The Greater Tokyo Area is the world's most populated metropolitan area. According to the Japan Statistics Bureau, the Kanto Major Metropolitan Area, one of various definitions of the Greater Tokyo Area, is defined as the area that consists of all municipalities that have > 1.5 % of their population (aged 15 and above) commuting to designated cities (Chiba, Kawasaki, Sagamihara, Saitama and Yokohama) or the 23 special wards of Tokyo Metropolis. The area's population was about 38 million in 2020, according to the Census. To our knowledge, no vehicle-based measurements of CH₄ and C₂H₆ have been ever reported within this area. In designing driving routes, we therefore prioritized coverage of large extent of the Greater Tokyo Area. In addition, particular focus was the area around the Yoyogi site (35.66° N, 139.68° E). Atmospheric CH₄ mole fraction measurement (not shown in this study) is ongoing at the site located in the central part of Tokyo, with the surrounding main land cover being residential buildings (Ishidoya et al., 2020; Sugawara et al., 2021). The Tokyo Bay area was also of interest, as part of the area is occupied by Japan's two major industrial zones (Keihin and Keiyo Industrial Zones) that hold heavy industries (e.g., steel mills, oil refineries, chemical plants and electricity generation). In the area, there is also Tokyo Bay-side Landfill in the Port of Tokyo, which is the final disposal site of solid wastes from the 23 special ward areas of Tokyo. In addition, the 23 special wards have 13 wastewater treatment plants, and they were also targets of the present measurement surveys. CH₄ emissions from the individual landfill and wastewater facilities are however beyond the scope of this study and will be investigated in a separate paper.

Our vehicle-based mobile measurements were conducted during daytime on 15 d (18–22 September, and 2–6, 16 and 18–21 October 2023: 14 weekdays and 1 Saturday). The measurement survey covered populated areas of Tokyo Metropolis as well as areas around the Tokyo Bay in Chiba and Kanagawa Prefectures, as shown in Fig. 1. The measurements were made on public roads only. The driving distance on each measurement day ranged from 71 to 208 km, depending on focus areas, and the total distance was 2012 km.

To measure atmospheric mole fractions of CH₄ and C₂H₆, we used a MIRA Ultra gas analyzer (Aeris Technologies, USA). The analyzer is a mid-infrared absorption spectrometer with a multi-pass cell (60 cm³ volume). The cell pressure is maintained at ∼ 240 mbar by an internal pump. More descriptions on the analyzer were previously given in Travis et al. (2020) and Commane et al. (2023).

To determine the CH₄ and C₂H₆ mole fractions in sample air, we prepared a suite of dry-air standard gases containing CH₄ and C₂H₆ at ambient to urban elevated mole fraction levels, ranging approximately from 2 to 5 µmolmol-1 (hereafter denoted as ppm) for CH₄ and 1 to 10 nmol mol⁻¹ (hereafter denoted as ppb) for C₂H₆. These standard gases were produced gravimetrically in 9.4 L aluminum cylinders by Taiyo Nippon Sanso JFP Corporation. The CH₄ and C₂H₆ mole fractions reported in this study are traceable to those in these standard gases. Prior to the measurements, the CH₄ mole fractions in the standard gases were measured by a cavity ring-down spectroscopy analyzer (G-2401, Picarro Inc., USA). The measurements confirmed that the CH₄ mole fractions in the standard gases show excellent linearity with reproducibility of < 1.0 ppb. The standard gases were also measured for C₂H₆ by a preconcentration and refocusing gas-chromatography mass-spectrometry (GC-MS) system (Umezawa et al., 2025b). The measurement results showed a good linearity of the standard gases with reproducibility of < 0.1 ppb.

The repeatability of the MIRA Ultra analyzer has been evaluated by Allan-Werle variance plots (e.g., Commane et al., 2023). We measured compressed air in a 47 L cylinder for about 24 h and calculated the Allan-Werle variance (not shown). The same experiment was made twice: 6–7 September 2023 and 21–22 February 2024. The first experiment was soon after delivery of the analyzer and the second was made after the modified version of the software (data acquisition frequency increased) being implemented. Repeatability of MIRA Ultra for 100 s was 0.02 ppb for CH₄ and 0.004 ppb for C₂H₆, which were later improved to 0.007 ppb for CH₄ and 0.003 ppb for C₂H₆. These values exceed those reported by Commane et al. (2023) for the same product.

The standard gases were measured periodically to evaluate longer-term stability of the analyzer. The instrument showed good stability during the mobile measurement campaign period in Tokyo (18 September–21 October 2023), with the measured CH₄ values of the standard gases agreed to the nominal value within ∼ 5 ppb at baseline level (∼ 2 ppm). In contrast, the measured values for the standard gases with higher CH₄ mole fractions (up to ∼ 5 ppm) showed differences up to ∼ 25 ppb. As this possible bias (∼ 0.025 ppm) is relatively small in comparison to the observed variability (> 1 ppm excess values with respect to the baseline), we apply no corrections for CH₄ measurements during the Tokyo mobile campaign measurements. It was also shown that measured C₂H₆ values by MIRA Ultra varied significantly from day to day. This suggests that raw C₂H₆ values reported from the analyzer are not compatible among different measurement days. We however confirmed that the measurement span covered by the standard gases remained constant (for the mole fraction range approximately from 1 to 10 ppb) over the course of measurements, indicating that the excess values in the C₂H₆ mole fraction were robust. We therefore report excess values only for C₂H₆.

We evaluated two MIRA Ultra analyzers of two institutions (National Institute for Environmental Studies, NIES and Osaka Metropolitan University, OMU) for sensitivity to humidity as previously reported by Commane et al. (2023). The NIES and OMU analyzers were operated for the measurements in Tokyo (this study) and Osaka (Ueyama et al., 2025), respectively. Figure 2 displays the CH₄ and C₂H₆ mole fraction dependence on the amount of water vapor. The mole fractions here are reported values by the instrument and applied no corrections. In this experiment, a compressed dry air was introduced into the analyzer. The air passed through a moisture exchanger (ME-110-72COMP-4, Perma Pure LLC) housed in a food container in which a humidor pack (84 % RH, Boveda Inc.) was placed. Warm up of the container increased humidity inside the container, which also increased the amount of water vapor in the sample flow of the air via the moisture exchanger. To examine response with low humidity, silica gel was placed in the container instead of the humidor pack to dry the sample air.

As shown in Fig. 2, CH₄ measurement by the analyzer showed a quadratic curve (grey and black lines) expressed by: 1NIES:Y=-0.05478+8.4828×10-6X-3.2856×10-10X22OMU:Y=-0.020178+4.0952×10-6X-2.07786×10-10X2 Commane et al. (2023) reported the humidity response of the MIRA Ultra analyzer, but interestingly, their experiment indicated a different curve in shape. They reported that the CH₄ value from the analyzer showed a quadratic but monotonically decreased curve with increasing humidity. In contrast, our results in Fig. 2 suggest that the measured CH₄ value has a plateau at water vapor range of approximately 5000–15 000 ppm. For the NIES MIRA Ultra, in the water vapor range of 7500–18 000 ppm, the measured CH₄ value falls within -0.01 ppm relative to the top of the quadratic convex curve positioned at ∼ 13 000 ppm water vapor. This implies that measurements with water vapor in the above target range would report CH₄ values with better reproducibility. We therefore consider that (1) calibration of the analyzer should be made by reference gases humidified to as close as the target range, and (2) CH₄ values reported by the analyzer should be corrected according to the water vapor values.

For C₂H₆, the measured value from the NIES MIRA Ultra showed a slight increase with increasing humidity expressed by: 3NIES:Y=-0.6521+4.8243×10-5X In the above water vapor range (7500–18 000 ppm), the C₂H₆ values are expected to fall within ±0.3 ppb from the nominal value. The C₂H₆ values reported by the analyzer should be corrected accordingly. In contrast, the MIRA Ultra analyzer of OMU did not show a consistent trend between experiments when the water vapor was increased and decreased (Fig. 2, bottom panel). Note that measurements by OMU MIRA Ultra are not included in this study and presented in the companion paper (Ueyama et al., 2025). It should be noted that the water response curve, as presented in this study, is likely different from instrument to instrument even within the same model product MIRA Ultra; we therefore recommend careful evaluation of each analyzer for high-accuracy measurements.

The vehicle-based measurement system consisted of the MIRA Ultra analyzer, a sample air inlet and connecting tubings, a GPS (Global Positioning System) receiver (16X-HVS, Garmin, USA), an anemometer (Portable Mini, Calypso Instruments, USA), a data logger (CR-1000X, Campbell Scientific, Inc., USA), and power supply. The car is equipped with the anemometer on the roof, the GPS receiver at the inner side of the front window, and the air inlet at front low part of the car (∼ 0.5 m height from the ground). Sample air is drawn by the internal pump of the MIRA Ultra instrument. Length of the tubing from the air inlet to MIRA Ultra is 6.25 m (5 m 1/8^′′ OD and 1.25 m 1/4^′′ OD). Power is supplied from the car power socket (+12 VDC) via a power inverter to MIRA Ultra and the data acquisition system. The main component of the data acquisition system is the data logger, which collects data from the anemometer, GPS and MIRA Ultra. Prior to measurement, a gas mixture that contains 1 % of hydrocarbons including CH₄ and C₂H₆ was sprayed to the air inlet to measure response time from the air inlet to MIRA Ultra. The response time was 8 s in most days with only a few days of 7 or 9 s. The data were located with the response time corrected for each measurement day.

In this study, we identified CH₄ enhancements in a similar manner to previous studies (von Fischer et al., 2017; Weller et al., 2019; Maazallahi et al., 2020; Defratyka et al., 2021; Wietzel and Schmidt 2023; Fernandez et al., 2022; Vogel et al., 2024; Ueyama et al., 2025). The baseline CH₄ mole fraction was defined as the 5th percentile of all the data obtained within ±2.5 min moving time window for individual data points. Subtracting the baseline from each measurement value allowed us to determine the CH₄ excess value. When the CH₄ excess exceeds 0.1 ppm, the data point is tagged as a Leak Indication (LI), as visualized in Fig. 4a. We then identified a Leak Point (LP) when 5 or more consecutive data points were tagged as LIs, which means that CH₄ excess values of > 0.1 ppm lasted for > 5 s at LPs, as our measurement was at 1 Hz. The central locations (latitude and longitude) and the maximum excess CH₄ value among these consecutive data points were assigned to represent each LP. The C₂H₆ mole fraction data were processed in the same manner to determine the C₂H₆ excess. To avoid possible influence from the vehicle exhaust, the data points with vehicle speed less than 1.5 km h⁻¹ was excluded.

It is noted that our threshold (0.1 ppm) is same as that in our companion paper (Ueyama et al., 2025), similar to that applied for measurements in German cities (5 % by Wietzel and Schmidt, 2023) and smaller than that for measurements in Paris, France (0.5 ppm by Defratyka et al., 2021) and Bucharest, Romania (0.2 ppm by Fernandez et al., 2022). In this regard, we found that the magnitude of CH₄ excess was generally smaller in the Tokyo area than in US and Europe cities, and the threshold (0.1 ppm) was appropriate for effective detection of LI as a CH₄ emitting location. For instance, a higher threshold of 0.2 ppm would miss ∼ 60 % of LPs with smaller enhancements. In contrast, a lower threshold of 0.05 ppm would more than double the LP counts, but, due to comparable magnitudes of the observed baseline CH₄ variability, the LP detection would entail larger uncertainty. As shown in Fig. 5, the LPs with excess CH₄ < 0.2 ppm comprise 41 % of all the LPs found during the measurement campaign. About 93 % of the LPs fell in the range of excess CH₄ values of < 1 ppm. On the other hand, we detected 28 LPs (5 %) exceeding excess CH₄ of 2 ppm with maximum of 19.8 ppm, of which 23 and 5 LPs were classified as biogenic and fossil fuel sources (see Sect. 5).

As shown in Fig. 4a and b, we observed coincident increases of CH₄ and C₂H₆ mole fractions, and their enhancement ratio varied at different LPs. We determined C₂H₆-to-CH₄ enhancement ratios (C2/C1 ratio) for individual LPs. The C2/C1 ratio was used for source attribution of each LP according to the source classification by Fernandez et al. (2022); LPs with the C2/C1 ratio of < 0.005, 0.005–0.1 and > 0.1 are attributed to biogenic, fossil fuel and combustion sources, respectively. Note that the gas composition reported by the local gas company corresponds to the C2/C1 ratio of 0.063. For the measurement day shown in Fig. 4, we have clusters of the data categorized into biogenic (C2/C1< 0.005), fossil fuel (C2/C1 ∼ 0.05) and combustion (C2/C1 ∼ 0.15).

Table 1 summarizes relevant statistics and the observed appearances of LPs for 5 selected wards. We note that the road distance covered by the present measurement campaign was only ∼ 12 % of the total roads of Tokyo Metropolis. The coverage for each ward ranged from 2.6 % to 54.5 % (see Fig. 7). As shown in Table 1, the best coverage was in Shibuya Ward, followed by Toshima, Koto, Chiyoda and Minato Wards (> 25 %). Among 469 LPs found in Tokyo Metropolis, count of LPs ranged from 17 to 66 in these 5 wards, which correspond to LP densities from 0.16 to 0.47 km⁻¹. The LP density is count of LPs per travel distance and indicate average frequency of CH₄ enhancement encounters in a target city (Vogel et al., 2024; Ueyama et al., 2025). When grouped into source categories, biogenic and fossil fuel LP densities ranged from 0.07 to 0.35 km⁻¹ and from 0.08 to 0.20 km⁻¹, respectively. The average LP density in Tokyo Metropolis (0.33 km⁻¹) is comparable to that observed in Osaka (0.39 km⁻¹), the second largest city of Japan (Ueyama et al., 2025). When classified by source categories, the average LP densities of biogenic and fossil fuel origins in Tokyo were 0.18 and 0.13 km⁻¹, respectively, in comparison to 0.13 and 0.24 km⁻¹ in Osaka. Vogel et al. (2024) presented comparisons of leak indication densities across cities in Europe and North America. Although not exactly compared due to differences in data processing, the fossil fuel LP density in Tokyo is apparently as large as those of most European cities with relatively small numbers of leak indications (roughly from Barcelona to Groningen in Fig. 2 of Vogel et al., 2024). Note that we counted CH₄ enhancements of > 0.1 ppm, which falls between thresholds used in the two classification methods of Vogel et al. (2024).

We analysed the observed frequency of the C2/C1 ratio in each ward. In many wards, as represented by Minato and Koto Wards, C2/C1 ratios close to zero (i.e., biogenic sources) were observed most frequently (Fig. 8). In contrast, Shibuya Ward showed distinct characteristics of the C2/C1 ratio with the highest frequency at ∼ 0.05, which is likely to correspond to fossil fuel sources. As a result, about three quarters of the LP counts in Shibuya Ward was attributed to the fossil fuel source (Table 1). Toshima and Chiyoda Wards also showed that about half of LPs was attributed to fossil fuel, although number of LPs were limited. The appearance of combustion LPs (C2/C1 ratio of > 0.1) was minor for all areas.

The different frequencies of the source attributions were plausibly associated with land uses and known CH₄ sources of the different wards. According to the report of Tokyo Metropolitan Government (Bureau of Urban Development 2023), Shibuya and Toshima Wards are mainly covered by residential land (68.7 % and 68.3 %, respectively), while the fraction is relatively low in Koto Ward (47.5 %). In addition, we note that our measurements prioritized coverage of residential areas for Shibuya Ward. The residential land of Koto Ward is occupied substantially by industrial areas (26.6 %), whereas the predominant coverage in other 4 wards are public, commercial, and residential spaces. It is also important to note that the present study surveyed streets near landfill sites (Tokyo Bay-side Landfill) and two wastewater treatment plants resided in Koto Ward, contributing to large number of biogenic LPs in the area. Minato Ward also has a wastewater treatment plant surveyed. In Koto Ward, 2 and 31 LPs were found in the vicinities of landfill site and wastewater treatment plants, respectively, among 44 biogenic LPs in total (Table 1). In Minato Ward, 12 of 22 LPs were identified near the wastewater treatment plant. In contrast, our measurement did not survey any known biogenic CH₄ source facilities in Shibuya and Toshima Wards. It should be noted that the above LP counts (and its density) include those found at proximities of biogenic facilities as mentioned above. Other biogenic LPs could be related to sewer networks and other water environment distributed in the survey areas. These facility and non-facility sources should be separately evaluated in a more sophisticated approach to estimating areal CH₄ emissions, but due to the limited data currently available, we do not distinguish these sources in the analysis below.

To estimate areal CH₄ emissions, we employ empirical equations as referred in section 3 and previous studies. Also relevant is LP density (km⁻¹) defined as number of the observed LPs divided by the road distance covered in each ward (Table 1). Below we estimate CH₄ emissions from the 5 selected wards (Chiyoda, Minato, Koto, Shibuya and Toshima) only, because estimates for other wards are considered to be far from representative due to limited data. Using the average maximum CH₄ excesses in the area (Table 2), we calculated emission rates (L min⁻¹ or g CH₄ min⁻¹) according to Eq. (4) and equations suggested by previous studies. In this study, we present estimates using the equation by Weller et al. (2019) for compatibility with our companion paper (Ueyama et al., 2025) and previous studies (Maazallahi et al., 2020; Vogel et al., 2024). It is noted that, when Equation 4 (our control release experiment) was employed, the estimated emissions would be about 10 times larger than those using the Weller et al. (2019) equation, indicating that conversion from CH₄ excess to emission rate with an empirical equation is the considerable source of uncertainty. As described in Sect. 3, suitable choice of such an empirical conversion equation requires consideration of important factors such as emission height and distance from the source in the target area. Use of Weller et al. (2019) equation assumes that most CH₄ emissions are represented by ground emissions occurring near the vehicle (roughly within 10 m). In contrast, Eq. (4) (our control experiment) assumes emissions some tens m away at 5 m height. The former might approximate emissions underground sources (e.g., natural gas distribution pipes or urban sewer networks), whereas the latter downwind advection from large facilities or local restaurants at significant distance (Ueyama et al., 2025). Since the survey areas have these mixtures, it is difficult to determine appropriateness of the equations for individual cases.

The areal flux was estimated by the following equation (Ueyama et al., 2025): 5F=ER×N×RD/A, where F is the areal flux (nmol m⁻² s⁻¹), ER is the emission rate, N is the number of LPs per travel distance (km⁻¹), RD is the total road distance (km), and A is the total area of the ward (km²). The estimated areal fluxes for the selected wards are tabulated in Table 2 and shown in Fig. 9a. As seen in Fig. 9a, reflecting significant counts of the measurement data downwind of the landfill site and wastewater treatment plants, the estimated fluxes for Minato and Koto Wards were contributed predominantly by biogenic sources. In contrast, emissions from residential areas in Shibuya and Toshima Wards were almost entirely attributed to fossil fuel sources.

The areal fluxes can be converted to annual areal emissions (kt CH₄ yr⁻¹) by multiplying the total areas of the individual wards as shown in Table 2 and Fig. 9b. Based on our measurements, it was indicated that Koto Ward had the largest CH₄ emission of 0.22 kt CH₄ yr⁻¹, followed by Shibuya (0.096 kt CH₄ yr⁻¹), and Minato (0.068 kt CH₄ yr⁻¹) Wards. As discussed earlier, our estimates suggested that the areal emissions were predominantly of biogenic origin for Chiyoda, Minato and Koto Wards, whereas fossil fuel sources dominated the total emissions of Shibuya and Toshima Wards. In comparison to our estimates, CH₄ emissions from these wards were reported to be 0.04, 0.08, 0.12, 0.04 and 0.04 kt CH₄ yr⁻¹ for respective wards (Table 2). Although these emission estimates are not considered to be conclusive due to large uncertainties, it is noteworthy that our biogenic emission estimates for the five selected wards were correlated with the reported emissions. This is apparently consistent with the reporting of the Tokyo Metropolitan Government that biogenic emissions account for most of the reported emissions according to the standard emission methodologies with nominal emission factors (Bureau of Environment of Tokyo Metropolitan Government, 2024). However, for Shibuya and Toshima Wards, where our emission estimates attributed primarily to fossil fuel sources, the discrepancies between the estimated and reported emissions were large. This implies that either the fossil fuel emissions observed in this study were not taken into account or emission factors for the relevant sectors employed for the report were smaller than actual.

Lastly, we stress that our emission estimates presented in this study are preliminary and that accurate evaluation of the emission reporting by the local government is currently difficult. As described earlier, our survey covered only ∼ 10 % of roads of the Tokyo Metropolis, and up to ∼ 50 % at a maximum for the wards with relatively high measurement coverages. In addition, the survey prioritized measurements in the proximities of known CH₄ sources (mainly biogenic), by which the result may be biased towards biogenic emissions for the target wards with such known sources (Minato and Koto Wards). For residential areas where fossil fuel-related CH₄ enhancements were frequently found (Shibuya and Toshima Wards), the measurements were not even in space, by which the current estimates may be underrepresentative for respective areas. We plan the second phase of mobile measurement campaign in Tokyo to increase the road coverage. It should be also noted that upscaled calculation from vehicle-based mobile measurements at ground level could underestimate citywide emissions, since such observations could miss CH₄ emission signals taking place at heights above street level (Ueyama et al., 2025). Moreover, as discussed earlier, the empirical conversion from the CH₄ excess to the emission rate (Sect. 3) is a source of large uncertainty, roughly a factor of 10. The equation will be better evaluated through comparison to the ongoing eddy covariance measurements at the site Yoyogi located in a residential area of Shibuya Ward (Sugawara et al., 2021), as performed by Ueyama et al. (2025). For preliminary comparison to Ueyama et al. (2025), CH₄ flux estimations for selected wards of Tokyo would fall within the range of variability found in various areas in Osaka and Sakai cities if the same conversion equation from excess CH₄ to emission rate was applied, which implies comparable magnitudes of urban CH₄ emissions in both megacity areas.