Refining NC State Park Unit Visitation Estimates Using Mobile Tracking Data

PAGE 9 North Carolina Division of Parks and Recreation Department of Natural and Cultural Resources July 2025 REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 9 CONTENTS 1 Introduction 4 2 Regression analysis results: table and 7 histograms 33.1. Interpretation of Four Regression Results 11 3.2. Parks with Low R-squared Values 18 3.3. Parks with High R-squared Values 19 Interpretation of results and discussion 11 Utilizing strong visitors per vehicle 20 multiplier - visitation estimates (VPVM-VE) and mobile tracking data (MTD) correlations to refine visitation estimates 4 4.1. Three Ways to Utilize the Observed Relationship 20 4.2. Hybrid Visitation Estimates (HVE) and Adjustment Factors 21 4.3. Table of Proposed Adjustment Factors 22 4.4. Table of Hybrid Visitation Estimates (HVE) with and without 24 inclusion of the intercept. 4.5. A Hybrid Visitation Estimate (HVE) that splits the percent 26 increase between HVEs with and without the intercept. NC State University. (2025). Refining NC State Park Unit Visitation Estimates using Mobile Tracking Data. Raleigh, NC: NC State University, College of Natural Resources. 5 6Appendix A. Regression Analysis Result Plots (Panels A-E) 30 Implications for management and policy 28 Recommendations 29 Suggested Citation PAGE 9 This project was made possible by a collaborative effort involving researchers at multiple institutions. NC State University Stacy Supak Lincoln Larson Charlynne Smith NC Division of Parks and Recreation Dave Head Vonda Martin The superintendents of all the parks involved in the study. ACKNOWLEDGEMENTS This research was funded by the NC Division of Park and Recreation (NC DNCR MOA 2933) Funding REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 3 PAGE 9 This report presents recommendations from previous research to integrate Visitors Per Vehicle Multiplier - Visitation Estimates (VPVM-VE) with additional data sources, such as Mobile Tracking Data (MTD) and trail counters, to generate more accurate visitation estimates. To accomplish this, we collaborated with PlayCore and the location intelligence company Placer.ai to obtain MTD, which includes geolocation data from cellular phones. They provided monthly MTD for 41 state park units (40 state parks + Occoneechee Mountain SRA) throughout 2022. Additionally, we retrieved the monthly VPVM-VE for the same year using the NC Park Attendance Dashboard. INTRODUCTION REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 4 PAGE 9 To evaluate the relationships between VPVM-VE and MTD, we conducted regression analysis for each of the 41 state park park units. The results, presented in plots (see Appendix A. Panels A-E) and a comprehensive table detailing R-squared values, slopes, and intercepts, are discussed in this report. Parks with high R-squared values are candidates for adjustment factors that can be used to create Hybrid visitation estimates (HVE). Parks with low R-squared values are identified and recommendations are made to help better understand the discrepancy between the VPVMVE and MTD datasets. Figure 1 and Table 1 compare Visitors Per Vehicle Multiplier - Visitation Estimates (VPVM-VE) and Mobile Tracking Data (MTD) for 2022, demonstrating consistently higher estimates via VPVM. Figure 1. Overage of the Visitors per Vehicle Multiplier - Visitation Estimates (VPVM-VE) compared to Mobile Tracking Data (MTD) for each park by month in 2022. REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 5 PAGE 9 DIVISION UNIT NAME PARK REGION VPVM-VE 2022 TOTAL MTD 2022 TOTAL VPVM-VE 2022 OVERAGE VPVM-VE 2022 % OVERAGE CAROLINA BEACH CABE COASTAL 856,418 242,190 614,228 71.72% CARVERS CREEK CACR COASTAL 151,734 47,856 103,878 68.46% CHIMNEY ROCK CHRO MOUNTAINS 371,276 466,953 -95,677 -25.77% CLIFFS OF THE NEUSE CLNE COASTAL 200,931 125,315 75,616 37.63% CROWDERS MOUNTAIN CRMO PIEDMONT 760,330 303,287 457,043 60.11% DISMAL SWAMP DISW COASTAL 43,551 14,723 28,828 66.19% ELK KNOB ELKN MOUNTAINS 46,005 25,789 20,216 43.94% ENO RIVER ENRI PIEDMONT 843,439 212,603 630,836 74.79% FALLS LAKE FALA PIEDMONT 1,244,806 596,032 648,774 52.12% FORT FISHER FOFI COASTAL 1,108,980 333,126 775,854 69.96% FORT MACON FOMA COASTAL 1,020,663 583,429 437,234 42.84% GOOSE CREEK GOCR COASTAL 146,155 106,838 39,317 26.90% GORGES GORG MOUNTAINS 180,052 91,421 88,631 49.23% GRANDFATHER MOUNTAIN GRMO MOUNTAINS 96,035 109,654 -13,619 -14.18% HAMMOCKS BEACH HABE COASTAL 210,236 98,485 111,751 53.16% HAW RIVER HARI PIEDMONT 80,733 42,470 38,263 47.39% HANGING ROCK HARO PIEDMONT 872,657 356,750 515,907 59.12% JONES LAKE JONE PIEDMONT 148,719 74,047 74,672 50.21% JORDAN LAKE JORD COASTAL 2,055,579 955,703 1,099,876 53.51% JOCKEY'S RIDGE JORI COASTAL 982,328 405,757 576,571 58.69% KERR LAKE KELA PIEDMONT 1,395,789 724,878 670,911 48.07% LAKE JAMES LAJA MOUNTAINS 530,415 278,354 252,061 47.52% LAKE NORMAN LANO PIEDMONT 881,370 229,764 651,606 73.93% LAKE WACCAMAW LAWA COASTAL 185,333 138,453 46,880 25.30% LUMBER RIVER LURI COASTAL 155,510 60,112 95,398 61.35% MAYO RIVER MARI PIEDMONT 95,789 29,016 66,773 69.71% MERCHANTS MILLPOND MEMI COASTAL 130,609 40,583 90,026 68.93% MEDOC MOUNTAIN MEMO COASTAL 173,627 47,634 125,993 72.57% MOUNT JEFFERSON MOJE MOUNTAINS 117,755 57,762 59,993 50.95% MOUNT MITCHELL MOMI MOUNTAINS 332,691 155,308 177,383 53.32% MORROW MOUNTAIN MOMO PIEDMONT 209,236 163,560 45,676 21.83% NEW RIVER NERI MOUNTAINS 365,702 96,123 269,579 73.72% OCCONEECHEE MOUNTAIN OCMO PIEDMONT 172,716 48,472 124,244 71.94% PETTIGREW PETT COASTAL 42,799 38,224 4,575 10.69% PILOT MOUNTAIN PIMO PIEDMONT 1,052,678 231,290 821,388 78.03% RAVEN ROCK RARO COASTAL 343,851 126,076 217,775 63.33% SINGLETARY LAKE SILA COASTAL 25,360 12,926 12,434 49.03% SOUTH MOUNTAINS SOMO MOUNTAINS 420,132 74,157 345,975 82.35% STONE MOUNTAIN STMO MOUNTAINS 409,478 150,668 258,810 63.20% WEYMOUTH WOODS SANDHILLS WEWO PIEDMONT 126,711 24,053 102,658 81.02% WILLIAM B. UMSTEAD WIUM PIEDMONT 826,817 277,034 549,783 66.49% Table 1. Results of analysis comparing aggregate 2022 Visitors per Vehicle Multiplier Visitation Estimates (VPVM-VE) and the aggregate 2022 Mobile Tracking Data (MTD) for each NC state park. The VPVM-VE 2022 Overage represents the subtraction of the MTD from the VPVM-VE. The PVM-VE 2022 % Overage represents the VPVM-VE 2022 Overage divided by the VPVM-VE. REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 6 PAGE 9 2. REGRESSION ANALYSIS RESULTS: TABLES AND HISTOGRAMS Table 2 presents the results of the analysis comparing VPVM-VE and MTD for each state park unit. The columns include Division Unit Name, Park, Region, R-squared values, slope values, and intercept values. R-squared values show how much the MTD helps explain the differences (or variance) in the VPVM-VE. In simple terms, it tells us how well the mobile data matches the vehicle count data. If the R-squared value is high (close to 1), it means the mobile visit data does a great job of explaining how many vehicles are in the park. In other words, the number of mobile visits is closely linked to the number of vehicles, and mobile data helps predict vehicle counts very accurately. If the R-squared value is low (closer to 0), this means the mobile data doesn’t explain much of the variation in the vehicle counts. This could mean there are other factors affecting the vehicle numbers that mobile data doesn’t capture, like unique visitor behaviors at the park, weather patterns, or areas with poor mobile coverage. For parks with low R-squared values, it might indicate that mobile tracking data doesn’t fully capture the visitation patterns at those parks, and further investigation is needed to understand what’s causing the discrepancies. Slope Values: Slope values show how strongly MTD and VPVM-VE are related. When the relationship between the two is strong (indicated by a high R-squared), the slope is useful for understanding the connection. If the slope is more than 1, it suggests that vehicle visits increase faster than mobile visits, while a slope less than 1 indicates that vehicle visits increase more slowly than mobile visits. However, if the R-squared value is low (the relationship is weak), the slope becomes less reliable, and its interpretation may not accurately reflect the data. In these cases, the slope should be interpreted with caution, as it may not accurately represent the relationship between the two datasets. Intercept Values: Intercept values show the baseline number of vehicle visits when no mobile visits are recorded. A high intercept means that even when there are no mobile visits (i.e., MTD shows zero), many vehicle visits are still counted at the park. This suggests that some visitors may not be captured by the mobile tracking data, such as visitors who don’t carry mobile devices or those who disable location tracking on their phones. On the other hand, a low or negative intercept suggests that either MTD is missing a significant portion of visitation or that vehicle visits are being overestimated by the VPVM-VE. In parks with high R-squared values, a high intercept indicates that while mobile data is capturing most visitation, there are still vehicle visits that mobile data misses. In parks with low R-squared values, this might indicate larger problems, like poor mobile coverage or incorrect vehicle counts. For a more detailed interpretation of state park units’ R-squared values, slope values, and intercept values, see Section 3: Interpretation of Results & Discussion. REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 7 PAGE 9 DIVISION UNIT NAME PARK REGION R–SQUARED SLOPE INTERCEPT CAROLINA BEACH CABE COASTAL 0.356175967 1.903509962 32.95057685 CARVERS CREEK CACR COASTAL 0.180358075 1.976318616 4.762941359 CHIMNEY ROCK CHRO MOUNTAINS 0.963429319 0.771201581 0.930092359 CLIFFS OF THE NEUSE CLNE COASTAL 0.291876654 0.874187919 7.615178416 CROWDERS MOUNTAIN CRMO PIEDMONT 0.970981882 2.017294904 12.37589005 DISMAL SWAMP DISW COASTAL 0.303158106 0.732309965 4.118119348 ELK KNOB ELKN MOUNTAINS 0.802748019 1.222975173 1.205474439 ENO RIVER ENRI PIEDMONT 0.619621341 1.932316384 36.05189498 FALLS LAKE FALA PIEDMONT 0.952996869 1.810468171 13.80908625 FORT FISHER FOFI COASTAL 0.93847867 2.308648738 28.32575672 FORT MACON FOMA COASTAL 0.900636442 1.12905223 30.16176556 GOOSE CREEK GOCR COASTAL 0.889054101 0.897343556 4.190384097 GORGES GORG MOUNTAINS 0.91229934 1.435055019 4.071486262 GRANDFATHER MOUNTAIN GRMO MOUNTAINS 0.929972426 0.786856639 0.812751839 HAMMOCKS BEACH HABE COASTAL 0.957566831 1.119022005 8.335759816 HAW RIVER HARI PIEDMONT 0.986636243 2.446579317 -0.013347613 HANGING ROCK HARO PIEDMONT 0.448692115 0.583369919 4.66310663 JONES LAKE JONE PIEDMONT 0938498425 1.778881438 21.71120035 JORDAN LAKE JORD COASTAL 0.950325975 1.469189504 3.327493732 JOCKEY'S RIDGE JORI COASTAL 0.985873483 2.000247982 11.99466691 KERR LAKE KELA PIEDMONT 0.8851289 1.402982853 31.56646627 LAKE JAMES LAJA MOUNTAINS 0.982883453 1.142852652 17.69144941 LAKE NORMAN LANO PIEDMONT 0.916733318 2.022467394 34.72331682 LAKE WACCAMAW LAWA COASTAL 0.860974038 0.471981221 9.998815331 LUMBER RIVER LURI COASTAL 0.610200163 2.068296901 2.598378055 MAYO RIVER MARI PIEDMONT 0.023040069 0.719718659 6.242136949 MERCHANTS MILLPOND MEMI COASTAL 0.825581864 1.874074059 7.029779688 MEDOC MOUNTAIN MEMO COASTAL 0.466315884 1.194807857 6.843342728 MOUNT JEFFERSON MOJE MOUNTAINS 0.032438296 0.478293306 10.91719SS7 MOUNT MITCHELL MOMI MOUNTAINS 0.961718442 2.013339204 0.121708407 MORROW MOUNTAIN MOMO PIEDMONT 0.979754092 2.06862387 0.951430335 NEW RIVER NERI MOUNTAINS 0.951363306 4.011990736 -1.661882125 OCCONEECHEE MOUNTAIN OCMO PIEDMONT 0.73347395 1.64899496 7.732159693 PETTIGREW PETT COASTAL 0.806075419 0.463937937 2.088786359 PILOT MOUNTAIN PIMO PIEDMONT 0.967784724 3.932517103 11.9271766 RAVEN ROCK RARO COASTAL 0.920099254 2.355839007 3.903020113 SINGLETARY LAKE SILA COASTAL 0.654587453 0.364280517 1.720942503 SOUTH MOUNTAINS SOMO MOUNTAINS 0.554531513 4.709919373 5.904875752 STONE MOUNTAIN STMO MOUNTAINS 0.866579741 2.054311807 8.329912385 WEYMOUTH WOODS SANDHILLS WEWO PIEDMONT 0.85722248 1.244288316 8.065177761 WILLIAM B. UMSTEAD WIUM PIEDMONT 0.259500919 1.756186476 28.35780298 Table 2. Results of analysis comparing Visitors per Vehicle Multiplier Visitation Estimates (VPVM-VE) and Mobile Tracking Data (MTD) for each NC state park unit in the calendar year 2022. R-squared values, slope values, and intercept values depict the nature of the relationship between the two data types. REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 8 PAGE 9 Figure 2. Correlation Histogram - VPVM-VE vs. MTD Monthly Visitation 2022. This histogram displays the distribution of R-squared values derived from the regression analysis of VPVM-VE compared to MTD for state parks units in 2022. The blue bars represent the frequency of parks within specified R-squared value ranges, illustrating the strength of the correlation between the two datasets, with values ranging from 0 (indicating no correlation) to 1 (indicating perfect correlation). The orange cumulative probability curve shows the cumulative percentage of parks with R-squared values below or equal to each specific value along the x-axis. Notably, approximately 55% of parks have R-squared values below 0.9, indicating that many parks exhibit weak correlations between VPVM-VE and MTD. REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 9 Two histograms, which are a special type of bar chart that shows how often something happens, illustrate the occurrences of R-squared values (Figure 2) and slope values (Figure 3) derived from the regression analysis. For example, each bar in the correlation figure shows how many parks fall into a certain range of correlation values. The taller the bar, the more parks have that level of correlation. If a bar is short, it means fewer parks have that value. These histograms help us understand how common or rare certain relationships are across parks in terms of their visitation data. PAGE 9 Figure 6. Lake Norman State Park Regression Result. This regression plot shows the relationship between VPVM-VE (in thousands) and MTD (in thousands) for Lake Norman State Park in the Piedmont region of North Carolina. The high R-squared value of 0.92 indicates a strong correlation between the two datasets. The moderately high slope of 2.02 suggests that for every 1,000 mobile visits, there are approximately 2,020 vehicle visits, indicating that the VPVMVE estimates are higher than expected based on MTD, but less so than at New River State Park (Figure 5). The highest intercept of 34.72 means that even when no mobile visits are recorded, the estimated baseline for vehicle visits is 34,720, suggesting that VPVM-VE may be overestimating visitation at times when mobile data is missing or underreported. The data points represent specific months in 2022, color-coded by month, with the symbol shape indicating season. The diagonal 1:1 line shows a perfect correlation between vehicle visits and mobile visits. REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 16 01020304050 Mobile Vis. - x1000 0 10 20 30 40 50 60 70 80 90 100 Vehicle Vis. - x1000 LANO (Ped) R 2 0.92 Slope : 2.02 Intercept : 34.72 PAGE 9 Figure 7 shows the comparison of visitation estimates for Morrow Mountain State Park (MOMO). The R-squared value for Morrow Mountain State Park (MOMO) is only 0.03, indicating that just 3% of the variance in vehicle visits can be explained by mobile visits. This extremely low correlation suggests that either: • MTD fails to effectively capture the visitation patterns, or • The raw Vehicle Counts do not accurately capture all visitors. • Both datasets may not adequately capture the true visitation patterns. The low R-squared value raises significant concerns about the reliability of both datasets, as they do not align to provide an accurate picture of park visitation. Given this weak relationship, the slope and intercept from the regression analysis are not meaningful indicators of visitation and should not be used in an HVE. Determining the reasons behind this discrepancy will be essential for improving the accuracy of visitation estimates at Morrow Mountain State Park. REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 17 010203040 Mobile Vis. - x1000 0 10 20 30 40 Vehicle Vis. - x1000 MOMO (Ped) R 2 0.03 Slope : 0.48 Intercept : 10.92 Figure 7. Morrow Mountain State Park Regression Result. This regression plot shows the relationship between VPVM-VE (in thousands) and MTD (in thousands) for Morrow Mountain State Park in the Piedmont region of North Carolina. The very low R-squared value of 0.03 indicates a very weak correlation between the two datasets, meaning that mobile visits (MTD) do not explain the variation in vehicle visits (VPVM-VE) at this park. This suggests that either the mobile data is not accurately capturing visitation at this park, or the vehicle counts do not reflect actual visitor behavior. The slope and intercept values are less meaningful in this case due to the weak correlation. Notably, data from December through May lie above the slope line, suggesting that during these months, vehicle visits are higher than what the mobile data predicts. In contrast, data from June through November fall below the slope line, indicating that vehicle visits are lower than what MTD estimates would suggest for those months. The data points represent specific months in 2022, color-coded by month, with the symbol shape indicating season. The diagonal 1:1 line shows a perfect correlation between vehicle visits and mobile visits. PAGE 9 3.2. Parks with Low R-squared Values Here is a list of parks that have R-squared values less than 0.8. For these parks, 2022 MTD offers a relatively poor reflection of vehicle count visitation estimates: REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 18 Piedmont Region • Occoneechee Mountain State Natural Area (OCMO): R-squared = 0.733 • Eno River State Park (ENRI): R-squared = 0.620 • Haw River State Park (HARI): R-squared = 0.449 • William B. Umstead State Park (WIUM): R-squared = 0.260 • Morrow Mountain State Park (MOMO): R-squared = 0.032 • Mayo River State Park (MARI) : R-squared = 0.023 Coastal Region • Singletary Lake State Park (SILA): R-squared = 0.655 • Lumber River State Park (LURI): R-squared = 0.610 • Merchants Millpond State Park (MEMI): R-squared = 0.466 • Carolina Beach State Park (CABE): R-squared = 0.356 • Dismal Swamp State Park (DISW): R-squared = 0.303 • Cliffs of the Neuse State Park (CLNE): R-squared = 0.292 • Carvers Creek State Park (CACR): R-squared = 0.180 Mountains Region • South Mountains State Park (SOMO): R-squared = 0.555 Low R-squared values indicate that mobile tracking data explains less than half of the variance in vehicle count estimates (VPVM-VE) for these parks. In simpler terms, this means that the mobile tracking data does not account for a significant portion of the differences in vehicle counts observed at these parks. For instance, if the R-squared value is 0.3, it suggests that only 30% of the changes in vehicle counts can be explained by the mobile tracking data, leaving 70% of the variance unexplained by this data source. This finding underscores the limitations of relying solely on mobile tracking data for accurate visitation estimates across the park system, as it may not fully capture visitor patterns influenced by various factors. Specifically, the low R-squared values suggest that mobile tracking data does not account for factors affecting vehicle counts, such as incomplete vehicle count capture due to unrecorded entrances and multipliers that may not reflect actual visitation patterns, particularly in parks with unique visitor demographics. Additionally, limitations in mobile tracking data itself— such as some visitors not carrying devices, demographic factors that reduce mobile technology usage, and untracked short visits—further contribute to the discrepancy between mobile visits and vehicle counts. Thus, both data sources have limitations that affect their ability to explain variance in visitation estimates. PAGE 9 3.3. Parks with High R-squared Values For the parks with R-squared values above 0.8, the intercepts range from -1.66 to 34.72, with a median of 8.07 and an average of 10.22. In analyzing these intercepts, it is important to note that they typically represent a single intercept from a regression model that combines both Mobile Tracking Data (MTD) and Visitors Per Vehicle Multiplier - Visitation Estimates (VPVM-VE). This single intercept signifies the expected visitation levels when both datasets indicate zero visits, providing a baseline for understanding visitation patterns. Here are a few reasons why intercepts may be high in the context of comparing MTD and VPVM-VE: REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 19 1. VPVM-VE Multipliers: The multipliers applied to vehicle counts in the VPVM-VE can inflate the baseline visitation numbers. These multipliers are intended to estimate total visitation based on assumed vehicle occupancy, while the MTD has no multiplier applied to adjust the number of phones captured to an estimation of visitation. 2. Regular Visitors Without Phones: Certain parks may attract a dedicated visitor base that regularly returns but does not use mobile devices. This behavior can result in a higher baseline visitation level reflected in the intercept, as these visitors are not captured in the MTD. 3. Limited Cell Phone Coverage: In some areas, poor or nonexistent cell phone coverage may prevent accurate tracking of mobile visits. This can lead to an underrepresentation of actual visitation in the MTD, contributing to a higher intercept when compared to VPVM-VE. 4. Counts of Non-Visitor Vehicles: VPVM-VE may include counts of vehicles that are not necessarily associated with park visitors, such as maintenance or delivery vehicles. This inclusion can artificially elevate the baseline visitation estimate. These factors highlight the complexities involved in accurately estimating park visitation and underscore the importance of understanding the underlying data sources and methodologies. PAGE 9 4. UTILIZING STRONG VISITORS PER VEHICLE MULTIPLER ESTIMATES (VPVM-VE) AND MOBILE TRACKING DATA (MTD) CORRELATIONS TO REFINE VISITATION ESTIMATES For parks that are highly correlated, even if the slope suggests VPVM-VE underestimates or overestimates actual visitation or intercept implies that there is always a significant VPVM-VE, even if the MTD shows zero activity, the observed relationship can potentially be utilized to refine visitation estimates through the creation of Adjustment Factors (AFs) and Hybrid Visitation Estimates (HVE). REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 20 4.1. Three Ways to Utilize the Observed Relationship The approach for calculating AFs and HVEs should depend on the context of the data and the availability of MTD. Option 1 When no additional mobile data is acquired, this formula takes VPVM−VE and creates a HVE: HVE = (VPVM−VE × Adjustment Factor) + Intercept This approach leverages future VPVM-VE to generate HVE, allowing the model to account for the relationship between vehicle counts and mobile visits. The intercept provides a baseline visitation level that may not be captured by vehicle counts alone. Option 2 If MTD is regularly acquired and it is a reliable indicator of visitation at this park, a preferred formula may be: HVE = (MTD × Adjustment Factor) + Intercept This method adjusts raw MTD to produce visitation estimates. The adjustment factor modifies MTD to reflect expected visitation levels, while the intercept accounts for baseline visitation that may exist even when mobile visits are low or absent. Option 3 Employing both datasets could prove useful. However, to utilize this formula for future estimates, one would need to regularly obtain MTD or assume similar visitation patterns from year to year, consistently applying the 2022 MTD figures in the equation: HVE = (MTD × w1) + (VPVM−VE × w2) + Intercept This approach allows for a weighted combination of both datasets, ensuring a robust visitation estimate that reflects the strengths of each data source. By applying different weights to MTD and VPVM-VE, this formula adapts to varying levels of data reliability and correlation between the two sources. The inclusion of the intercept enhances the accuracy of HVE by accounting for baseline visitation levels, making this formula a flexible option for parks with consistent mobile tracking data and varying visitation patterns. PAGE 9REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 21 4.2. Hybrid Visitation Estimates (HVE) and Adjustment Factor Because Option 1 from the previous section uses the least resources as it requires no additional mobile data, this report will cover creating HVE based solely on future VPVM−VE using the regression equations coefficient in the calculation: HVE = (VPVM−VE × Adjustment Factor) + Intercept This section outlines the development of an adjustment factor to improve the accuracy of the Visitors Per Vehicle Multiplier - Visitation Estimates (VPVM-VE) by incorporating insights from Mobile Tracking Data (MTD). By leveraging the observed relationships between VPVM-VE and MTD, we can create park-specific adjustment factors that refine visitation estimates previously based solely on vehicle counts. This approach aims to provide a more accurate representation of actual visitor behavior within each park. Parks exhibiting an R-squared value above 0.8 are good candidates for adjustment and the creation of a Hybrid Visitation Estimate (HVE), given the strong correlations observed. For each highly correlated park, we can tailor the adjustment factor based on the slope of the regression to enhance the accuracy of visitation estimates. We propose the following formula for the Adjustment Factor: Adjustment Factor = 1 − ((slope − k ) / 10) Where k is a reference value that varies depending on the slope: • For slopes between 0.75 and 1.25, k = 1 • For slopes above 1.25, k = 0.95 • For slopes below 0.75, k = 1.05 The distance between the slope and k informs the magnitude of the adjustment. The further the slope is from k, the more significant the adjustment, allowing the formula to flexibly respond to the specific visitation dynamics of each park. A slope of 1 indicates that for every additional mobile visit, there is an equivalent increase in vehicle visits, suggesting that mobile data accurately captures visitation patterns and no adjustment is needed. By using k = 1, we reinforce this reliability, ensuring the adjustment factor does not significantly alter the visitation estimate. When slopes fall below 1, it means that for each additional mobile visit recorded, there are fewer vehicle visits than expected based on vehicle count estimates. This suggests that the mobile tracking data (MTD) isn’t capturing all the visitors at the park for several reasons. Many people might not have mobile devices with them or may have their location services turned off, leading to uncounted visits. Others might avoid using mobile apps due to privacy concerns or simply because they find them hard to use. Additionally, certain parks might attract visitors who are less likely to use technology, like older adults or families. Quick visits, such as stopping for a rest or a picnic, might also be missed by mobile tracking, leading to significant gaps between vehicle counts and mobile visits. Furthermore, the way vehicle counts are estimated may not consider all the people traveling in each vehicle, which could further confuse the comparison between the two data sources. PAGE 9REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 22 In these cases, we should increase the Hybrid Visitation Estimate (HVE) to better match the higher visitor numbers indicated by vehicle counts. This adjustment effectively pushes the slope of the regression line closer to an idealized 1-to-1 relationship, thereby increasing the HVE relative to the VPVM-VE. On the other hand, when the slope is above 1, it indicates that each additional mobile visit corresponds to a greater number of vehicle visits, meaning that the number of vehicle visits increases at a faster rate than the number of mobile visits. This indicates that mobile tracking may not fully capture the total number of visitors for similar reasons, such as differences in who visits the park and how often they use technology. In this situation, the adjustment factor is used to lower the HVE, acknowledging that vehicle counts might reflect a higher level of visitation than what mobile data shows. By doing so, the HVE effectively pushes the slope of the regression line towards the idealized 1-to-1 relationship, decreasing the HVE relative to the VPVM-VE. For example, if we set k = 0.95, the adjustment accounts for the disparity between the two datasets, ensuring that the HVE is calibrated to more accurately reflect actual park usage without inflating the visitor numbers beyond what the vehicle counts indicate. Using this formula, the Hybrid Visitation Estimate (HVE) can be derived for each park meeting the R-squared and slope criteria: HVE = VPVM−VE × Adjustment Factor + Intercept This formula applies the observed relationships in each park to adjust visitation estimates, ensuring they are better aligned with actual visitor patterns as indicated by both data sources. PAGE 9REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 23 PARK REGION RSQUARED SLOPE INTERCEPT K ADJUSTMENT FACTOR (AF) VPVM−VE (JAN 2022) MTD (JAN 2022) NERI MOUNTAINS 0.951 4.01 -1.66 0.95 0.69 6,664 3,121 PIMO PIEDMONT 0.968 3.93 11.93 0.95 0.70 51,015 8,430 HARO PIEDMONT 0.987 2.45 -0.01 0.95 0.85 23,461 8,961 RARO COASTAL 0.920 2.36 3.90 0.95 0.86 16,553 6,353 FOFI COASTAL 0.938 2.31 28.33 0.95 0.86 26,896 3,510 MOMI MOUNTAINS 0.980 2.07 0.95 0.95 0.89 2,938 1,084 STMO MOUNTAINS 0.867 2.05 8.33 0.95 0.89 13,652 2,349 LANO PIEDMONT 0.917 2.02 34.72 0.95 0.89 44,371 4,809 CRMO PIEDMONT 0.971 2.02 12.38 0.95 0.89 43,088 14,439 MOJE MOUNTAINS 0.962 2.01 0.12 0.95 0.89 2,386 1,841 JORD PIEDMONT 0.986 2.00 11.99 0.95 0.89 22,915 12,069 MEMO COASTAL 0.826 1.87 7.03 0.95 0.91 11,556 1,444 FALA PIEDMONT 0.953 1.81 13.81 0.95 0.91 18,432 7,325 JORI COASTAL 0.938 1.78 21.71 0.95 0.92 24,104 4,051 JONE COASTAL 0.950 1.47 13.33 0.95 0.95 6,006 2,220 GORG MOUNTAINS 0.912 1.44 4.07 0.95 0.95 5,230 2,095 KELA PIEDMONT 0.885 1.40 31.57 0.95 0.95 21,788 5,527 WEWO PIEDMONT 0.857 1.24 8.07 10.98 10,058 1,619 ELKN MOUNTAINS 0.803 1.22 1.21 10.98 2,712 1,312 LAJA MOUNTAINS 0.983 1.14 17.69 10.99 22,706 4,235 FOMA COASTAL 0.901 1.13 30.16 10.99 44,793 12,530 HABE COASTAL 0.958 1.12 8.34 10.99 9,900 2,300 GOCR COASTAL 0.889 0.90 4.19 11.01 4,852 2,839 GRMO MOUNTAINS 0.930 0.79 0.81 11.02 2,120 3,012 CHRO MOUNTAINS 0.963 0.77 0.93 11.02 12,121 13,228 LAWA COASTAL 0.861 0.47 10.00 1.05 1.06 10,149 2,427 PETT COASTAL 0.806 0.46 2.09 1.05 1.06 2,647 838 4.3. Table of Proposed Adjustment Factors Table 3 presents parks that meet the criteria of having an R-squared value greater than 0.8. The table includes each park, its Region, R-squared value, Slope, and Intercept, along with their respective k values and calculated Adjustment Factors (AF). The final two columns on the right are the original VPVM−VE and MTD for 2022. Many of these fields will be used as input for HVE calculations in the following tables in Section 5. Table 3. Visitation adjustment factors for NC State Park units with R-squared values > 0.8 for comparisons of Visitors per Vehicle Multiplier - Visitation Estimates (VPVM-VE) and Mobile Tracking Data (MTD). REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 23 PAGE 9 4.4. Table of Hybrid Visitation Estimates (HVE) with and without Inclusion of the Intercepts Given the discussion of reasons why intercepts may be high in Section 4.3, it is important to examine how the intercept influences the increase and percent increase for each park. In Table 4 below, the green field has the parks Intercept, AF, Slope, and VPVM−VE estimates for January 2022. The Orange model presents the Hybrid Visitation Model 1 (HVE1) which does not utilize the intercept. The Purple model presents the Hybrid Visitation Model 1 (HVE1) which does utilize the intercept. For each model, there is an HVE estimate using January 2022 data and the formula HVE = VPVM−VE × Adjustment Factor, both with and without the intercept. Then, the difference between the HVEs and VPVM−VE, as well as the percentage increase of the HVEs from the VPVM−VE, are provided in the table below. For HVE1, the HVE1 % Increase (Jan 2022) ranged from -31% to 2%, while the HVE2 % Increase (Jan 2022) ranged from -15% to 140%. For most parks, the percent increase of the HVE is higher (and in many cases much higher) for HVE2 (with the intercept) compared to HV1(without the intercept), with exceptions for only HARO and NERI, which have negative intercepts. Given the observed influence of the intercept in the HVE2 % Increase and the reasons discussed in section 4.3 for why intercepts may be high in the context of comparing MTD and VPVM-VE, a desired estimate may land somewhere between these two models. Table 4. Comparison of Hybrid Visitation Estimates (HVE) with and without the Inclusion of the Intercept for NC State Park Units (January 2022). This table compares two Hybrid Visitation Estimate (HVE) models calculated for North Carolina state park units in January 2022, with and without the inclusion of the intercept from the regression results. The table provides key data for each state park unit (highlighted in green), including the intercept, adjustment factor (AF), slope, and VPVM-VE data used in the regression analysis. It then presents two HVE models: (in orange) HVE1, which excludes the intercept, and (in purple) HVE2, which includes the intercept. The table also shows the difference and percentage increase for both HVE1 and HVE2 in comparison to the VPVM-VE, offering insights into how the inclusion of the intercept impacts the refinement of visitation estimates across various state park units. REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 24 PAGE 9REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 25 PARK INTERCEPT AF SLOPE VPVM−VE (JAN 2022) HVE1 = [VPVM−VE (JAN 2022) *AF] HVE1 (JAN 2022) INCREASE HVE1 % INCREASE (JAN 2022) HVE2 = [VPVM−VE (JAN 2022) *AF] + INTERCEPT HVE2 (JAN 2022) INCREASE HVE2 % INCREASE (JAN 2022) NERI -1.66 0.69 4.01 6,664 4,623 -2,041 -31% 2,962 -3,702 56% PIMO 11.93 0.70 3.93 51,015 35,800 -15,215 -30% 47,727 -3,288 -6% HARO -0.01 0.85 2.45 23,461 19,950 -3,511 -15% 19,937 -3,524 -15% RARO 3.90 0.86 2.36 16,553 14,226 -2,327 -14% 18,129 1,576 10% FOFI 28.33 0.86 2.31 26,896 23,242 -3,654 -14% 51,568 24,672 92% MOMI 0.95 0.89 2.07 2,938 2,609 -329 -11% 3,561 623 21% STMO 8.33 0.89 2.05 13,652 12,144 -1,508 -11% 20,474 6,822 50% LANO 34.72 0.89 2.02 44,371 39,612 -4,759 -11% 74,336 29,965 68% CRMO 12.38 0.89 2.02 43.088 38,489 -4,599 -11% 50,865 7,777 18% MOJE 10.12 0.89 2.01 2,386 2,132 -254 -11% 2,254 -132 -6% JORD 11.99 0.89 2.00 22,915 20,508 -2,407 -11% 32,503 9,588 42% MEMO 17.03 0.91 1.87 11,556 10,488 -1,068 -9% 17,518 5,962 52% FALA 13.81 0.91 1.81 18,432 16,846 -1,586 -9% 30,655 12.223 66% JORI 21.71 0.92 1.78 24,104 22,106 -1,998 -8% 43,817 19,713 82% JONE 13.33 0.95 1.47 6,006 5,694 -312 -5% 9,022 3.016 50% GORG 4.07 0.95 1.44 5,230 4,976 -254 -5% 9,048 3,818 73% KELA 31.57 0.95 1.40 21,788 20,801 -987 -5% 52,368 30.580 140% WEWO 8.07 0.98 1.24 10,058 9,812 -246 -2% 17,877 7.819 78% ELKN 1.21 0.98 1.22 2,712 2,652 -60 -2% 3,857 1,145 42% LAJA 17.69 0.99 1.14 22,706 22,382 -324 -1% 40,073 17.367 76% FOMA 30.16 0.99 1.13 44,793 44,215 -578 -1% 74,377 29,584 66% HABE 8.34 0.99 1.12 9,900 9,782 -118 -1% 18,118 8.218 83% GOCR 4.19 1.01 0.90 4,852 4,902 50 1% 9,092 4,240 87% GRMO 0.81 1.02 0.79 2,120 2,165 45 2% 2,978 1858 40% CHRO 0.93 1.02 0.77 12,121 12,398 277 2% 13,328 1,207 10% LAWA 10.00 1.06 0.47 10,149 10,736 587 6% 20,734 10,585 104% PETT 2.09 1.06 0.46 2,647 2,802 155 6% 4,891 2,244 85% Table 4. Comparison of Hybrid Visitation Estimates (HVE) with and without the Inclusion of the Intercept for NC State Park Units (January 2022). This table compares two Hybrid Visitation Estimate (HVE) models calculated for North Carolina state park units in January 2022, with and without the inclusion of the intercept from the regression results. The table provides key data for each state park unit (highlighted in green), including the intercept, adjustment factor (AF), slope, and VPVM-VE data used in the regression analysis. It then presents two HVE models: (in orange) HVE1, which excludes the intercept, and (in purple) HVE2, which includes the intercept. The table also shows the difference and percentage increase for both HVE1 and HVE2 in comparison to the VPVM-VE, offering insights into how the inclusion of the intercept impacts the refinement of visitation estimates across various state park units. PAGE 9 50 100 150 50 100 150 Car Count Vis. - x1000 JORI (Cst) R20.94 Slope : 1.78 Intercept : 21.71 10 20 30 5 10 15 20 25 30 LAWA (Cst) R20.86 Slope : 0.47 Intercept : 10.00 5101520 5 10 15 20 LURI (Cst) R20.61 Slope : 2.07 Intercept : 2.60 51015 5 10 15 Car Count Vis. - x1000 MEMI (Cst) R20.47 Slope : 1.19 Intercept : 6.84 5101520 5 10 15 20 MEMO (Cst) R20.83 Slope : 1.87 Intercept : 7.03 2468 2 4 6 8PETT (Cst) R20.81 Slope : 0.46 Intercept : 2.09 10 20 30 40 Mobile Data Vis. - x1000 10 20 30 40 Car Count Vis. - x1000 RARO (Cst) R20.92 Slope : 2.36 Intercept : 3.90 1234 Mobile Data Vis. - x1000 1 2 3 4SILA (Cst) R20.65 Slope : 0.36 Intercept : 1.72 20 40 60 80 Mobile Data Vis. - x1000 20 40 60 80 CRMO (Ped) R20.97 Slope : 2.02 Intercept : 12.38 Panel B - Coastal & Piedmont REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 32 PAGE 9REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 29 20 40 60 80 100 20 40 60 80 100 Car Count Vis. - x1000 ENRI (Ped) R20.62 Slope : 1.93 Intercept : 36.05 50 100 150 200 50 100 150 200 FALA (Ped) R 20.95 Slope : 1.81 Intercept : 13.81 2468 2 4 6 8 HARI (Ped) R20.45 Slope : 0.58 Intercept : 4.66 20 40 60 80 100 120 20 40 60 80 100 120 Car Count Vis. - x1000 HARO (Ped) R20.99 Slope : 2.45 Intercept : -0.01 100 200 300 100 200 300 JORD (Ped) R20.99 Slope : 2.00 Intercept : 11.99 50 100 150 200 50 100 150 200 KELA (Ped) R20.89 Slope : 1.40 Intercept : 31.57 20 40 60 80 100 Mobile Data Vis. - x1000 20 40 60 80 100 Car Count Vis. - x1000 LANO (Ped) R20.92 Slope : 2.02 Intercept : 34.72 5101520 Mobile Data Vis. - x1000 5 10 15 20 MARI (Ped) R20.02 Slope : 0.72 Intercept : 6.24 10 20 30 40 Mobile Data Vis. - x1000 10 20 30 40 MOMO (Ped) R 20.03 Slope 0.48 Intercept : 10.92 Panel C - Piedmont REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 33 PAGE 9REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 29 5101520 5 10 15 20 Car Count Vis. - x1000 OCMO (Ped) R20.73 Slope : 1.65 Intercept : 7.73 50 100 150 50 100 150 PIMO (Ped) R20.97 Slope : 3.93 Intercept : 11.93 51 01 5 5 10 15 WEWO (Ped) R20.86 Slope : 1.24 Intercept : 8.07 20 40 60 80 100 20 40 60 80 100 Car Count Vis. - x1000 WIUM (Ped) R20.26 Slope : 1.76 Intercept : 28.36 20 40 60 20 30 40 50 60 70 CHRO (Mnt) R20.96 Slope : 0.77 Intercept : 0.93 246 1 2 3 4 5 6 7ELKN (Mnt) R20.80 Slope : 1.22 Intercept : 1.21 510152025 Mobile Data Vis. - x1000 5 10 15 20 25 Car Count Vis. - x1000 GORG (Mnt) R20.91 Slope : 1.44 Intercept : 4.07 5101520 Mobile Data Vis. - x1000 5 10 15 20 GRMO (Mnt) R20.93 Slope : 0.79 Intercept : 0.81 20 40 60 80 Mobile Data Vis. - x1000 20 40 60 80 LAJA (Mnt) R20.98 Slope : 1.14 Intercept : 17.69 Panel D - Piedmont & Mountains REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 34 PAGE 9REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 29 51015 5 10 15 Car Count Vis. - x1000 MOJE (Mnt) R20.96 Slope : 2.01 Intercept : 0.12 20 40 60 20 40 60 MOMI (Mnt) R 20.98 Slope : 2.07 Intercept : 0.95 20 40 60 80 Mobile Data Vis. - x1000 20 40 60 80 NERI (Mnt) R20.95 Slope : 4.01 Intercept : -1.66 10 20 30 40 50 Mobile Data Vis. - x1000 10 20 30 40 50 Car Count Vis. - x1000 SOMO (Mnt) R20.55 Slope : 4.71 Intercept : 5.90 10 20 30 40 50 Mobile Data Vis. - x1000 10 20 30 40 50 STMO (Mnt) R20.87 Slope : 2.05 Intercept : 8.33 Panel E - Mountains REFINING NC STATE PARK UNIT VISITATION ESTIMATES USING MOBILE TRACKING DATA PAGE 35