside this normal range. The statistics show that in North China the forecasting efficiency is low for quiescence type abnormal, the enhanced type abnormal has a higher correct alarm rate, and the precursory criterion both by enhanced and quiescence type abnormal has a higher forecasting efficiency. This means that the precursory of small or moderate earthquakes before large earthquakes mainly displays as enhanced activity. The results also show that the pattern change of small or moderate earthquake spatial distribution seems to be related to coming large earthquakes more tightly.