GENERAL ELECTRIC 
SALMON FALLS RIVER DAM,
Somersworth, NH

In August 1991, a survey network was established to monitor horizontal and vertical movement of the General Electric Salmon Falls River dam located in Somersworth, New Hampshire. The survey network was designed to detect deformations in the horizontal and vertical alignment of the crest of the dam, as well as overall movements of the dam with respect to the surrounding ground. The highly built-up nature of the area around the dam site and the limited intervisibility between control stations present challenges to the detection of overall movements. The granite block dam is approximately 300 feet long and ranges in height from approximately 3 feet to 10 feet.

CONTROL NETWORK DESCRIPTION

Horizontal
The horizontal control network was designed in two components, a network of 6 station set out to monitor for overall movement of the dam and a network of 8 stations along the top of the dam set out to monitor gross bending deformation in the crest of the dam. The movement network consists of 5 occupied stations and 1 intersection station which is also one of the dam crest stations. A total of 10 horizontal distances and 16 angles link the 6 control station as a network. . The 8 dam crest stations consist of 3/8" by 2" lag screws set in lead shields in 5/8" diameter holes bored into the cap stones on the dam. A 9/64" diameter hole was bored into the head of each screw. Screws are set at approximately 40 foot intervals along the dam crest. Right angle offsets were measured to each of these stations from the line defined by two of the movement network stations.

        Vertical

The vertical control network consists of three bench marks and the 8 dam crest stations. The benchmarks are located as follows:

1. Maine Department of Transportation BMf BNB-9-A, second order, class II, (NGVD29), set in the northeasterly concrete abutment of the Berwick/Somersworth Bridge,
2. USGS BM (unnamed) set in the concrete abutment of the spillway at the westerly end of the dam,
3. A chiseled square set in bedrock southwesterly of control station 3.

FIELD SURVEY

Horizontal:
Stations in the overall movement network were linked together with 10 horizontal distances and 16 angles. Angles were measured with a LEITZ SET3 total station between fixed targets set at all intervisible stations. Four positions were measured at each occupied station with a rejection limit of 04" from the mean. The instrument was re-leveled at the start of each position and the horizontal circle was advanced 45° . The means of the 4 positions were used in computations. All optical plummets and tribrach level vials were adjusted immediately prior to the start of work. The instrument was shaded with an umbrella at all occupied stations.

Distances were measured with the same instrument to fixed targets at all intervisible stations. Horizontal distances, slope distances and a pair of zenith angles were measured in both directions at all intervisible stations. All distance observations were made in meters (MMM.mmm) and later converted to U.S. survey feet. The means of the reduced slope distances and the horizontal distances were used in computations. Atmospheric corrections based on observed temperatures and pressures were entered into the instrument at each occupied station. The instrument was tested on the NGS Sanford, Maine EDM calibration baseline one month prior to this project and found to be acceptable. A cyclic error calibration was also conducted on this instrument within one month of this project. The cyclic error was found to be less than 1 millimeter. A zero error calibration was conducted simultaneously with this project by measuring 8 combinations of partial distances between station 3 and station 5 to the dam crest stations. This error appears to be about 0.0018 meters. No corrections were applied to observed data as all calibration parameters were thought to be at the noise level.

Right angle offsets were measured to dam crest stations from both ends of the baseline defined by control station 5 and control station 3. This was accomplished by rotating a specially constructed 4-foot rod in a horizontal plane around a center anchored in the 9/64" diameter hole in the top of each lag screw and observing the shortest offset. A weighted mean of the two observations was used for computations.

Vertical:
Vertical control points and dam crest stations were tied together in a vertical network observed with a LEITZ B2C automatic level and a precise 10-foot single face staff. The staff has not been calibrated by an independent laboratory; however, it was compared for overall length with a precise NEDO invar scale staff which is specified by the manufacturer to have an overall length accuracy of 0.00002 meters with random inaccuracies not exceeding 0.00001 meters. The rod used for the survey was found to have a length accuracy of 0.001 feet between the 1-foot and 9-foot graduations. A peg test (collimation test) was conducted on the level and the results applied to level loop adjustments. Level loops were run by the three-wire method to second order, class I standards except that backsight and foresight lengths could not be balanced for dam crest stations. Collimation, earth curvature, and refraction corrections were applied to these observations. Observations were made to dam crest stations from each end of the dam and averaged.

The overall movement horizontal network was adjusted using a minimum constraints weighted least-squares adjustment. The - minimum constraints selected were north and east coordinate of station 1 held fixed and azimuth station 1 to station 2 held fixed. Standard deviations of 0° 00'02" were applied for all observed angles and 0.010 feet for all distances. The adjustment was successful based on statistical testing of the aposteriori variance of unit weight with the chi-square distribution for 17 degrees of freedom at the 95% (two-tai1ed) level. The range residuals on the adjusted distances was 0.001 feet to 0.011 feet with a RMS error of 0.005 feet. The range of residuals on the adjusted angles was 0° 00'00.2" to 0° 00'02.8" with a RMS error of 0° 00'01.3". The maximum error ellipse relative to the constraint previous outlined was associated with station 2 with a semi-major axis of 0.004 feet at the 68% level. Maximum loop misclosure for the triangles which are defined within the network was 0.007 feet based upon adjusted observations.

Baseline offsets to dam crest stations are the weighted means of directly observed quantities. The only correction applied to observed quantities was for rod zero error. Weighting of observations was based upon sighting distance to the rod. Weights ranging from 1 to 3 were used. Short sights were given the highest weight. These offsets represent distances to fixed points on the dam crest relative to two fixed ground stations. Offsets to a best-fit line were also computed to access dam crest bending in the event that stations 3 and 5 or all stations move. These offsets were computed by a linear regression model normalized to the 8 dam crest station geometry. That is to say, the x axis of the final coordinate system is parallel to the best fit line and therefore offsets are perpendicular to the best fit line. (91-159-00)

The vertical network was computed and adjusted using the MDOT benchmark held as the minimum constraint.

Post Script: Since 1991, the network has been reobserved and analyzed four times.  (Client, General Electric)

In all, about half of the points could be double observed. Five days of precise measurements were made in winter conditions to points accessible only by climbing the steel skeleton. We checked field data using a weighted least squares adjustment conducted in three phases. First, we performed and tested a minimal constraint adjustment of the control network for internal consistency and adherence to the error budget. Second, observations to the points on the skeleton were added and the whole was readjusted. (Redundant observations were entered as observations to uniquely defined points; thus two positions for each of the twice-located points were produced.) Third, coordinate comparisons of these positions were made. 

We found the maximum deviation never exceeded ¼". The final adjustment of the network incorporated all observations and all redundancies. The network and observations were found to be statistically consistent at the 95% confidence level with 253 degrees of freedom.  (Client, Architectural Skylight Company)

We recently completed a boundary survey of the entire 2,000-acre Wapack National Wildlife Refuge in southern New Hampshire and a survey of three parcels with a combined area of 2,000 acres, located in Magalloway Plantation, to be added to the Lake Umbagog, National Wildlife Refuge.

During the fall of 2000, we completed a survey of tracts located in Whitefield and Jefferson, N.H., associated with the Conte NWR. Work included preparation of tract reports and subdivision plans and applications for both towns. A very tight performance schedule was  compounded by the size of the project and field conditions, limited availability of Planning Board meetings, public notice requirements for subdivision, and coordination of approvals with two Planning Boards. Despite these obstacles, work was completed on time and within budget. (Client, US Fish & Wildlife Service)

Additional site mapping and topographic survey work was performed to map the location and elevations of the railroad switch located southwesterly of the southwesterly corner of Building #92 and the location of structures inside dry-dock 1.  (Client, Whitman, Requardt and Assciates, LLP)