1 Introduction
Hainan Province, China’s largest special economic zone, pilot free trade zone, and national ecological civilization pilot zone, has established the most rigorous ecological protection protocols in China through the enforcement of the Hainan Free Trade Port Law. The
Implementation Plan for the National Ecological Civilization Pilot Zone set forth the objective for Hainan Province to achieve world-leading standards in ecological environment quality and resource utilization efficiency by 2035. Despite Hainan’s superior air quality relative to China’s mainland, with 2023 annual mean concentrations of SO
2 (3.2 μg/m
3), NO
2 (8.1 μg/m
3), and CO (0.6 mg/m
3) consistently meeting WHO guidelines (WHO, 2021), gaps still existed with the global benchmark. The province’s annual PM
2.5 concentration (12.0 μg/m
3) in 2023 (The People’s Government of Hainan Province, 2024) exceeded both the WHO guideline (5 μg/m
3) and the proposed world-class standard of 10 μg/m
3 established by Song et al. (
2024) for ecological civilization pilot zones. Moreover, emerging evidence suggested that health risks persist even at low PM
2.5 concentrations (below 10 μg/m
3), with some studies indicating potentially greater per-unit health impacts at these lower exposure levels (
Strak et al., 2021;
Weichenthal et al., 2022;
Zhao et al., 2024;
Zhu et al., 2024) thereby highlighting the critical need for sustained pollution mitigation efforts even in relatively clean regions. Consequently, Hainan Province required advanced control strategies to further reduce PM
2.5 concentrations.
Under low PM
2.5 concentration conditions, background values constituted a critical determinant of air quality. The background concentration of PM
2.5 typically referred to the concentration level unaffected by local emission sources, constituted by natural emissions, such as biogenic volatile organic compounds (BVOCs), biomass burning, sea salt emissions, soil NO
x emissions, soil dust, lightning activity, and contributed by regional transport (
Ortiz and Friedrich, 2013;
Han et al., 2015;
Martín-Cruz et al., 2020). Estimating Hainan’s PM
2.5 background concentration provided a crucial reference for assessing the province’s potential for air quality improvement, and constituted a critical foundation for designing effective emission reduction strategies to achieve world-leading air quality goals by 2035. Hainan, situated in the subtropical coastal zone, experienced relatively high natural source emissions including BVOCs and sea salt emissions (
Li et al., 2017). Therefore, the background PM
2.5 concentration in Hainan was significantly influenced by natural sources. Additionally, studies indicated that regional transport contributed substantially to Hainan’s PM
2.5 background concentration (
Zhang et al., 2014;
Shi et al., 2017;
Cao et al., 2023;
Xu et al., 2024). Recent assessments employing observation-based source apportionment methods revealed that natural sources accounted for 26% of PM
2.5 concentration (The People’s Government of Hainan Province, 2021). However, conventional observation-based source apportionment methods such as Positive Matrix Factorization (PMF) and Chemical Mass Balance (CMB) face substantial limitations in distinguishing between local and regional transport contributions, and in identifying the contribution of BVOCs (
Wang et al., 2018). Although chemical transport models (e.g., CMAQ, WRF-Chem) have demonstrated superior advantages in resolving source-receptor relationships through sensitivity simulations (
Tchepel et al., 2010;
Guo et al., 2017;
Coelho et al., 2023), no study has yet utilized these models to systematically quantify the contributions of natural emissions and regional transport to PM
2.5 concentrations in Hainan, hindering the determination of its PM
2.5 background concen-tration. Furthermore, given that the concept of background concentration can serve different purposes in scientific and policy contexts, adopting distinct definitions of background value for Hainan was necessary and meaningful.
To address these research gaps, this study integrated multiple emission inventories, conducted sensitivity scenario simulations using the CMAQ model, and constrained the modeling results with various observational data (including regular state-controlled stations, background stations, and BVOC measure-ments), thereby enhancing the reliability of source apportionment. Notably, we propose three distinct definitions of PM2.5 background values, each serving different policy-assessment purposes. Furthermore, a future emission scenario was incorporated to evaluate the feasibility of achieving world-leading air quality objectives, providing a forward-looking perspective for policy-making. The research aimed to: (1) assess impacts of natural emissions and regional transport on PM2.5 concentrations across Hainan Province; (2) deter-mine the PM2.5 background concentration in Hainan and assess the impact of meteorological condition fluctuations on background PM2.5 level; (3) compare Hainan’s PM2.5 background values with those from other clean regions worldwide. The research findings would provide critical scientific support for Hainan’s transition toward world-class air quality standards while maintaining its ecological leadership.
2 Materials and methods
2.1 Model configurations
The CMAQv5.3.2 model was employed to investigate contributions of anthropogenic and natural emission sources to PM2.5 concentrations in Hainan Province through a nested-domain modeling approach. The modeling framework consisted of two nested domains. The outer domain covered continental China and adjacent Southeast Asian regions with 27 km × 27 km horizontal grid resolution, while the inner domain focused on southern provinces of China, and neighboring Southeast Asian countries including Vietnam, Laos, Myanmar, and Thailand, extending over portions of the South China Sea with higher 9 km × 9 km spatial resolution. Meteorological inputs for the air quality simulations were generated using the Weather Research and Forecasting model version 3.9.1 (WRFv3.9.1). Initial and boundary conditions for the WRFv3.9.1 model were obtained from 1° × 1° Final Global Tropospheric Analysis data (FNL) global analysis dataset provided by the National Centers for Environmental Prediction (NCEP), updated at 6-hour intervals. The four-dimensional data assimilation (FDDA) was applied using the NCEP Automated Data Processing (ADP) Operational Global Surface Observational Data (ds461.0). The WRF configuration incorporated multiple physical parameterizations including the Pleim-Xiu land surface scheme, Asymmetric Convective Model version 2 (ACM2) boundary layer scheme, Grell-Freitas (GF) cumulus parameterization scheme, Rapid Radiative Transfer Model for Global climate models (RRTMG) radiation scheme, and Morrison double-moment cloud micro-physics. The outer domain received boundary conditions from global-scale GEOS-Chem simulations, while the inner domain utilized concentration outputs from the parent domain simulations. The CMAQ implementation adopted the SAPRC07 gas-phase chemical mechanism coupled with the AERO07+2D_VBS aerosol module to simulate gas-phase chemistry and aerosol processes. To reduce potential influences from initial conditions, all simulation periods incor-porated a 5-d spin-up phase prior to formal analysis periods.
2.2 Emission inventories
The anthropogenic emissions for Hainan emission inventory integrated local air pollutants emissions in Hainan Province and provincial-level air pollutants emissions from the Air Benefit and Cost Attainment Assessment System - Emission Inventory version 2.0 (ABaCAS-EI v2.0) dataset developed by Tsinghua University (
Li et al., 2023). The anthropogenic emissions of SO
2, NO
x, and particulate matter (PM) for Hainan were mostly acquired from the Hainan Research Academy of Environmental Sciences. Specifically, point sources encompassed power plant and industrial process emissions, while non-point sources included on-road vehicle, dust, and commercial cooking emissions. Pollutants emissions from other sectors, VOCs emissions across all sectors in Hainan, and the anthropogenic emissions for China’s mainland other than Hainan were adopted from the ABaCAS-EI database (
Li et al., 2023). The sector-specific air pollutants emissions in Hainan Province for 2019 were presented in Fig. S1. The foreign anthropogenic emissions were characterized using the Emissions Database for Global Atmospheric Research (EDGAR), jointly released by the Netherlands Environmental Assessment Agency and the European Commission’s Joint Research Centre (
Crippa et al., 2020). The shipping emissions were obtained from the multi-scale Shipping Emission Inventory Model (SEIM) with a spatial resolution of 0.1 degrees, covering SO
2, NO
x, CO, VOCs, PM
2.5, OC, and BC (
Liu et al., 2016).
Biomass burning emissions were quantified using the Fire Inventory from National Center for Atmospheric Research (NCAR) version 2.5 (FINNv2.5), which calculated gas and particulate emissions from forests and agriculture fires using satellite observations of active fires, land cover, emission factors, and estimated fuel loads (
Wiedinmyer et al., 2023). With a high spatiotemporal resolution of 1 km spatially and daily temporally, FINNv2.5 effectively captured the spatio-temporal variations in biomass burning (
Wiedinmyer et al., 2023). Biogenic emissions were simulated using MEGANv3.2 (
Crippa et al., 2020), and sea salt aerosol emissions were estimated via the CMAQ sea salt module following wind-speed dependent emission schemes (
Ovadnevaite et al., 2014). The soil NO
x emissions were calculated online as a function of temperature using MEGANv3.2 (
Guenther et al., 2006). The lightning-induced NO
x emissions were estimated using observed lightning flashes along with the assumed NO
x production per flash (
Kang et al., 2019).
2.3 Observation data sources
Multiple sources of observational data were used in this study to achieve comprehensive model validation. Hourly PM
2.5 concentration measurements from 2019 were collected from national control stations in Haikou and Sanya. To enhance the spatiotemporal resolution of validation, daily mean concentrations of inorganic components (e.g., SO
42−, NO
3−, CL
−, NH
4+, K
+, Ca
2+, Mg
2+, Na
+) for Haikou and Wuzhishan in 2021 were acquired from Hainan University. For regional background validation, this study integrated two complementary observational datasets provided by the Hainan Provincial Environmental Monitoring Center (HPEMC). The first dataset comprised daily average PM
2.5 concentrations obtaining from the Wuzhishan atmospheric background station (18.81 °N, 109.54 °E), a pristine monitoring site. The second dataset incorporated seasonal PM
2.5 averages from the Wenchang Puqian Regional Atmospheric Super Station (19.62 °N, 110.89 °E), which had been scientifically established as a representative regional background monitoring facility for Hainan Province. To validate biogenic emission contributions, hourly isoprene concentrations monitored in Sanya (18.25 °N, 109.51°E) during 2019 acquiring from the HPEMC were collected. In addition, radiocarbon (
14C) measurements from a coastal background station and an urban station in Shenzhen (
Huang et al., 2020;
He et al., 2022) were incorporated to quantitatively constrain biogenic contributions.
2.4 Quantification of emission source contributions and background values
The source contributions to PM
2.5 concentrations in Hainan were quantified using the brute-force method. Based on previous findings demonstrating significant impacts of biomass burning from Southeast Asia on Hainan’s air quality (
Li et al., 2017;
Duc et al., 2021;
Qin et al., 2024), we specifically investigated both local and regional biomass burning contributions to PM
2.5 concentrations in Hainan. Based on preliminary assessments (
Murray, 2016;
Huang et al., 2023), soil NO
x emissions, lightning NO
x emissions and soil dust were found to contribute minimally to PM
2.5 in Hainan. Consequently, these sources were included in the emission inputs but were excluded from contribution attribution analysis. The natural sources investigated here comprise biomass burning with wildfire predominance, VOCs emitted from vegetation and sea salt aerosols originating from marine emissions. A series of sensitivity experiments comprising one baseline scenario (Base) and six sensitivity scenarios (Table 1) was established to isolate specific source contributions. The Base scenario incorporated all emission sources, while each sensitivity scenario systematically excluded specific source categories. Differences between the baseline scenario and sensitivity scenarios of No_Local_ANTH, No_Trans_ANTH, No_Local_BB, No_Trans_BB, No_BVOCs, and No_Sea salt respectively represented the contributions of local anthropogenic emissions, regional anthropogenic emissions, local biomass burning, regional biomass burning, BVOCs emissions, and sea salt emissions to PM
2.5 in Hainan. The “local emissions” denoted emissions originating within Hainan Province, while “regional emissions” referred to emissions from China’s mainland outside Hainan and Southeast Asia within the model domain.
To evaluate seasonal variations in PM2.5 concen-trations across different emission scenarios, distinct temporal sampling strategies were imple-mented. For the No_Local_ANTH, No_Trans_ANTH, No_BVOCs, and No_Sea salt scenarios, simulations from January, April, July, and October 2019 were selected as seasonal representatives for winter, spring, summer, and autumn, respectively. However, given the substantial monthly variability characteristic of biomass burning emissions relative to other emission sources, we adopted a more comprehensive approach for the Base, No_Local_BB, and No_Trans_BB scenarios to minimize seasonal representation bias. These biomass burning-related scenarios employed full-year 2019 simulations with seasonal classifications as follows: winter (January, February, December), spring (March, April, May), summer (June, July, August), and autumn (September, October, November). This extended sampling framework ensures more robust charac-terization of biomass burning contributions to Hainan’s PM2.5 concentrations while maintaining temporal compar-ability across all investigated scenarios.
In the current study, we adopted three definitions to quantify PM
2.5 background concentrations in Hainan Province, each addressing a specific scientific or policy-management objective. The first definition characterized Hainan’s background concentration as PM
2.5 levels after removing local anthropogenic emissions (labeled as background I). This was a conventional definitions widely adopted in existing studies. The second definition (background II) derived PM
2.5 background values by excluding anthropogenic emissions from both Hainan and its surrounding regions (all Chinese provinces and Southeast Asian countries within the second model domain that were adjacent to Hainan). This represented a theoretical background level under atmospheric conditions without anthro-pogenic interference, primarily driven by natural emissions. The third definition quantified Hainan’s PM
2.5 background value by excluding local anthro-pogenic emissions within Hainan while applying the 2035 anthropogenic emissions under the “enhanced control policies” (ECP) scenario to Hainan’s surrounding regions (background III). This ECP scenario assumed smooth implementation of emission reduction measures in China’s mainland. It was designed to implement the National Air Quality Continuous Improvement Action Plan, Carbon Peak Implementation Plan, and Beautiful China Initiative. Specifically, the ECP scenario assumed continued future Gross Domestic Product (GDP) growth, albeit at a gradually slowing pace as the economy matures. The energy mix would be significantly optimized, with the share of clean energy installed capacity in the power sector projected to reach 94% by 2035. For the transportation sector, the transition to new energy vehicles would be accelerated, assuming that the share of new energy vehicles in new car sales would gradually approach 100% by 2035. For industrial and other sectors, advanced end-of-pipe control techn-ologies and cleaner production processes would be adopted in the future. This emission scenario forecasted a 35%−65% reduction in anthropogenic emissions in Hainan Province by 2035 compared to the 2019 baseline. Nevertheless, this scenario was subject to uncertainties, primarily including the stringency and timeliness of future policy implementation, the pace of technological advancement and penetration (e.g., in clean energy storage efficiency and pollution control technologies), deviations of actual economic growth from projections. Detailed information regarding the ECP emission scenario can be found in
Song et al. (2024). This prospective definition of background III provided a realistic projection of the background concentration expected by 2035, contingent upon the successful enactment of regional emission reduction policies across mainland China. It directly supported the policy imperative of evaluating the attainability of Hainan’s goal to achieve world-leading air quality objectives by 2035. Since publicly reported PM
2.5 concentrations in Hainan Province were derived from the average of 18 cities monitoring stations, we calculated the background PM
2.5 concentrations based on model grid averages corresponding to these stations. This station-based approach aligned with China’s environmental management protocols, where air quality assessments rely on monitoring station data. By providing background concentrations directly com-parable to official reports, our results delivered actionable scientific guidance for Hainan’s policy-makers. In addition, we also provided PM
2.5 background values based on the average values across all grids in Hainan Province for comparing with other background value studies. When calculating the PM
2.5 background values for the 18 cities in Hainan Province, we corrected the simulation results using observational data, following the method of
Song et al. (2024). The adjustment was implemented by multiplying the simulated PM
2.5 concentrations under the scenarios of No_Local_ANTH, No_ANTH, and No_Local_ANTH_2035 by the ratio of BASE scenario simulations to observed concentrations for each city.
2.5 Selection of the most and least favorable meteorological years
Since meteorological conditions could influence background values, we also investigated their impact on PM2.5 background values in Hainan. For each month from 2014 to 2023, we identified the four months with the highest and lowest observed PM2.5 concentrations in Haikou and Sanya, respectively defined as heavy pollution months and clean months. The average values and anomalies of relative humidity (RH) and northeasterly wind during the heavy pollution months and clean months were calculated for each year from 2014 to 2023. Analysis of annual anomalies in RH and northeasterly wind frequency during heavy pollution months and clean months revealed consistent patterns (Fig. S2). During heavy pollution months, RH exhibited near-universal negative anomalies while northeasterly winds showed predominantly positive anomalies across the years, indicating that low RH and northeasterly winds contribute to pollutant formation or accumu-lation. For clean months, the relationship with RH anomalies was weaker, but northeasterly winds consistently displayed negative anomalies, demon-strating that southwesterly winds facilitate pollutant dispersion. Based primarily on the annual northeasterly wind anomalies, 2020 was identified as the most meteorologically favorable year for Haikou and Sanya. Conversely, 2019 was determined to be the most meteorologically unfavorable year, as the combined influence of low RH and persistent northeasterly winds directly contributed to the peak PM2.5 concentrations observed during its heavy pollution months.
3 Results and discussion
3.1 Model performance
Figure 1(a) presents comparisons of the daily average PM2.5 concentrations derived from model simulations and observations at Haikou and Sanya sites. The model demonstrated satisfactory performances for PM2.5 across all sites, with normalized mean bias (NMB) values not exceeding ±15.0%. The model accurately captured most PM2.5 concentration peaks at Haikou and Sanya sites. The simulated temporal variation of PM2.5 concentrations at these sites were in close agreement with the observed data, achieving correlation coefficient (R) values not less than 0.7. We further evaluated the model’s performance in simulating inorganic ionic components. Figure 1(b) demonstrates that the model successfully reproduced the concen-tration levels and seasonal characteristics of inorganic ionic species at both the Haikou and Wuzhishan sites, yielding NMBs of 3.0% and −25.8%, respectively. The model simulated concentrations of inorganic salts (ammonium, nitrate, and sulfate) within reasonable ranges at both locations, with NMBs of 8.4% and 24.2% and RMSEs of 1.1 μg/m3 and 0.79 μg/m3 for Haikou and Wuzhishan, respectively. The model captured the observed SO42− concentrations well, overestimated NO3− at both Haikou and Wuzhishan, and underestimated NH4+. For K+, Ca2+, and Mg2+ ions, the simulations generally fell within acceptable bias ranges, showing NMBs of 37.2% and −38.9% and corresponding RMSEs of 0.19 and 0.18 μg/m3 at Haikou and Wuzhishan, respectively. As the dominant component of sea salt, the accurate simulation of NaCl concentrations served as a critical prerequisite for reliably assessing the model’s capability to quantify sea salt contributions to PM2.5. As shown in Fig. 1(b), the model successfully reproduced both concentration levels and Seasonal patterns of NaCl, with overall NMBs and RMSEs of −11.2% and 0.25 μg/m3 for Haikou and −6.6% and 0.11 μg/m3 for Wuzhishan.
Isoprene, recognized as the predominant BVOCs in terms of global emissions, constituted approximately 50% of total BVOCs emissions (
Guenther et al., 2006). A comparison between predicted and observed isoprene concentrations was widely adopted as an effective approach for evaluating the accuracy of biogenic emission inventories (
Zhang et al., 2020). Here, we compared the simulated and observed daily mean isoprene concentrations in Sanya, with the results presented in Fig. 1(c). The model reasonably reproduced the isoprene concentrations and their temporal variations in Sanya, exhibiting a NMB of −14.8% and a strong correlation with observations (
R = 0.7). This observation constraint enhanced our confidence on the biogenic emission inventory for Hainan used in this study and the model’s simulation of BVOCs. Accurate simulation of contributions from natural sources was critical for determining the background PM
2.5 concentrations in Hainan. Thus, we further constrained the model simulations of BVOCs contributions using observational analytical results of
14C. The
14C content in aerosols served as an effective tracer for distinguishing carbon contributions from non-fossil fuel sources (including BVOCs and biomass burning emissions) and fossil fuel sources. This provided observational evidence to quantify the contributions of BVOCs and biomass burning to organic aerosol (OA). Given that
14C isotopic measurements primarily occurred during non-biomass-burning periods (
Huang et al., 2020;
He et al., 2022), the identified non-fossil carbon predominantly originated from BVOCs emissions. Comparative analyses were performed at two monitoring sites, the Shenzhen Coastal Background Station representing suburban environments and the Shenzhen Research Academy Site characterizing urban areas. As illustrated in Fig. 1(d), the model simulations revealed BVOCs contributions of 22.5% to organic carbon (OC) at the coastal background station, which was close to the 25.4% estimated through
14C isotopic observations. At the urban site, simulated BVOCs contributions reached 20.3%, showing good agreement with observational estimates of 18.5%. These comparative analyses demonstrated the reliability of the applied methodology in resolving biogenic and anthropogenic influences on aerosol. The simulated isoprene concentrations and BVOCs contributions to OC demonstrated strong consistency with observational data, supporting the reliability of our subsequent quantification of BVOCs contributions to PM
2.5 in Hainan.
To validate the model’s simulation of background PM2.5 at the Hainan atmospheric monitoring station, we compared simulated PM2.5 concentrations with observational data from Wuzhishan background site and the Wenchang Background Station. The model reproduced the characteristics of PM2.5 concentration at Wuzhishan site, with a NMB of 8.0%. At the Wenchang site, the model also demonstrated close agreement with the monitoring site data in both annual averages (with a NMB value of 4.2%) and seasonal variations in Fig. 1(e). This confirmed its reliability in characterizing background concentrations in Hainan.
3.2 Source contributions to PM2.5 concentration
Figure 2 illustrates the spatial distribution of annual contributions from six emission sources to annual PM2.5 concentrations, while Figure 3 quantifies their seasonal contributions to PM2.5 concentrations in Hainan. Local anthropogenic emissions constituted the largest source of PM2.5 in Hainan, particularly in northern regions where they contributed over 60% to PM2.5 concen-trations. The annual average contribution of local anthropogenic emissions to Hainan’s PM2.5 concentration was 6.75 μg/m3 (49.5%). Local anthro-pogenic emissions exhibited pronounced seasonal variability in their contributions to PM2.5 concentrations across Hainan Province. Under prevailing southerly winds in summer, regional transport of mainland pollutants subsided substantially, allowing local emissions to dominate PM2.5 concentrations in Hainan at 63.9%. While during summer and winter, intensified regional transport decreased local anthropogenic contributions to 39.0%–46.7% of PM2.5 concentrations. As shown in Fig. S1, local biomass burning exhibited substantially lower annual emissions of NOx, SO2, VOCs, and primary PM2.5 (0.5%–4.3%) in Hainan compared to other anthropogenic sources (23.7%– 86.9%). Consequently, local biomass burning contri-buted limited to Hainan’s annual PM2.5 concentration, averaging 0.12 μg/m3 (0.9% of total). Seasonal analysis identified spring as the period of maximum influence of local biomass burning on Hainan, accounting for 2.1% of PM2.5 concentrations (Fig. 3).
Anthropogenic emissions from surrounding regions constituted the second-largest PM
2.5 source in Hainan, just second to local anthropogenic emissions, demon-strating an annual mean concentration of 2.51 μg/m
3 (18.4% of total). The contribution of anthropogenic emissions from surrounding regions exhibited pronounced seasonal variability, with contributions reaching 21.8% in autumn and further intensifying to 30.2% during winter. This seasonal amplification aligned with the northeasterly monsoon patterns, which transported pollutants from the Pearl River Delta region to Hainan during cold seasons. Conversely, the contri-butions of anthropogenic emissions from surrounding regions were significantly lower in warmer periods, accounting for merely 4.1% and 8.8% of PM
2.5 concentrations during spring and summer, respectively. Biomass burning constituted a major PM
2.5 source within Southeast Asian countries, particularly Laos, Myanmar, Thailand, and Vietnam, with the highest annual contribution reaching 58% in Laos. This contribution spatial pattern correlated with the intensity of fire activity mainly caused by forest fires (
Le et al., 2020;
Yin, 2020;
Duc et al., 2021). These high biomass burning emissions in Southeast Asia subsequently influenced Hainan’s air quality through regional transport (
Yin, 2020;
Phuong et al., 2021), contributing 1.48 μg/m
3 (10.9% of total PM
2.5) annually to PM
2.5 concentrations in Hainan. Seasonal analysis revealed spring as the peak impact period (31.9%), when enhanced biomass burning in the Southeast Asia regions coincided with westerly transport pathways to Hainan (
Zhang et al., 2012). For example,
Duc et al. (2021) utilized satellite remote sensing data and the WRF-Chem model, demonstrated that pollutants from biomass burning in Thailand and Laos during spring, 2019 were uplift to high altitude along the Truong Son mountain under westerly wind, and impacted downwind regions including Hainan through long-range transboundary transport. Combined contributions of biomass burning and anthropogenic emissions from Hainan’s surrounding regions accounted for 29.3% (3.99 μg/m
3) of the annual mean PM
2.5 concentrations, highlighting the critical role of regional transport in shaping air quality in Hainan (
Zhang et al., 2012,
2014;
Zhou et al., 2018). BVOCs emissions contributed 10.4% (1.41 μg/m
3) annually to PM
2.5 concentration in Hainan, with seasonal peaks reaching 16.5% during summer months due to intensified vegetation metabolic activity. Sea salt emissions predominantly affected coastal zones, contributing 8.6% (1.17 μg/m
3) annually to Hainan’s PM
2.5 concentration, consistent with the source apportionment results of 7% identified by
Shi et al. (2017).
3.3 PM2.5 background concentration in Hainan
Following the definitions for PM2.5 background concentration described in Section 2.4, we determined two types of PM2.5 background values for Hainan. One was calculated based on the average concentration across all grids within Hainan and another was derived from the average value of the 18 cities. As illustrated in Fig. 4(a), our analysis showed that the 18 cities-averaged PM2.5 background I, II, and III value were 6.9, 4.2, and 5.2 μg/m3, respectively. The corresponding values derived from the province grid-averaged were 6.7, 4.6, and 4.9 μg/m3.The annual PM2.5 background concentrations obtained using the 18 cities-averaged and the grid-averaged showed minor differences, despite evidently higher observed urban PM2.5 concentrations compared to suburban levels across Hainan at present. This occurred because after removing local anthropogenic emissions, the remaining sources exerted a similar influence on PM2.5 concentrations across urban, suburban, and rural areas of Hainan. Seasonal analysis demonstrated elevated PM2.5 background concentrations during spring, autumn, winter compared to summer. This seasonal variation was associated with the seasonal patterns of biogenic emissions and atmospheric dispersion conditions.
Additionally, we compared the current PM2.5 concentrations with the background values across 18 cities in Hainan (Fig. S3). After excluding local anthropogenic emissions within Hainan, the background PM2.5 concentrations in Hainan ranged from 2.6 to 9.4 μg/m3. After removing anthropogenic emissions from Hainan and the surrounding regions of Hainan, the background concentrations across the cities ranged between 1.9 and 7.3 μg/m3. Under the 2035 policy scenario for the surrounding regions combined with the removal of local Hainan anthropogenic emissions, the background values varied from 2.2 to 7.8 μg/m3. Compared to current levels, the background concentrations calculated using the three definitions decreased by 28.7%–71.8%. Notably, seven cities including Haikou, Sanya, Qionghai, Wanning, Tunchang, Chengmai, and Lingao, showed PM2.5 decreases exceeding 50%, demonstrating significant potential for anthropogenic emission reductions in these cities. Overall, Hainan’s relatively low PM2.5 background levels indicated its considerable remaining potential for anthropogenic emission reductions. Sustained efforts to further decrease PM2.5 concentrations could position Hainan to achieve world-leading air quality targets by 2035.
Considering the influence of meteorological conditions on PM2.5 background values, we analyzed the PM2.5 background values in Hainan under the most favorable and least favorable meteorological years between 2014 and 2023. According to Section 2.5, 2019 served as the baseline simulation year and represented the most meteorologically unfavorable conditions, while 2020 was identified as the most favorable year. It should be noted that the PM2.5 background values analyzed here refer to PM2.5 concentrations after excluding local anthropogenic emissions in Hainan. The simulation analysis indicated that compared to the 2019 baseline, the favorable meteorological conditions in 2020 resulted in a reduction of 1.6 and 0.8 μg/m3 in the annual mean PM2.5 background value for the grid-averaged and the 18 cities-averaged, respectively (Fig. 5). Seasonally, the favorable meteorological conditions in 2020 resulted in an increase in PM2.5 background values during spring, while decreases during summer, autumn, and winter. Overall, accounting for meteorological variability, the PM2.5 background levels in other years in Hainan were likely lower than those derived from 2019. This implied greater potential for Hainan to reduce emissions in the future.
3.4 Comparison of PM2.5 background concentrations with other clean areas
To understand the level of PM
2.5 background concentrations in Hainan, we compared Hainan’s PM
2.5 background concentration with previously reported PM
2.5 background values from other clean regions in China and other countries. The results are presented in Table 2. The PM
2.5 background values in Table 2 were derived from multiple estimation approaches, including observational sampling methods where background air quality monitoring sites were established in areas with minimal human activity such as mountains and islands, denoted in Table 2 as “Observation”. Compared to the PM
2.5 concentrations measured at atmospheric back-ground stations in China, Hainan’s PM
2.5 background concentration (4.2–6.9 μg/m
3) was significantly lower than values observed at major Chinese atmospheric background stations, including Shenlongjia, Hubei (9.2 μg/m
3;
Gao et al., 2020), Menyuan, Qinghai (12.7 μg/m
3;
Gao et al., 2020), Wuyi Mountain (17.3 μg/m
3;
Gao et al., 2020), Changbai Mountain (17.6 ± 12.6 μg/m
3;
Liu et al., 2018), and Tibetan sites (Namtso: 11.2 ± 6.9 μg/m
3; Zangdongnan: 12.3 ± 8.0 μg/m
3;
Liu et al., 2018). It was slightly below Lijiang, Yunnan (7.6 μg/m
3;
Gao et al., 2020), and comparable to the PM
2.5 background levels observed at remote background sites in Sichuan (Hailuogou: 6.7 μg/m
3;
Gao et al., 2020; Gongga Mountain: 6.5 ± 6.2 μg/m
3;
Cheng et al., 2019). The comparisons indicated that Hainan’s PM
2.5 background value was at a relatively low level compared to other atmospheric background stations in China. It also reflected that, despite most of China’s background monitoring sites were situated far from urban areas and lack notable nearby anthropogenic emissions, they were still influenced to some extent by emissions from surrounding regions. Globally, Hainan’s background PM
2.5 concentrations were comparable to the European background concentrations (Birkenes: 4.3 ± 1.1 μg/m
3; Hyytiälä: 5.5 ± 1.3 μg/m
3;
Cusack et al., 2012; regional mean: 4.8 ± 2.4 μg/m
3;
Putaud et al., 2004) but remained below North American sites (Colorado: 7.9 μg/m
3;
Stowell et al., 2019; Iberian Peninsula: 8.6 μg/m
3;
Cusack et al., 2012). The comparison between urban background sites (e.g., London, Milan, Paris, Düsseldorf) and the Hainan involved inherent differences in spatial representativeness and monitoring conditions. However, it should be noted that this comparison was not intended to equate the two types of sites directly, but rather to provide a broader context for interpreting the PM
2.5 background levels. Despite differences in site classification, this approach helped to underscore the relatively clean atmospheric background in Hainan compared to other clean regions.
In addition to PM
2.5 background value comparisons from atmospheric background monitoring sites, we also collected background PM
2.5 values for global pristine areas derived from various numerical calculation methods. For instance, Hainan’s PM
2.5 background concentrations (4.2–6.9 μg/m
3) were comparable to or slightly higher than values reported for major European cities (
Gao et al., 2020) including London (3.2 μg/m
3), Milan (5.0 μg/m
3), Paris (4.8 μg/m
3), and Dusseldorf (5.6 μg/m
3), which were estimated using the three-dimensional PMCAMx transport model coupled with Particulate Matter Source Apportionment Technology (PSAT). The PM
2.5 background concentrations I (6.7 and 6.9 μg/m
3) estimated for Hainan was comparable to background levels in Italy (6.2 μg/m
3;
Cesari et al., 2016) determined through PMF source apportionment. The values were higher than the value reported in Central Israel (4.3 ± 2.5 μg/m
3;
Belachsen and Broday, 2024), where background values were determined through factor identification via Non-negative Matrix Factorization method and assessment of air mass source contributions to PM
2.5. Similarly, the Hainan estimates were higher than the background value obtained for the Canary Archipelago (4.6 ± 0.4 μg/m
3;
Martín-Cruz et al., 2020), which was estimated through statistical clustering of the lowest concentration cluster in PM
2.5 time-series observations. Hainan’s PM
2.5 background level was lower than meteorological-filtered values derived from direct measurements during periods meeting specific lower-layer wind field conditions with no pollution events for Aletai atmospheric background station in Xinjiang (7.8 μg/m
3;
Wang et al., 2019). Using a similar local emission zeroing approach through CMAQ model simulation, background concentration I (6.7 and 6.9 μg/m
3) identified for Hainan in this study was markedly lower than the reported background level from Taiwan, China (9.0 μg/m
3;
Chen et al., 2014). However, another study applying the Advanced Global Atmospheric Gas Experiment (AGAGE) method, which estimated background concentrations by selecting periods with low PM
2.5 values, reported a PM
2.5 background of 4.4 ± 2.6 μg/m
3 for Taiwan, China (
Wang et al., 2020). This was lower than Hainan’s PM
2.5 background I (6.7 and 6.9 μg/m
3) and background III (4.9 and 5.2 μg/m
3), whereas it aligned with background concentration II (4.6 and 4.2 μg/m
3). The substantial disparity between two background values from Taiwan was likely attributable to differences in analysis years and methodologies. This underscored the need for caution when comparing PM
2.5 background levels derived through different research approaches. Overall, despite methodological variations in PM
2.5 background concentration determination across different studies, comparative analyses with both regional background values using the same CMAQ model simulation approach and atmospheric background station measurements demonstrated Hainan’s PM
2.5 back-ground concentrations at comparable or lower levels.
3.5 Implications for future air quality management
When determining future emission reduction targets, it was necessary to consider background concentrations and their variability influenced by meteorological conditions. It should be noted that the three definitions of PM2.5 background values served different purposes. Background I is widely used and directly reflects the current influence of external sources. Background II represented a theoretical lower bound, useful for assessing natural baselines. Background III, which incorporated future emission reductions in surrounding regions, was particularly relevant for policy assessment, as it reflected achievable background levels under realistic regional cooperation scenarios. However, it relied heavily on the accuracy of future emission projections. The selection of an appropriate background definition should align with the spatial scale and policy objective. For city-level management, background I may be more practical as it reflected regional transport influences. For provincial or regional scale, background III offered a forward-looking reference aligned with regional emission policies. If the goal was to assess the maximum potential for air quality improvement, background II provided a theoretical reference under idealized conditions. Our research demonstrated that Hainan’s PM2.5 background concentrations were generally low using the three definitions. When meteorological variability was considered, these back-ground concentrations may decrease further. Therefore, there was significant potential for further improvement in Hainan’s air quality in the future. Continuous implementation of specific control measures targeting energy combustion, industrial processes, transportation sources, and agricultural sources was required to further reduce air pollutants emissions.
In addition, Hainan experienced substantial contri-butions from anthropogenic emissions transported from surrounding regions, particularly during autumn and winter when PM
2.5 concentrations were high (
Ping et al., 2023). Therefore, reducing the pollution from outside the region was considered crucial for achieving Hainan’s air quality goals. It was recommended that Hainan strengthen regional joint prevention and control efforts with other provinces in southeastern China. Specifically, in addressing cross-regional pollution transport, cooperation should be enhanced through mechanisms including information sharing, policy coordination, and joint law enforcement. For instance, under the broader vision of the Beautiful China initiative, Hainan could collaborate with neighboring provinces like Guangdong, Guangxi and Fujian to focus on strengthening PM
2.5 control measures during winter to mitigate the impact of external pollution on Hainan’s air quality. Furthermore, given Hainan’s geographical location and the evident impacts from Southeast Asian biomass burning and anthropogenic emissions, exploring the potential of establishing international cooperation mechanisms with neighboring Southeast Asian countries (particularly Vietnam, Laos, and Myanmar) on anthropogenic emission reduction and wildfire prevention was highly recommended.
Driven by Hainan’s strategic positioning as a National Ecological Civilization Demonstration Zone requiring world-class air quality, we further assessed the feasibility of achieving the newly proposed world-class ambient air quality standards (AAQS) by 2035 (
Song et al., 2024). Adopting the emission scenarios established by
Song et al. (2024), we implemented four future emission scenarios based on 2019 emission levels. For Hainan, three scenarios were applied, including “business as usual” (BAU), PC, and “maximum technically feasible reduction” (MTFR) scenarios. For surrounding regions of Hainan in China’s mainland, PC and “conservative” (CONS) scenarios were adopted. We focused on PM
2.5 concentrations in Hainan when implementing its BAU, PC, and MTFR scenarios while surrounding mainland regions maintained the PC scenario. Additionally, we assessed the compliance of PM
2.5 concentrations in Hainan under its PC scenario coupled with surrounding regions’ CONS scenario, which involves less stringent end-of-pipe control technologies and conservative emission reduction strategies. Using CMAQ sensitivity simulations, we evaluated compliance of annual mean PM
2.5 concentrations (against 10 μg/m
3) and the annual 95th percentile of daily average PM
2.5 concentrations (against 20 μg/m
3) across all 18 cities in Hainan under these four scenarios. For a more reasonable assessment of achievability, the projected PM
2.5 concentrations under future scenarios were corrected using observational data, which was multiplied by the ratio of the simulated value for 2035 to the simulated value for 2019. Simulation results showed that when Hainan and its surrounding regions simultaneously implemented policy control scenarios (PC_PC, PC_MTFR), both the annual mean PM
2.5 concentrations and the annual 95th percentile of daily PM
2.5 concentrations met the threshold values in all 18 Hainan cities (Fig. S4). Under the CONS_PC scenario, the annual 95th percentile of daily concentrations still met the threshold across Hainan. Furthermore, the annual mean PM
2.5 concentrations met the threshold value in all cities except Dongfang. These findings indicated that achieving the world-class air quality standard in Hainan was a realistic prospect. By taking a leading role in China’s air quality management efforts, Hainan could provide valuable experiences for air quality improvement in other regions of China and the world.
4 Conclusions
Determining background PM2.5 concentration was crucial for setting Hainan’s future emission reduction targets and achieving world-leading air quality standards. Hence, this study determined Hainan’s background PM2.5 concentration in 2019 by quantifying contributions from natural emissions and regional transport using the CMAQ model constrained by observational data. Source apportionment results indicated that natural sources contributed 19.0% to Hainan’s annual mean PM2.5 concentration. This comprised a 10.4% contribution from BVOCs and 8.6% from marine aerosols. Local anthropogenic emissions (49.5%) and regional transport (29.3%) accounted for the majority of Hainan’s annual PM2.5 concentration. The regional transport originated predominantly from anthropogenic emissions in surrounding regions (18.4%) and regional biomass burning (10.9%), while local biomass burning made a relatively small contribution (0.9%).
This study employed three definition methods to calculate Hainan’s PM2.5 background concentration, namely: 1) excluding anthropogenic emissions within Hainan; 2) excluding anthropogenic emissions within both Hainan and its surrounding regions; and 3) appli-cation of projected 2035 emission scenarios to the surrounding regions while excluding anthropogenic emissions within Hainan. Under these three definitions, the estimated annual mean PM2.5 background values averaged across the 18 cities in Hainan were 6.9, 4.2, and 5.2 μg/m3, respectively. Correspondingly, the grid-averaged PM2.5 background values for Hainan were 6.7, 4.6, and 4.9 μg/m3, respectively. Meteorological vari-ability analysis revealed further decrease in Hainan’s PM2.5 background concentration under 2020’s favo-rable meteorological conditions compared to 2019’s unfavorable conditions used above. Compared to background levels in other island regions globally, Hainan’s PM2.5 background concentration was found to be comparable.
Hainan’s PM2.5 background concentration was maintained at a low level, indicating significant potential for further air quality improvement. When both Hainan and its surrounding regions employed the policy scenario, the model demonstrated that the annual average PM2.5 concentration and the annual 95th percentile of daily average PM2.5 concentrations across all 18 cities in Hainan could meet the newly proposed AAQS standards. This finding indicated that achieving world-leading air quality levels in Hainan by 2035 was prospective.
The Author(s) 2025. This article is published with open access at link.springer.com and journal.hep.com.cn
PDF
(4095KB)
