Abstract:

To conduct annual forest inventories (AFI) is a changing trend in the forest monitoring systems in the world. Because ground plots measured each year in an AFI system are much less than in a traditional inventory system, it is essential to improve the estimation precision by using new techniques. The issue discussed in this paper was about how to improve the area estimation precision in an AFI year by configuring a three-phase sampling framework which takes good use of the relationship between remotely sensed imagery plot data, ground plot data measured in different time. The three-phase sampling approach proposed here takes the remotely sensed imagery plot data as the first-phase sample, the data of all plots in the population measured former as the second-phase sample, and the data of plots measured in the AFI year as the third-phase sample. The third-phase sample was a subset of the first and the second samples, and the second-phase sample was a subset of the first-phase sample. First, an area estimation was made using a two-phase sampling technique and the first and the second sample data. Then this estimation was used to another two-phase sampling which used the data of the second and the third samples. These two steps maked up the complete three-phase sampling technique. In the first AFI year, the second-phase sample was composed of the plot data measured in the last periodic inventory. From the second AFI year and on, the data of the second-phase sample could be from different years. The paper presents three-phase sampling estimation formula and variance and covariance formula. A study cased in Zhejiang Province, China, showed the fact that the three-phase sampling could produce the highest AFI area estimation precision.

Key words: annual forest inventory, three-phase sampling, remote sensing, area estimation