Combining Incomplete Information From Independent Assessment Surveys for Estimating Masonry Deterioration

Construction materials used in building structures, such as masonry, wood, and reinforced concrete, deteriorate over time because of many factors including poor design, defective materials or manufacture, and poor workmanship. This article is concerned with estimates of masonry deterioration and the effects of covariates on the damage to bricks on the walls of Ž ve multiple-story buildings of a residential complex located in the Bronx, New York. In this case study, the damage of primary interest was a “spall,” a physical separation of a portion of the brick face from the body of the brick. Eventually, the face becomes so damaged that portions fall off. The result is an unattractive appearance and a hazard to passersby. In this study, spall damage was assessed by means of three different and independent condition assessment surveys: an expensive, precise, and hence very limited scaffold drop survey and two additional inexpensive, but more detailed photographic and visual surveys. The photographic survey was obtained by photographing the walls of the entire residential complex, and the visual survey was done by individuals walking around the periphery of each building and making a visual assessment of the damage to each wall. In the photographic survey, a large amount of incomplete data was unavoidable because of either poor photo angles or various physical obstructions. A binomial regression model using four categorical explanatory variables or factors was Ž tted to the observed photographic spall data. Sparseness of the data, the presence of outliers, and overdispersion were major problems encountered in selecting and Ž tting a suitable model. A small pilot survey, in which the relevant portions of the photographic and visual surveys were matched to 11 drop locations of the scaffold survey, recorded spall counts using each survey method. From this pilot survey, photographic and visual spall data were calibrated to the scaffold drop survey data. It was determined that of the two surveys, only the photographic spall survey was needed to predict scaffold spalls. The estimate of total damage from the photographic survey was then adjusted using the calibration results. Finally, a multiple imputation procedure was used to impute values for the missing data and obtain an estimate (and its standard error) of the true spall rate over the entire residential complex. Sources of uncertainty to factŽ nders in legal trials are discussed and illustrated through the present case.