Effects of Data Sparsity and Spatiotemporal Variability on Hazard Maps of Workplace Noise

Personal sampling, considered a state-of-the-art technique to assess worker exposures to occupational hazards, is often conducted for the duration of a work shift so that time-weighted average (TWA) exposures may be evaluated relative to published occupational exposure limits (OELs). Such cross-shift measurements, however, provide little information on the spatial variability of exposures, except after a very large number of samples. Hazard maps, contour plots (or similar depiction) of hazard intensity throughout the workplace, have gained popularity as a way to locate sources and to visualize spatial variability of physical and chemical hazards within a facility. However, these maps are often generated from short duration measures and have little ability to assess temporal variability. To assess the potential bias that results from the use of short-duration measurements to represent the TWA in a hazard map, noise intensity measurements were collected at high spatial and temporal resolution in two facilities. Static monitors were distributed throughout the facility and used to capture the temporal variability at these locations. Roving monitors (typical of the hazard mapping process) captured spatial variability over multiple traverses through the facility. The differences in hazards maps generated with different sampling techniques were evaluated. Hazard maps produced from sparse, roving monitor data were in good agreement with the TWA hazard maps at the facility with low temporal variability. Estimated values were within 5 dB of the TWA over approximately 90% of the facility. However, at the facility with higher temporal variability, large differences between hazard maps were observed for different traverses through the facility. On the second day of sampling, estimates were at least 5 dB different than the TWA for more than half of the locations within the facility. The temporal variability of noise was found to have a greater influence on map accuracy than the spatial sampling resolution.

个体采样（Personal sampling）作为评估劳动者职业危害接触水平的前沿技术，通常以单个工作班为周期开展，以便结合已发布的职业接触限值（OELs）对时间加权平均（TWA）接触水平进行评估。然而，这类跨班采样仅能提供极有限的接触空间变异性信息，除非采集样本的数量极为庞大。危害分布图，即展示工作场所全域危害强度的等高线图（或同类可视化形式），如今已成为定位危害源、可视化设施内物理与化学危害空间变异性的常用手段。但此类分布图通常基于短时测量生成，几乎无法评估时间变异性。为评估使用短时测量值代表危害分布图中时间加权平均水平所引入的潜在偏差，研究团队在两处工业设施中采集了兼具高空间与时间分辨率的噪声强度数据，在设施内全域布设固定监测仪以捕获各点位的时间变异性，同时采用流动监测仪（危害绘图流程中的典型设备）通过多次遍历设施来捕捉空间变异性。研究对不同采样技术生成的危害分布图之间的差异进行了评估，结果显示在时间变异性较低的设施中，基于稀疏流动监测仪数据生成的危害分布图与基于时间加权平均的危害图吻合度良好，约90%的设施区域内，估算值与时间加权平均水平的差值在5分贝（dB）以内；但在时间变异性较高的设施中，不同遍历路径生成的危害分布图间存在显著差异，采样第二日设施内超半数点位的估算值与时间加权平均水平的差值至少达到5分贝。研究最终发现，噪声的时间变异性对危害分布图精度的影响要大于空间采样分辨率的影响。