We chose an elevation near the upper limit of most

redeeming value was to remind us that we should not

buildings as our case study elevation. Had we done

apply our elevation correction factor above the tree

these case studies at lower elevations, failure to apply

line.

the elevation correction factor would have resulted in

Each of the three CRREL researchers and the three

inappropriately low design loads for some of the build-

SENH structural engineers involved was provided with

ings in each town.

a copy of the "data and methodology" report mentioned

previously (Tobiasson & Greatorex 1997), several

3 CASE STUDY FORMS AND GUIDELINES

representative case studies done by CRREL previously,

and written suggestions by Tobiasson and Greatorex

Case study forms were computer-generated for each

for conducting case studies, a copy of which can be

town. Figures 3 and 4 present such forms for the town

obtained from CRREL.

of Salisbury. The first page (Fig. 3) contains the data

We began by working on 40 towns, about half of

available in the vicinity. For many towns, that tabula-

which were in the rugged northern portion of the state

tion contains data from neighboring states. For

and the rest in the rolling hills of southwestern New

Salisbury, periods of record range from 4 to 44 years;

Hampshire. We each conducted our analysis in our own

about half the stations are NWS and half non-NWS,

way and forwarded our "preliminary" ground snow

and ground snow loads are available in the vicinity at

load answers to a third party at CRREL, who tallied

elevations from 350 ft (107 m) to 1500 ft (457 m),

them without divulging the author of each value, and

bracketing the 900 ft (274 m) elevation chosen for

then sent the tally to us. We then reassessed our an-

Salisbury.

swers in light of those of the five others, and then sent

The final page (Fig. 4) of each case study contains

in our "semi-final" answers, which were tallied in a

two plots of ground snow load vs. elevation. The up-

similar fashion, then returned to us. We met shortly

per plot contains just the data from the nearest six to

thereafter to discuss our various methods of analysis

eight stations, while the lower plot contains all the data

and our answers and to arrive at a final answer for

available within a 25-mile (40-km) radius, plus any

each of the 40 towns. As a result of our first meeting,

NWS first order data within 50 miles (80 km). As

we each made some changes to our method of analy-

shown in Figure 4, the elevation of interest is high-

sis. We then repeated the process for the remaining

lighted on the plots, as is a straight line of best fit us-

100 towns being studied.

ing least squares and the best fit value of the ground

snow load at the elevation of interest. For some towns

the ground snow load "answer" is similar on the upper

4 DIFFERING WAYS OF ARRIVING

and lower plots but for other towns it is quite differ-

AT ANSWERS

ent.

Ground snow loads generally increase at higher el-

The three individuals representing CRREL had done

evations up to the tree line. Above the tree line, they

many case studies and were comfortable with the case

may decrease because of wind action. The upper plot

study forms and the guidelines for analysis. They

in Figure 4 has a negative "slope" (i.e., elevation cor-

closely followed the instructions, giving more weight

rection factor) of 1.67 lb/ft2 per 100 ft (0.26 kN/m2

to closer stations and stations with longer periods of

per 100 m). The few data points on the "nearest 6"

record. They gave little weight to stations with less

plot result in an unrealistic slope and thus the ground

than about 15 years of record and they gave little weight

snow load answer of 68 lb/ft2 (3.3 kN/m2) is not to be

to stations where the ratio of the 50-year ground snow

trusted. The lower "all values" plot in Figure 4 con-

load (i.e., Pg on the case study tabulation) to the larg-

tains enough data points to generate a physically more

est ground snow load ever measured there (i.e., the

realistic slope of 2.5 lb/ft2 per 100 ft (0.39 kN/m2 per

Record Max value, Pmax, on the case study tabulation)

100 m) and, thus, a believable ground snow load of 80

was greater than 1.6. They flagged such stations on

lb/ft2 (3.8 kN/m2).

the upper plot and added a few stations somewhat fur-

Data from near the 6288-ft (1917-m) summit of Mt.

ther away, but with longer periods of record, to re-

Washington created problems. The tabulated ground

place them. Often, more stations were added than were

snow load there is only 56 lb/ft2 (2.7 kN/m2), which is

eliminated. Then they either "eyeballed" or calculated

far below the ground snow load at many other places

a new line of best fit in their quest for that case study's

at elevations well below 1000 ft (305 m). The high

answer. When "eyeballing" in a line of best fit, they

gave it a slope of between 2 and 2.5 lb/ft2 per 100 ft

winds on that treeless summit result in ground snow

(0.31 to 0.39 kN/m2 per 100 m), based on the written

load measurements that are much too low to be used

for our purposes. Several plots containing the Mt.

suggestions mentioned above. Two of them found it

Washington value have a negative slope and the ground

valuable to bound the good data by upper and lower

snow load answer suffers as a result. While Mt. Wash-

lines at one of these slopes. Their answer was usually

ington and a few other stations frustrated us, their im-

somewhat above the midpoint of the upper and lower

plications were worth considering. Mt. Washington's

bounds at the case study elevation. The third individual

devoted additional attention to the geographical posi-

315