Evaluating the Use of Spatially Precise Diurnal Population Data in Aviation Noise Studies (2020)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

13 3 Evaluation of Data Sources 3.1 Introduction In this section, we identify and evaluate potential sources of spatiotemporal population distribution data that could be utilized for aviation noise studies. While the focus of our evaluation was on diurnal movement, we also investigated any population data that has a temporal aspect—be it for smaller increments of the day or for seasonal movements. The key points on which we conducted our evaluation are the following: • Data Availability: We evaluated the ability to obtain the relevant data sets, the pertinent costs, and ease of access. • Data Quality: We considered the data’s level of detail, completeness, and accuracy, and identified strengths and limitations. • Data Applicability: We considered how the data could be applied to aviation noise studies. That is, what types of analysis does the data lend itself to? • Data Compatibility: We assessed the ease or difficulty in transforming the data into an AEDT population data set, keeping in mind the fact that metrics used in aviation noise studies have a need for similar resolution and availability. Since no one model can accurately capture all movements of people throughout the day and night, another criterion we have considered in our evaluation of individual spatiotemporal population data sets is the degree to which the data set represents underlying variability. In addition, we considered the application of data sources based on samples or smaller geographic areas, rather than fully detailed data sets. This enabled us to easily transform smaller samples and construct AEDT input decks encompassing or representing diurnal variations. This ability allowed for visual representations within AEDT as a precursor to conducting sample noise analyses. Lastly, while we did consider developing hybrid data sets that produce spatiotemporal population data by merging static population data with other data sources (e.g., city planning data), we discovered that many of the existing sources of spatiotemporal data already follow this approach. For example, ORNL’s LandScan data set is developed by beginning with the U.S. Census data and applying a variety of methods and additional data inputs to generate a data set that includes daytime distributions. There are at least three broad categories of spatiotemporal population distribution data that are available in a somewhat ready-to-use form. These three categories are: • Category 1: Manually derived or sampled population data such as the ACS or the LEHD data. • Category 2: Derivative population distributions that use static data, algorithmic methods, and other inputs (e.g., satellite observations) to derive time-variant population distributions by using static population data as the base set. Examples of these include ORNL’s LandScan and ESRI Daytime Population data sets.

14 • Category 3: High-resolution location data that is based on emerging technologies such as cell tower triangulation or device-recorded GPS positions. The basis for the comparison of these data sets is the current baseline of the decennial U.S. Census population data used at the census block level. A census block is the smallest geographic unit used by the U.S. Census Bureau for tabulation of 100-percent data (data collected from all houses, rather than a sample of houses). The number of blocks in the United States, including Puerto Rico, for the 2010 census was just over 11 million. Census blocks are grouped into block groups, which are grouped into census tracts, which typically are subdivisions of individual counties. Blocks are typically bounded by streets, roads, or creeks. In cities, a census block may correspond to a city block; however, in rural areas where there are fewer roads, blocks may be limited by other features. The population of a census block varies greatly. As of the 2010 census, there were almost five million blocks with a reported population of zero, while a block that is entirely occupied by an apartment complex might have several hundred inhabitants. In this section, we present an overview of one or more data sources evaluated in each category as well as our findings pertinent to the evaluation criteria. 3.2 Category 1: American Community Survey The ACS is an ongoing survey, conducted by the U.S. Census Bureau, which collects information about the population on a yearly basis. It is used by many public-sector, private-sector, and not- for-profit stakeholders to allocate funding, track shifting demographics, plan for emergencies, and learn about local communities. Sent to approximately 295,000 addresses monthly (or 3.5 million per year), it is the largest household survey that the Census Bureau administers. The ACS is different from the decennial census in that it is conducted every year to track social and economic trends and geographic movements. A core attribute of the ACS pertinent to this research is the information collected from respondents regarding their physical home and work addresses. 3.2.1 Data Availability ACS data is free to obtain and download from the official Census website. Data are available in both ACS summary files in comma-delimited (CSV) text format and more detailed text formats available via File Transfer Protocol (FTP). The ACS collects data on an ongoing basis, January through December, and releases new data sets every year. 3.2.2 Data Quality Data available from the ACS are much coarser than desired for aviation noise studies. Commuter-adjusted population estimates are available only at the county level. In particular, the data are extrapolated to show county-to-county flows. In addition, the survey does not collect information on any non-work activities (e.g., school) that take respondents away from the residence. This means that all non-work activities are excluded from the county-to-county flow data. Lastly, since the survey is only sent to about 3.5 million addresses, out of a total of about 137 million, it represents less than 3% of the total number of housing units.

15 3.2.3 Data Applicability Our initial evaluation of the ACS data suggests that they are not at a level of resolution and fidelity that can be immediately useful for aviation noise studies. Since aviation noise studies typically utilize census block points, the resampling of ACS county-level daytime values would almost certainly result in excessive smoothing and over-simplification. In addition, the coverage of only 3% of the working population’s households is likely insufficient to accurately account for movements of all demographics. 3.2.4 Data Compatibility Transforming raw ACS data into formats suitable for modeling with AEDT or noise modeling tools is a simple process. However, the resolution of the data at the county level precludes an accurate assessment of block-level daytime locations of persons. Any resampling will have to resample county-level counts down to the block level, scale the 3% coverage up to 100%, and account for movements by non-workers. 3.2.5 Conclusions We conclude that while the ACS could be useful for supplementary analyses of worker movements, it is probably unsuited to support the development and use of spatiotemporal population data for aviation noise modeling. 3.3 Category 1: LEHD Origin–Destination Employment Statistics (LODES) The LEHD data are the result of a partnership between the Census Bureau and individual U.S. states to provide high-quality local labor market information and to improve the Census Bureau’s economic and demographic data programs. LEHD data is obtained by taking state labor market information and data on employment and residence from the U.S. Social Security Administration and applying it to the U.S. Census data to generate local employment dynamics data. LEHD data are not directly available and are only made available to qualified researchers. However, in addition to these restricted data sets, LEHD creates public-use data sets and online tools that are easily accessible. One of these is the LEHD Origin-Destination Employment Statistics (LODES) data set (see Figure 3).

16 Figure 3: LODES Movement for the City of San Francisco 3.3.1 Data Availability It does not cost anything to obtain and download LODES data from the Census website. Data are available by version (version 7 is enumerated by 2010 census blocks and earlier versions are enumerated by the 2000 census blocks), U.S. state, years of data (2002–2015), and job type (all jobs, primary jobs, private jobs, federal jobs, etc.). Data files are state-based and organized into three types: origin–destination (OD), residence area characteristics, and workplace area characteristics, all at census block geographic detail. Of key interest is the OD data because it links where people live to where they work, by census block. At the time of the writing of this document, data are available for most states for the years 2002–2015. New versions of LODES and the accompanying data are released on an ongoing basis. However, year-specific data appear to lag about two years behind the current calendar year. 3.3.2 Data Quality A promising characteristic of the LODES data set is that data are available for individual census blocks. The OD data provides worker movement data that lists all worker movement from an originating (home/residence) block to a destination (work) block. Data points can be filtered by other criteria such as age, earnings, job type, or industry. A limitation is that since the data are only collected for workers, all non-work activities are excluded from the block-to-block flow. A sample analysis for blocks within the Atlanta metropolitan area shows that of about 3.4 million people included in the analysis area, the LODES data accounts for movement of about 1.5 million persons

17 moving from their home blocks to other blocks for work and about 1.8 million persons moving to blocks inside the analysis area for work. Even taking into account non-workers (e.g., tourists, children, and volunteers), the LODES data are accounting for the movement of about half of all the residents, thereby allowing for reasonable estimates of daytime population locations. 3.3.3 Data Applicability Given that the data are available at the census block level, the LODES data sets are suitable for use in aviation noise studies without a lot of additional effort. However, to account for full coverage of the movement by different demographics, it might be necessary to develop additional methods to model the movement of non-workers. The LODES data are in a form that is suitable for any aviation noise studies that use census block data. In particular, they will provide a reasonable estimate of the daytime distributions of populations. 3.3.4 Data Compatibility To support this research, we developed simple scripts to parse and transform LODES data files into formats suitable to be imported into AEDT. Since the underlying LODES data are census block points, the centroids are coincident with the census block points obtained from the U.S. Census. 3.3.5 Conclusions The LODES data set is promising as a source of daytime population distributions at the desired level of resolution (census block points). Not only is it free, it is easily accessible, easily transformed, and is continually updated. Additional work might be necessary to model the movements of non-workers; however, as a reasonable source of daytime population distributions, it is probably useful as a source of diurnal population data for use in aviation noise modeling. 3.4 Category 2: ORNL’s LandScan™ Global ORNL’s LandScan Global is a community standard for global population distribution data. For the United States, it is a rasterized data set of approximately 1 km (30 x 30 arc seconds) spatial resolution, and it gives an ambient population representing the average number of people who are likely to be present at any given time during a 24-hour period for typical days, weeks, and seasons. The database is updated annually based on a number of inputs and released to the broader user community around October. 3.4.1 Data Availability LandScan data are available at no cost to students and to researchers working in academia. Users outside these categories are required to purchase the data set that they would need. The current pricing for the global data set is about $6,500 and for the United States only about $800. The data are not available as smaller data sets, e.g., state, metropolitan area, or city. The data are provided in ESRI grid format and ESRI binary raster format. GIS software such as ArcGIS (subscription) or QGIS (free, open source) is required to be able to open the files and use them. LandScan Global updates are typically released on an annual basis.

18 3.4.2 Data Quality The LandScan algorithm uses spatial data and imagery analysis technologies and a multi-variable dasymetric modeling approach to disaggregate census counts within administrative boundaries. Based upon the spatial data and the socioeconomic and cultural understanding of an area, cells are preferentially weighted for the possible occurrence of population during a day. The population distribution model calculates a “likelihood” coefficient for each cell and applies the coefficients to the census counts, which are employed as control totals for appropriate areas. The LandScan methodology includes a manual verification and modification process to improve the spatial precision and relative magnitude of the population distribution. A large component/input into LandScan is information gleaned from satellite imagery (car counts, density of people, etc.). Satellite imagery needs to be collected during the daytime; specifically, most is collected at around 10:30 a.m. from most sensors. Since information from imagery is critical to LandScan population counts, even though the nominal value of the counts is an ambient 24-hour count, the grid cell count is more heavily weighted towards representing a daytime population. Since the base data set for LandScan is the U.S. Census, for any distinctly bounded geographic area, the total count of LandScan population is almost identical to the census population. The global data set is a rasterized 1 km x 1 km grid with 20,880 rows and 43,200 columns covering North 84 degrees to South 90 degrees latitude, and West 180 degrees to East 180 degrees longitude, and contains 902,016,000 values with each value being the number of people per cell. At this resolution, the LandScan population centroids are at a fidelity that is nominally much coarser than the census block resolution. This variability is dependent on the location such that in rural areas, a census block may encompass a larger area than a LandScan centroid, whereas in densely populated urban areas, a LandScan centroid may represent 10 or more census block points. Figure 4 depicts LandScan centroids (green points) overlaid on census block points close to San Francisco International Airport.

19 Figure 4: LandScan Cells and Centroids Overlaid on Census Block Centroids 3.4.3 Data Applicability Researchers at ORNL have employed detailed research, imagery, analysis, and modeling techniques to develop the LandScan data set from the base census data. The result is a sophisticated derivative of the census population representing the ambient 24-hour distribution. The data can be used, as presented as a regular grid, for contour generation and population tabulation within desired contours. It can also be resampled onto census block centroids to be used for standard noise studies that use census block points. This process is discussed next as part of Data Compatibility. 3.4.4 Data Compatibility To be suitable for use in aviation noise modeling, it is preferable that the LandScan grid points are resampled onto census block points. For the purposes of this report, we carried out this exercise for sample study areas such as San Francisco, Atlanta, and Washington, D.C. The population points we used as the baseline are census block centroids from the 2010 census for the chosen analysis area. We then identified the LandScan cell that each population centroid is in using a nearest neighbor analysis with the raster cell center point. The LandScan population for each cell was then evenly distributed among all census block centroids in each cell to generate a LandScan version of census block point centroids. Simple scripts were written to transform these regenerated centroids into population receptor sets that can be input into AEDT.

20 3.4.5 Conclusions The LandScan data set represents the cutting edge of spatiotemporal population distribution development. The development of the data encompasses diurnal movement, weekend movement, and even seasonal movement. While it nominally represents a 24-hour average, it is weighted towards emphasizing daytime distributions. After appropriate transformation, the data set might represent a promising alternative to the standard census distribution. 3.5 Category 2: ORNL’s LandScan™ USA ORNL’s LandScan USA is a restricted data set available to the military, Department of Homeland Security, and other federal and state agencies. Provided only for the United States, it is a rasterized data set of approximately 90 m (3 x 3 arc seconds) spatial resolution. LandScan USA provides two layers with modeled population distributions – a nighttime layer and a daytime layer—both on the same 3” x 3” rasterized grid. The data set is updated annually based on a number of inputs and made available to authorized users through the Homeland Security portal. 3.5.1 Data Availability LandScan USA data are restricted and only available to authorized military, federal, and state agencies through the Homeland Security portal. Users outside these categories will be required to request authorization through an appropriate government sponsor. We obtained the data for use in this research. The data are provided in ESRI grid format and as ESRI binary raster format. GIS software such as ArcGIS (subscription) or QGIS (free, open source) is required to be able to open the files and use them. LandScan USA updates are typically released on an annual basis. 3.5.2 Data Quality The data set for the U.S. is a rasterized 3-arc second grid (about 90 meters square) with each value being the number of people per cell. The LandScan algorithm uses spatial data and imagery analysis technologies and a multi-variable dasymetric modeling approach to disaggregate census counts within administrative boundaries. The database contains both a nighttime residential as well as a baseline daytime population distribution that incorporates movement of workers and students. Based upon the spatial data and the socioeconomic and cultural understanding of an area, cells are preferentially weighted for the possible occurrence of population during the day and at night. The population distribution model calculates a “likelihood” coefficient for each cell and applies the coefficients to the census counts, which are employed as control totals for appropriate areas. The LandScan methodology includes a manual verification and modification process to improve the spatial precision and relative magnitude of the population distribution. In the LandScan USA model, each census block is characterized using the land cover data to estimate the individual percentages of urban (residential, commercial, and industrial classes) and non-urban (agricultural, forests, and other classes) along with the census block population and number of housing units. Based on these evaluations, each census block is allocated to a sub-model that uses a specific allocation algorithm that relates such characterizations to cultural and settlement geographic understandings.

21 The data set for the U.S. is a rasterized 3-arc second grid (about 90 meters square) with each value being the number of people per cell. At this resolution, the LandScan population centroids are at a fidelity that is comparable to the census block resolution in urban areas. Figure 5 depicts LandScan USA centroids (gray points) overlaid on census block points (black) close to Las Vegas McCarran airport. The points containing zero population have been removed to reduce the number of points to be modeled in AEDT. Figure 5: LandScan USA Centroids (gray) Overlaid on Census Block Centroids (black) 3.5.3 Data Applicability Researchers at ORNL have employed detailed research, imagery, analysis, and modeling techniques to develop the LandScan data set from the base census data. The result is a sophisticated derivative of the census population containing daytime and nighttime distributions. The data can be used, as presented as a regular grid, for contour generation and population tabulation within desired contours. Since the grid resolution is comparable to the resolution of urban census block points it can likely be used in lieu of census block centroids.

22 3.5.4 Data Compatibility The LandScan USA data set is at a resolution that is probably suitable for use as-is with aviation noise studies. Since the data set contains both a daytime layer and a nighttime layer, if desired it can also be resampled onto census block centroids to be used for standard noise studies that use census block points. For the purposes of our evaluation we extracted and transformed subsets of the data into AEDT input files. With appropriate tools such as ESRI’s ArcGIS or the open source QGIS, it is relatively straightforward to carry out the data extraction and transformation. A visual display in AEDT contrasting color-coded (by population count) of resampled LandScan USA daytime centroids against LandScan USA nighttime centroids shows an easily discernible movement of population between the centroids. 3.5.5 Conclusions The LandScan USA data set represents the state of the art in high-fidelity spatiotemporal population distribution development. The development of the data encompasses temporal movement of different demographics. Of significant value are the separate daytime and nighttime layers. The data set is of a resolution and quality that is a desirable alternative to the standard census distribution. However, until the restrictions on its usage are lifted it is likely to be unavailable to most aviation noise practitioners. 3.6 Category 2: ESRI Daytime Population ESRI is an international supplier of GIS software, web GIS and geodatabase management applications, and related data products. ESRI has developed its daytime population data set to highlight how commuting diasporas expand or contract a given area’s population and can radically change the day and night demographic profiles (see Figure 6). Figure 6: Example of ESRI’s Daytime Population Data near Washington, D.C.

23 3.6.1 Data Availability The daytime population data set is available as a single data set for the entire U.S. for a cost of $800. The data are supplied in either CSV format or Microsoft Excel format. The data are available for the U.S., state, county, census tract, block group, census subdivision, places, ZIP codes, core-based statistical area, congressional district, and designated market area levels of geography. Population tables included in the data are daytime population workers, daytime population residents, total daytime population, and daytime population density (population per square mile). The data are purchased under license and the terms of the license allow for up to five users to work with the raw data. The raw data cannot be shared with others, re-sold, nor made available on a website, unless it is in a static format such as a report, map, or graphic. If new data sets are derived from the raw data, the terms of the license do not apply to the derived data set. ESRI’s Daytime Population data set is part of ESRI’s Demographics—a suite of products and tools that is updated annually. 3.6.2 Data Quality ESRI’s Daytime Population data set estimates daytime population by utilizing base data from the U.S. Census and applying a mix of estimates from ESRI’s updated demographics, the ACS, and business data from Infogroup. ESRI defines daytime population as the total number of “residents” and “workers” present at any given location during normal business hours. The daytime residents include population less than 16 years of age, working-age persons who are unemployed or not in the labor force (e.g., retirees, homemakers, college students, and group populations in nursing homes, juvenile detention centers, and homeless shelters). Therefore, children at schools or day care centers are not counted. The modeling process incorporates the methodological distinction between workers and persons employed. The former represents persons working throughout the workday, while the latter also includes persons employed but absent from work for various reasons such as illness, personal business, or vacation. 3.6.3 Data Applicability The highest level of fidelity available in the daytime population data set is the Census Block Group. For general usage with aviation noise, the data will have to be resampled to the finer grain census block level. The default resampling would be an even distribution to all block points within a Census Block Group. Once allocated to block points, the data can be used for most aviation noise modeling analyses that use population data. 3.6.4 Data Compatibility The ESRI daytime data set is provided either in ArcGIS or CSV format. With the CSV format, a very small amount of effort was necessary to extract and transform daytime population data into the resolution and format suitable to import into AEDT for noise modeling. 3.6.5 Conclusions The ESRI Daytime Population data set is a promising source of diurnal population data so far. Their daytime distribution can be used in conjunction with the census data set (nominally considered to be a nighttime distribution). The resolution of the data is coarser than desired but

24 can be resampled to finer grain census block points. Lastly, the cost to obtain the data set is quite reasonable and should be affordable to most aviation noise practitioners. 3.7 Category 3: AirSage Activity Density AirSage is a company that specializes in providing mobile-location-based insights to researchers, advertisers, planners, and other consumers of population movement. AirSage purchases location data from data aggregators who collect device location data. This location data is collected by a variety of different software applications installed on consumer devices such as mobile phones, tablets, fitness devices, and smart watches. The product of most interest to this research is their Activity Density data set. 3.7.1 Data Availability AirSage’s Activity Density product provides population density in geographic grids (1,000 meters, 100 meters, or 10 meters) for a specific geographic bounding box and time period (e.g., calendar month or entire year). Continuing conversations and correspondence with AirSage have indicated that prices scale with the size of the area and the requested time period. As of August 2018, they indicate that data for one month for a single county will cost in the range of $5,000 and data for one month for a Metropolitan Service Area such as the San Francisco Bay Area will be in the $15,000–$20,000 range. Their standard time period configured for their Activity Density product is one month. Requesting smaller subsets, e.g., two days or one week, will likely cost more than the base minimum of $5,000. Historical data are available back to January of 2017. Grid resolution—1,000 meters, 100 meters, 10 meters—for the selected area also factors into the cost. According to AirSage, they collect and aggregate data on a near-real-time rolling basis and subsequently make those data available as part of their products and services. 3.7.2 Data Quality AirSage’s Activity Density product is derived by identifying distinct devices and the movement of each distinct device. Data for unique devices are aggregated from a multitude of providers, quality checked to de-duplicate reported locations, and anonymized. Distinctions are made between home/residential locations, work locations, end points (stationary points neither at home or work), and transient locations (moving points). The home locations are used to compute weights to represent the entire U.S. population, including Alaska, Hawaii, and Puerto Rico. Weights are calculated for calendar months at the census tract level. For example, if 200 unique devices are recorded (by the census) in a census tract with 1,000 people, then all movements and locations are scaled by a factor of five to scale to 1,000 people. Nationwide, the average coverage of devices recorded by AirSage to persons recorded by the census is about 20%.

25 3.7.3 Data Applicability At the appropriate resolution, e.g., 10 meters or 100 meters, the Activity Density data set can be used to develop fairly accurate spatiotemporal distributions of population for different times of the day, week, and year. These data could be transformed into population centroids at a resolution much better than anything currently used for aviation noise modeling. These data could be used both as regular grids for contour generation or utilized as-is as population centroids. If desired, they could even be resampled onto existing census population centroids. One concern is that coverage is about 20%. Further investigation will be necessary to evaluate the correctness of the extrapolated distribution. 3.7.4 Data Compatibility At the current time, we have not procured data from AirSage. Based on their product sheet, the data are provided in one CSV file or a series of CSV files. If the coordinates and population counts are provided, transforming these data into receptor set inputs for AEDT should be straightforward. The data could also be imported into GIS software for use with noise contours. 3.7.5 Conclusions At this moment in time, obtaining suitable spatiotemporal population data, for aviation noise modeling, from a company such as AirSage could be cost-prohibitive for most aviation noise studies. As this is an emerging field, data availability and cost may evolve. 3.8 Category 3: Other Mobile Data Providers We reached out to other potential providers of high-resolution, spatiotemporal mobile phone data to see if we could purchase or otherwise obtain sample data. In the industry, these companies are known as “location aggregators” that obtain location data directly from major providers such as AT&T, Sprint, T-Mobile, and Verizon. The companies we contacted included LocationSmart, Mobile Walla, SAP Consumer Insights, and Zumigo. Of these, SAP let us know that their Consumer Insights product was no longer available for the U.S. At the time of the writing of this document, we have not received a response from the other three location aggregators. Obtaining location data from these aggregators might be a difficult endeavor because of recent data breaches. Major cell phone companies are terminating location data sharing agreements with third parties because of the aggregators leaking or selling precise, real-time location data that they should have neither leaked nor sold (Fung 2018). 3.9 Conclusions In our research we focused on discovering and evaluating existing sources of spatiotemporal population data that could be suitable for use in aviation noise studies. That is, we only tangentially evaluated models and methods for modifying population distributions to represent diurnal variations. Suitable data sources are few and far between. Specifically, there are very few available data sources that adequately and realistically cover the movement of all demographics and at the desired fidelity or resolution.

26 ACS and LODES, provided by the U.S. Census Bureau, are two free sources of data that we evaluated. ACS provides estimates of county-to-county worker movement; therefore, it is unsuited for direct usage in noise modeling. LODES turned out to be well developed, and the data are in a form that is almost ready to use for aviation noise studies that normally use census population data. The population distributions shown in Figure 7 use LODES data. The only drawback to the LODES data is that it does not include non-workers. Additional models or methods would have to be developed to extend coverage to all demographics. Figure 7: LODES Residential (left) & Workplace (right) Population Distributions Going into our research, LandScan was a promising data source that provides comprehensive demographic coverage. The resolution of the LandScan Global data set is quite coarse. At a 30-arc second resolution, the data has to be resampled to higher-resolution census block points (see Figure 8). Additionally, the LandScan Global data set provides a single layer of 24-hour ambient population. This data set is probably not as valuable for use in noise modeling involving the use of population centroids. However, LandScan USA has a much finer grid at a 3-arc-second resolution and provides both daytime and nighttime population distributions. A drawback of LandScan USA is that it remains a restricted data set available only to the military, federal agencies, emergency planners, and authorized researchers. If the restrictions on its use are lifted the LandScan USA data set would be useful to researchers.

27 Figure 8: Census 2010 Population (left) and Resampled LandScan data (right) ESRI’s Daytime Population set seems promising as they seem to use models and methods similar to LandScan. With population centroids at the Census Block Group level, the resolution is good. Our evaluation of the ESRI daytime population set reveals that it is a very promising source of daytime population distributions when paired with normal census data. Device-based data, such as AirSage’s Activity Density product, provide accurate, high-fidelity spatiotemporal population distribution data. However, at this time, 20% of the population covered. We anticipate that as device usage and data collection increase, the availability and coverage of high-fidelity spatiotemporal population data will become a viable data source for aviation noise modeling.