APPENDIX — ADEQUACY OF DATA 



651 



Special Surveys Made in 15 States 

 AND Coastal Alaska 



For the remaining 15 States and Coastal Alaska, 

 little or no Forest Survey data were available. 

 Seven of these States were covered by special 

 surveys using regular Forest Survey' procedures, 

 but with coverage that was less intensive than 

 usual, often being confined to sample counties. 

 In two States, there were some basic data from 

 surveys conducted by State agencies, and these 

 were supplemented by using aerial photographs, 

 public land records, or new ground plot sampling 

 as required. The other six are Rocky Mountain 

 States with much of their commercial forest land 

 in national-forest holdings. Recent timber inven- 

 tory data on these and on other public and private 

 holdings provided the principal basis for the 

 estimates, but aerial photographs were interpreted 

 and ground plots were measured where such data 

 were insufficient. 



In Coastal Alaska, areas of forest types and 

 stand-size classes were determined from aerial 

 photographs covering 71 percent of the commercial 

 forest land area. Average volumes and growth 

 rates were obtained from a relatively light sample 

 of ground plots and from national-forest inventory 

 data. 



Adequacy of Data Depends Chiefly 

 ON Intensity of Sampling 



The sampling error of the estimate of commercial 

 forest land in the United States and Coastal 

 Alaska is 0.2 percent (table 83). For sawtimber 

 volume, it is 0.8 percent, and for growing stock 

 volume, 0.6 percent. Corresponding sampling 

 errors for net annual growth are 3.1 percent and 

 2.2 percent. 



For the 21 Eastern States completed by the 

 Forest Survey, sampling errors averaged 2.2 

 percent per million acres of commercial forest 

 land, compared to an accuracy goal of 3.0 percent 

 per million acres. Sampling error of growing 

 stock averaged 4.7 percent per billion feet com- 

 pared to a goal of 5.0 percent. In the two States 

 in the West completed by the Forest Survey, the 

 sampling errors of commercial forest area esti- 

 mates averaged 2.4 percent per million acres com- 

 pared to a goal of 3.0 percent. Growing stock 

 error was 12.9 percent compared to a goal of 10.0 

 percent per billion cubic feet set for these States. 



For States in which Forest Survey coverage 

 was sufficiently advanced to extend estimates to 

 the whole State, the sampling accuracy goal 

 varied from 3.0 to 4.5 percent per million acres 

 of commercial forest land, and from 5.0 to 12.5 

 percent per billion cubic feet of growing stock. 

 In two of these States for which sampling errors 

 were computed, North Carolina and Virginia, 



the commercial forest area sampling errors aver- 

 aged 4.9 percent per million acres compared to a 

 goal of 4.5 percent; growing stock errors averaged 

 7.6 percent per bUlion feet compared to a goal of 

 7.5 percent. 



On the basis of the above comparisons for 25 

 States, it is believed that sampling accuracy goals 

 in the remaining 23 States and Coastal Alaska 

 were likewise achieved satisfactorily, and the 

 sampling errors are entered in table 83 on this 

 basis. Goals for the 15 States where little or no 

 Forest vSurvey data were available were generally 

 set at 6 percent per million acres of commercial 

 forest, and varied from 10 to 15 percent per billion 

 cubic feet. 



Estimates of the sampling errors of net annual 

 growth were calculated for five States. For these 

 States, the sampling error per billion cubic feet 

 was less than half the sampling error indicated for 

 growing stock volume. However, the sampling 

 error goals per billion cubic feet were the same 

 for net annual growth as for growing stock volume, 

 and it is on this basis that sampling errors for 

 growth are estimated for the other 43 States. 

 While this would appear to give conservative 

 estimates, judging from the comparison available 

 for five States, this safety margin is adopted to 

 make allowance for the large and usually unknown 

 variability in the mortality component of net 

 growth, and also for possible errors in adjusting 

 both mortality and growth for a particular year 

 to the trend level. 



The sampling error of board-foot growth was 

 computed by multiplying the sampling error of 

 cubic-foot growth in a State by 1.31, this ratio 

 being based on data from States where the 

 sampling errors of both sawtimber and growing 

 stock volumes were calculated. 



The sampling errors of breakdowns of commer- 

 cial forest area by stand-size class, stocking class, 

 and forest type group can be approximated from 

 the relationship shown in figure 1. The steps are: 

 (1) Note the smallest geographic unit of which the 

 breakdown is a part and for which the sampling 

 error is given in table 83. (2) Compute the per- 

 centage that the breakdown contributes to the 

 total, and read from figui'e 1 the corresponding 

 factor. (3) Multiply the sampling error of the 

 total by the factor. This product is the approxi- 

 mate sampling error of the breakdown. For 

 example, the sampling error of the estimate of 

 total commercial forest area in Missouri, 15,064 

 thousand acres, is 0.7 percent. Of this area, 

 2,033 thousand acres, or 13 percent, is classed as 

 sawtimber. From figure 1, the multiplying factor 

 for 13 percent of the total is 2.8. The approximate 

 sampling error for area in sawtimber stands is 

 therefore 2.8 X 0.7, or 2.0 percent. 



The sampling errors of timber volume by species 

 and tree size also can be approximated from 

 figure 1. For example, Douglas-fir makes up 49 



