160

140

120

100

80

60

40

Slope = 2.1 lb/ft2 per 100 ft

20

0

0

500

1000

1500

2000

2500

**Elevation (feet)**

Figure 6. The elevation correction factor for the 236 highest quality stations used in

our analyses was 2.1 lb/ft2 per 100 ft (0.33 kN/m2 per 100 m). (To convert lb/ft2 to

kN/m2, mulitply by 0.0479, and for ft to m, multiply by 0.3048.)

process we developed to arrive at answers tended to

When several points at about the elevation of interest

fell above the trend line, he increased his preliminary

bring each of us to about the same answer. We expect

that if any one of us had used our method of analysis

answer.

alone, without receiving feedback from the others along

The other two SENH structural engineers consid-

the way, we may have arrived at significantly differ-

ered both NWS and non-NWS data, but one of them

ent answers for some towns. Thus, we conclude that

gave more weight to the non-NWS information be-

there is merit in involving several individuals in a way

cause it eliminated the step of having to relate snow

that they periodically receive anonymous feedback

depths to snow loads (see equation 1 in Tobiasson &

from each other. This process allowed the group to

Greatorex 1997). Both of these individuals developed

determine most answers before our meetings and pre-

selection criteria that eliminated from consideration a

cluded the need to discuss many of the case studies at

number of the stations on the case study form. The

those meetings. When we met, we concentrated on the

acceptance criteria of one individual were (1) at least

few case studies on which we had remaining concerns

15 years of record, (2) less than 15 (sometimes 20)

or disagreements. This left time for us to explore ways

miles (24, sometimes 32 km) away and (3) Pg/Pmax

of improving the process, ways of simplifying our find-

ratio no more than 1.75 for non-NWS stations and no

ings, and ways of incorporating them into the national

more than 1.5 for NWS stations. The other individual's

standard (i.e., ASCE 7-95) and into practice within

acceptance criteria were (1) at least 20 years of record,

New Hampshire. It also allowed us time to discuss our

(2) less than 15 miles away, and (3) Pg/Pmax ratio no

increasing understanding of ground snow loads in New

more than 1.5. Both then adjusted each selected ground

Hampshire.

snow load to the case study elevation by using an el-

evation correction factor of from 2.0 to 2.5 lb/ft2 per

100 ft of elevation difference (0.31 to 0.39 kN/m2 per

5 ADDITIONAL INVESTIGATIONS

100 m). Both then determined the average value of

For 69 of the 302 stations shown in Figure 2, where a

the ground snow load at that elevation for all the sta-

50-year ground snow load is available, the Pg/Pmax ratio

tions selected. In the vicinity of Mt. Washington, where

exceeded 1.5. Often, the 50-year ground snow load at

a station or two had a value quite different from this

average, a second average was often calculated, elimi-

such stations greatly exceeded other ground snow loads

in the vicinity. For example, the upper outlier in the

nating the outliers. One individual developed separate

lower plot in Figure 4 has a high Pg/Pmax ratio of 1.7.

averages for all data and for "non-NWS" data and gave

more weight to the "non-NWS" average. He always

Responding to this complication proved to be the

most controversial aspect of our analysis. To better

plotted all the data he analyzed and frequently referred

understand what was happening, we examined prob-

back to the case study plots before finalizing his answer.

A review of each individual's final answers indi-

ability plots of several of these stations and determined

that, for them, the log-normal distribution used

cates that no one's approach caused them to be consis-

to generate the ground snow load values on the case

tently much lower or much higher than the group's

final answer. Thus, quantitatively, it appears that the

study forms does not fit the actual trend in lower

318