BACKGROUND Spatially accurate population data are critical for determining health impacts from many known risk factors. However, the utility of the increasing spatial resolution of disease mapping and environmental exposures is limited by the lack of receptor population data at similar sub-census block spatial scales. METHODS Here we apply an innovative method (Population Allocation by Occupied Domicile Estimation - ABODE) to disaggregate U.S. Census populations by allocating an average person per household to geospatially-identified residential housing units (RHU). We considered two possible sources of RHU location data: address point locations and building footprint centroids. We compared the performance of ABODE with the common proportional population allocation (PPA) method for estimating the nighttime residential populations within 200 m radii and setback areas (100 - 300 ft) around active underground natural gas storage (UGS) wells (n = 9834) in six U.S. states. RESULTS Address location data generally outperformed building footprint data in predicting total counts of census residential housing units, with correlations ranging from 0.67 to 0.81 at the census block level. Using residentially-sited addresses only, ABODE estimated upwards of 20,000 physical households with between 48,126 and 53,250 people living within 200 m of active UGS wells - likely encompassing the size of a proposed UGS Wellhead Safety Zone. Across the 9834 active wells assessed, ABODE estimated between 5074 and 10,198 more people living in these areas compare to PPA, and the difference was significant at the individual well level (p = < 0.0001). By either population estimation method, OH exhibits a substantial degree of hyperlocal land use conflict between populations and UGS wells - more so than other states assessed. In some rare cases, population estimates differed by more than 100 people for the small 200 m2 well-areas. ABODE's explicit accounting of physical households confirmed over 50% of PPA predictions as false positives indicated by non-zero predictions in areas absent physical RHUs. CONCLUSIONS Compared to PPA - in allocating identical population data at sub-census block spatial scales -ABODE provides a more precise population at risk (PAR) estimate with higher confidence estimates of populations at greatest risk. 65% of UGS wells occupy residential urban and suburban areas indicating the unique land use conflicts presented by UGS systems that likely continue to experience population encroachment. Overall, ABODE confirms tens of thousands of homes and residents are likely located within the proposed UGS Wellhead Safety Zone - and in some cases within state's oil and gas well surface setback distances - of active UGS wells.

中文翻译：

---

住房单位级别的人口分配：在PA，OH，NY，WV，MI和CA的地下天然气存储井周围的估算。

背景技术空间准确的人口数据对于确定来自许多已知风险因素的健康影响至关重要。但是，由于缺乏类似的人口普查分区空间尺度上的受体种群数据，限制了疾病制图和环境暴露的空间分辨率的提高。方法在这里，我们采用一种创新的方法（通过居住的住所估计的人口分配-ABODE），通过将每个家庭的平均人均分配给地理空间确定的住宅住房单位（RHU）来对美国人口普查人口进行分类。我们考虑了RHU位置数据的两个可能来源：地址点位置和建筑物占地面积质心。我们将ABODE的性能与通用比例人口分配（PPA）方法进行了比较，以估算半径200 m半径内的夜间居民人口以及活跃的地下天然气存储（UGS）井（n = 9834）附近的受挫区域（100-300 ft）在美国的六个州。结果在预测人口普查住宅单元的总数时，地址位置数据通常胜过建筑足迹数据，在人口普查区一级，相关性在0.67到0.81之间。仅使用居住地地址，ABODE估计有20,000户有形家庭，居住在UGS活跃井200 m内的48,126至53,250人，可能包括拟建的UGS井口安全区的大小。在评估的9834口活跃井中，ABODE估计在5074至10之间，与PPA相比，住在这些地区的人口增加了198人，差异在单个井水平上是显着的（p = <0.0001）。通过这两种人口估计方法，OH都会在人口和UGS井之间表现出很大程度的超局部土地利用冲突-比其他评估州更严重。在一些罕见的情况下，对于200平方米的小井域，人口估计相差100多人。ABODE对有形家庭的明确核算证实了PPA预测的50％以上为假阳性，这是在没有有形RHU的区域中非零预测所指示的。结论与PPA相比-在亚人口普查区块空间尺度上分配相同的人口数据时-ABODE提供了更精确的风险人群（PAR）估计，并且对最高风险人群的置信度估计更高。UGS井中有65％的井位于居民区和郊区，这表明UGS系统存在独特的土地使用冲突，可能会继续遭受人口侵蚀。总体而言，ABODE确认，数以万计的房屋和居民可能位于拟议的UGS井口安全区内-某些情况下位于州的油气井地面后退距离之内-处于活动状态的UGS井。