We analyse the temporal variability of size-resolved ions and NAIS-measured particle concentrations using instrumental facilities established at the Atmospheric Observatory – Centre for Atmospheric Sciences (CAS-AO) (Rathore et al., 2025). The site is in Sonipat district, Haryana, India (28.99° N, 77.01° E; ∼ 125 ma.s.l.). It is a suburban site located ∼ 65 km from the IIT Delhi main campus and ∼ 10 km northwest of the northern boundary of Delhi. Agricultural land surrounds the site, while National Highways 44 and 34 intersect approximately 500 m from the sampling inlet. These highways represent the dominant nearby source of traffic-related emissions, in addition to advected regional sources. Industrial chimneys are also located beyond the intersecting highways. Residential areas lie approximately 300–400 m from the site and may contribute diesel-generator and household emissions. Small water bodies, including ponds and fish farms, are present nearby, and the Yamuna River flows within approximately 10 km of the site, potentially influencing local humidity and vapour precursors for particle formation and growth.

Figure 1 illustrates the measurement site's geographical context from Google Maps, including a zoomed-in view highlighting nearby roads and infrastructure (left), and a broader view showing its position within India (right). From a meteorological perspective, the CAS-AO acts as a receptor for air masses originating from upwind regions, including parts of the Middle East, Pakistan, and northern India. Its location, combined with proximity to major highways, agricultural activity, industrial sources, and residential areas, makes it well suited for investigating the interplay between local emissions, regional transport, and atmospheric processing of aerosols under subtropical conditions. The dataset used for the present paper is from May 2023 to December 2023.

In this study, we used a Neutral Cluster and Air Ion Spectrometer (NAIS) and a Scanning Mobility Particle Sizer (SMPS). We operated the instruments on two independent sampling lines. We measured meteorological parameters using an Automatic Weather Station (AWS) installed on the rooftop of the same building (I-Tech Park building IIT Delhi's Sonipat campus). Unless otherwise specified, all diameters referenced in this study refer to mobility diameters. A brief description of the instruments is as follows:

In this study, the seasons are demarcated according to specific time frames. Winter spans December–February, spring March–April, summer May–June, monsoon July–September, and autumn October–November, following the seasonal grouping used in previous Delhi aerosol analyses (Gani et al., 2020). We divide the diurnal cycle into two time periods: day (06:00–18:00 LT) and night (18:00–06:00 LT). All temporal references throughout this study are expressed in local time (LT), specifically in Indian Standard Time (IST; GMT+05:30). This structured approach to seasonal and temporal divisions facilitates a more comprehensive understanding of the data and enables us to draw meaningful comparisons across different time frames. We have used the ions and total particles data from NAIS negative channel.

Data coverage for the different instruments and derived quantities is summarised in Fig. S1 in the Supplement. Although the overall measurement period extended from May to December 2023, the number of valid days available for individual analyses differed because of instrument interruptions, missing data periods, precipitation-affected ion measurements, and gaps during the daytime growth window used for event classification. The NAIS-based NPF classification was performed using 108 valid NAIS days. SMPS measurements usable for the concurrent NAIS-SMPS analysis were available for a limited subset of the campaign (26 d), mainly during August–November 2023.

Data observed from AWS is presented in Fig. S2 which shows the typical seasonal diurnal variability of key meteorological parameters with significant variations in temperature (T), relative humidity (RH), wind speeds, and wind direction throughout the year. Winter (December–February) is characterised by low temperatures (∼ 2–18 °C), high relative humidity (∼ 35 %–95 %), and low wind speeds (∼ 1.5–4 ms-1). During winter, the site frequently experiences shallow atmospheric boundary layers (depth < 100 m), particularly at night and in the morning. In contrast, the summer months (April to June) are marked by intense heat (T ∼ 25 to 40 °C) and dry conditions (RH ∼ 30 %–70 %). The monsoon season, spanning from July to mid-September, brings episodic heavy rainfall to the majority of the IGP. This period is characterised by slightly lower temperatures (T ∼ 20 to 35 °C) compared to the summer. There may be some occurrence of relatively high temperatures, but mostly these are infrequent. Notably, the winds predominantly come from the northwest throughout most of the year, but during the summer and monsoon, they shift to south during the nighttime. Conditions unfavourable for atmospheric dispersion, such as low wind speeds, shallow boundary-layer heights, and stagnant high-pressure periods, often enhance pollutant accumulation, leading to exceptionally high PM2.5 concentrations (Guttikunda and Gurjar, 2012; Trivedi et al., 2014).

Researchers have proposed different classification frameworks in the literature to classify the observed particle number size distribution (PNSD) and identify the NPF events. Classification of PNSD is essential because it helps understand the frequency and characteristics of NPF events and enables quantification of particle formation and growth rates. However, the interpretation of NPF frequency is not solely a property of the atmospheric conditions but also depends on the choice of detection methodology, instrumentation sensitivity, and the specific classification logic applied.

One of the earliest systematic and widely adopted approaches for classifying NPF events was introduced by Dal Maso et al. (2005). This scheme is based on visual inspection of daily PSD and identifies an event when a distinct nucleation mode appears, persists for more than 1 h, and shows continuous growth toward larger diameters. Events are further categorised according to whether growth and formation rates can be reliably quantified: Class Ia denotes well-defined events, Class Ib denotes events that are visible but less clear than Class Ia, and Class II denotes less distinct events with weaker growth signatures. Days without such features are classified as non-events. This classification involves visual interpretation and is therefore inherently subjective and time intensive, but it remains a useful first-step diagnostic approach because it provides a structured and physically consistent way to separate event from non-event days in long-term aerosol datasets.

In the present study, we adopted the Dal Maso et al. (2005) classification framework with minor adjustments suited to our dataset. Distinguishing between subcategories Class Ia and Class Ib was not feasible, as most event-days exhibited an early and sharp burst of particles that obscured a clear separation between very strong and moderately strong events. Therefore, we grouped all clearly observable and well-developed growth events into Class I, while days showing NPF signatures under background influence or with less distinct growth were classified as Class II. Days without identifiable nucleation and growth were categorised as non-events. Days containing data gaps longer than 30 min were excluded. Days with even 30 min missing data during the critical daytime growth period were also discarded to avoid biased classification.

Another classification method, Dada et al. (2018) detects initial cluster formation signals by searching periods with elevated 2–4 nm ion concentrations and investigates NPF signals using periods with elevated 7–25 nm particle concentrations. Based on daily PNSD, the method classifies days into four categories: Regional Events: Elevated period observed in 2–4 nm ions and subsequently in 7–25 nm particles (local formation with growth), Ion Bursts: 2–4 nm ions present but no elevated concentration observed in larger particles, Transported events: 7–25 nm particles present without local ion formation (carried from elsewhere), and the Non-Events: Neither sufficient ion nor particle elevated concentration periods detected. This automated method identifies exact start, peak, and end times of events while excluding precipitation periods that interfere with ion measurements.

We applied this classification to observations from CAS-AO in a slightly modified way. Using monthly mean particle number concentrations above 12 nm, we observed a substantial contribution from background aerosol loading and instrument noise, particularly during the winter months, making it difficult to isolate the NPF growth signal in the full 7–25 nm range prescribed by Dada et al. (2018). Therefore, we restricted the particle-growth criterion to the 7–12 nm range. This site-specific adaptation improves robustness for the present high-background dataset, but it also means that the resulting classes are not directly equivalent to studies applying the standard 7–25 nm Dada et al. (2018) bounds.

An alternative and recently developed technique for analysing NPF events is the nanoparticle ranking analysis method proposed by Aliaga et al. (2023). This approach offers a quantitative and objective framework to evaluate the likelihood and strength of NPF events by examining daily changes in particle number concentrations within the 2.5–5 nm size range (ΔN2.5–5). This parameter is calculated from smoothed time series by contrasting background (night-time) and active (daytime) periods, yielding a single metric for each day that reflects the intensity of potential NPF. The days are then ranked according to this parameter and grouped into percentile classes or Gaussian modes of log(ΔN2.5–5) which allows the probability and strength of NPF to be examined continuously rather than via a binary event/non-event classification. This method produces a continuous variable that facilitates the probabilistic assessment of both strong and weak NPF occurrences.

In this study, we applied the nanoparticle ranking analysis following the framework of (Aliaga et al., 2023) to characterise NPF behaviour at CAS-AO. The ranking parameter was derived from the 3–5 nm particle number concentration (the lowest reliable size bin of our instrument), calculated as the difference between the daytime maximum concentration (06:00–18:00 LT) and the nighttime median background concentration (21:00–06:00 LT). This size range avoids contamination from the continuous molecular cluster mode at the lower end and reduces influence from primary anthropogenic emission sources, including traffic, at the upper end. A 2 h rolling median smoothing was applied prior to the calculation to reduce short-term variability. The resulting ΔN3–5 values were ranked across the full observational period and grouped into 5 % percentile intervals.

Growth rate (GR) is defined as the temporal rate of change of particle diameter (dDp/dt), expressed in nmh-1, representing the speed at which newly formed particles increase in size during growth events. In this study, GR was estimated using two independent approaches: maximum concentration method (using Gaussian function fitting) and appearance time method (using sigmoid function) (Kulmala et al., 2012). We have used particle number concentrations with size ranges (3–7, and 7–20 nm). Particle number concentration has been smoothed using a rolling median filter to avoid noisy signals and temporary fluctuations. For the maximum concentration approach, the temporal evolution of each size bin concentration was fitted using a least-squares Gaussian function to determine the peak time. The growth rate was then calculated by performing a linear regression between peak times and corresponding particle diameters, expressed in nmh-1. For the appearance method, the rising portion of each size-bin concentration curve was fitted with a sigmoid function to determine the 50 % appearance time (t50). Growth rates were derived by regressing t50 values against particle diameter. The resulting growth rates distributions were visualised using box-and-strip plots to represent central tendency, spread, and event-to-event variability while preserving individual data points. Before fitting, we examined the distribution's evolution over time to identify the window that represents the growth. Each fit and its peak-specific diameter range (in both categories) were reviewed to confirm that a representative time was reliably detected.

For GR estimation, the modified Dal Maso et al. (2005) classification was used to identify candidate event days. Class I and Class II days were considered because both categories can contain visually traceable growth features. Each candidate event day was then inspected individually. The nominal 3–7 and 7–20 nm intervals were used as initial search ranges, but the full interval was not imposed if part of the range was affected by noise, weak mode separation, plume-like variability, or discontinuous concentration enhancements. Instead, different contiguous sub-ranges and fitting windows within the nominal intervals were tested, and GR was calculated only for the portion that showed a coherent diameter–time growth pattern representative of NPF. Days or size intervals dominated by short-lived bursts or pseudo-growth signals were excluded from the GR statistics. This procedure yielded 17 NAIS event days suitable for at least one GR estimate.

The condensation sink (CS) was calculated from the aerosol PNSD following the diameter-based formulation described by Ezhova et al. (2018); Pirjola et al. (1999): CS = 2πD∑β(Dp)DpN(Dp), where D is the vapour diffusion coefficient, Dp is particle diameter, β(Dp) is the transition-regime correction factor, and N(Dp) is the particle number concentration in each size bin. The PNSD were obtained by combining NAIS measurements below 20 nm with SMPS measurements above 20 nm, resulting in a continuous size spectrum spanning the nucleation and accumulation modes for concurrent-instrument days.

For the event/non-event CS comparison, daily CS values were summarised as the median CS during 06:00–12:00 LT. This morning window was selected to represent the pre-onset, onset, and early-growth period of NPF while avoiding later afternoon boundary-layer accumulation and local emission effects. Of the 26 concurrent NAIS-SMPS days, 11 were NAIS event days and 15 were NAIS non-event days; 2 of the event days (6 September and 16 October 2023) were treated as disagreement/ambiguous cases because the SMPS particle size distribution did not support a clean event-day interpretation, leaving 9 event days and 15 non-event days for the condensation-sink comparison. Condensation sinks were therefore calculated only from these 9 event days and 15 non-event days, because particles outside the NAIS size range contribute substantially to these sinks. For CS calculations, NAIS particle data were used up to 20 nm and SMPS data above 20 nm, producing a merged PNSD of approximately 2.5–570 nm for concurrent-instrument days.

The particle formation rate (J) represents the rate at which new particles are formed at a given size and is expressed in cm-3s-1. Formation rates were determined for both 3 nm (J3) and 7 nm (J7), with the J3 equation shown as the representative formulation. These calculations were performed only for three case-study event days with concurrent NAIS and SMPS data and sufficient data quality for dN/dt, CoagS, and the growth-out-of-range term. For these days, the SMPS was available and a usable GR could be estimated, so CoagS and J were calculated from the merged NAIS + SMPS PNSD rather than from a truncated NAIS-only distribution. These values are therefore presented as case-study estimates rather than seasonal or campaign-wide averages.

The formation rate was calculated as: 1J3=dN3–7dt+CoagS⋅N3–7+GRΔDp⋅N3–7 where N3–7 is the particle number concentration in the 3–7 nm size range, dN3–7dt is the temporal change in concentration, CoagS is the coagulation sink representing losses due to coagulation with pre-existing particles, GR is the particle growth rate (nmh-1), and ΔDp is the width of the size interval (nm).

The first term represents the net increase in particle concentration, the second term accounts for losses due to coagulation, and the third term corrects for particles growing out of the size range. The coagulation sink was calculated using standard coagulation theory for interactions between newly formed particles and the background aerosol population (Kulmala et al., 2012; Lehtinen et al., 2007). For the three Table 1 case-study days, this term used the combined NAIS + SMPS PNSD, covering the available particle sizes from the NAIS lower particle range to the upper SMPS range (∼ 2.5–570 nm). To reduce the influence of short-term fluctuations and measurement noise, particle number concentrations were smoothed before calculating dN/dt. For J3, particle sizes between 3 and 7 nm were used; for J7, particle sizes between 7 and 20 nm were used. Formation rates were therefore calculated only for the smaller case-study subset of event days for which all required terms (CS from SMPS-NAIS data, GR from NAIS data) could be determined reliably.

Particle and ion number concentrations measured by the NAIS at CAS-AO exhibit pronounced diurnal and seasonal variability (Fig. 2). NAIS-measured particle concentrations in Fig. 2a show clear daytime enhancements during May–June, while these features become more episodic and delayed during August–September. In contrast, October–December exhibits weak and poorly structured diurnal variability, with no clearly defined growth patterns.

In May–June, particle growth typically begins early in the day (around 06:00 LT), while in August–September the onset appears somewhat later and more variable. This seasonal contrast is consistent with stronger photochemical activity and more favourable dilution conditions during the warmer months, whereas colder months are more often affected by high background aerosol loading and weaker boundary-layer mixing.

Ion number concentrations in Fig. 2b display broadly similar but generally more subdued features. Distinct midday enhancements are primarily observed during May and June, while distributions become progressively more uniform toward winter. Overall, the warmer months show clearer and more continuous growth signatures, indicating a strong seasonal modulation of NPF-related processes at the site. The smallest diameter ranges (below ∼ 3 nm for total particles and ∼ 2 nm for ions) are influenced by corona-generated charger ions and instrumental noise. Concentrations above 12 nm are dominated by pre-existing Aitken mode particles and background aerosol loading, making it difficult to isolate NPF signals at larger sizes. Horizontal dashed lines indicate the diameter ranges selected for subsequent analysis (3–12 nm for total particles and 2–4 nm for ions). These demarcated windows capture the size range where NPF signatures are most reliably observed and help refine the diameter thresholds used in standard NPF classification approaches.

Building on the size–time evolution shown in Fig. 2, the monthly statistical distributions of number concentrations within the selected diameter ranges are presented in Fig. 3. The figure shows total particles (7–12 and 3–5 nm) and ions (2–4 nm) as paired daytime and nighttime violin plots representing kernel density estimates. Within each violin, a box plot indicates the monthly median (central line), interquartile range (box), and 5 %–95 % percentiles (whiskers) for daytime (06:00–18:00 LT) and nighttime (18:00–06:00 LT) concentrations. Within the retained 3–12 nm range identified in Fig. 2, Fig. 3 presents two narrower sub-windows, 3–5 and 7–12 nm. The 3–5 nm window isolates particles closest to the NAIS lower detection limit, capturing the earliest stage of a growing mode while limiting the contribution of larger, locally influenced particles, including traffic emissions, that increasingly populate the distribution above about 5 nm in this urban-influenced setting.

The 7–12 nm window remains conservatively below the approximately 12 nm threshold beyond which background Aitken-mode aerosol dominates the signal, as established in Fig. 2. Across the 7–12 nm size range in Fig. 3a, particle concentrations show a seasonal and diurnal contrast. During summer (May–June), daytime median concentrations reach on the order of 3–5 × 10³ cm-3, approximately two to three times higher than nighttime levels (∼ 1–2 × 10³ cm-3). The distributions are also substantially broader during the day, with upper percentiles extending well above 8 × 10³ cm-3, indicating episodic but intense production events rather than uniform background enhancement. In contrast, August and September exhibit more moderate daytime increases (median ∼ 1–2 × 10³ cm-3) with reduced day–night contrast, while October–December show comparatively compact distributions and, in some cases (November–December), slightly higher nighttime medians, suggesting suppressed NPF and a stronger influence of background accumulation.

A similar pattern is observed for the 3–5 nm size range in Fig. 3b, which represents freshly formed particles (relatively). In May and June, daytime medians approach ∼ 3 × 10³ cm-3 and exceed nighttime values by a factor of ∼ 3–3.5, consistent with frequent and strong NPF activity. The enhancement weakens progressively toward late monsoon and post-monsoon months, with September and October showing only marginal day–night differences (∼ 1.0–1.1 ratio). By November and December, nighttime concentrations become comparable to or slightly higher than daytime levels, reflecting the absence of sustained nucleation bursts and more stable aerosol conditions. In contrast, ion concentrations in the 2–4 nm range in Fig. 3c remain much lower in absolute magnitude (generally 5–12 cm-3), with moderate daytime enhancement during May–August (day / night ratios ∼ 1.5–2.5). However, unlike particles, ions display weaker seasonal variability and in late autumn–winter (October–December) occasionally exhibit higher nighttime medians, suggesting that ion abundance alone does not scale directly with the intensity of particle growth. Together, these distributional analyses quantitatively confirm the seasonal and diurnal patterns inferred from the heatmaps: strong and sustained daytime particle production during warmer months, a gradual weakening through the monsoon transition, and comparatively stable, background-dominated conditions in winter.

New particle formation (NPF) events were identified across the observational period (May–December 2023) at CAS-AO, Sonipat, representing a suburban site located approximately 50 km upwind of Delhi in the IGP (Rathore et al., 2025). The site experiences substantial seasonal variability in meteorological conditions and aerosol loading (Sect. 3.1), both of which modulate the frequency and intensity of NPF. Given that the IGP is characterised by elevated background aerosol concentrations and high condensation sinks, the detection and classification of NPF events is non-trivial and depends significantly on the choice of methodology. To address this, we applied three complementary classification frameworks of increasing complexity and objectivity, as described in the methodology section, and the results are presented below.

During the post-monsoon and winter transition period over the Indo-Gangetic Plain, aerosol loadings are typically elevated and dominated by accumulation-mode particles, leading to enhanced condensation and coagulation sinks. Consequently, clear and sustained NPF events are generally less frequent during this season which has been demonstrated by the classification results in Sect. 3.2. Nevertheless, intermittent NPF events are occasionally observed under favourable combinations of reduced sink strength and supportive meteorological conditions. To illustrate the processes enabling NPF under such otherwise unfavourable background conditions, we examine in detail the period from 5–10 October 2023, which includes a well-defined NPF event on 7 October 2023.

Figure 11 illustrates the time-diameter evolution of particle number concentrations and sink parameters during 5–10 October 2023, encompassing both NPF and non-event days. The figure combines size-resolved measurements from (a) SMPS, (b) NAIS negative-ion channel, and the corresponding (c) condensation sinks. The SMPS and NAIS spectrograms illustrate the temporal evolution of particle populations across the nucleation, Aitken, and accumulation modes. Dashed horizontal lines mark reference diameters at 2 and 25 nm. All panels share a common time axis, enabling direct comparison between PNSD and sink dynamics over the same period.

Figure 11d–f summarises the synoptic and local meteorological conditions during the same period encompassing the NPF event on 7 October 2023. Wind speeds exhibit moderate variability throughout the period, with the event-day characterised by predominantly north-westerly flow and moderate wind speeds, conditions associated with enhanced ventilation and lower background aerosol accumulation. Wind direction shows intermittent variability on non-event days, suggesting changing air mass influences that may contribute to fluctuating aerosol loadings. Concurrent low relative humidity (∼ 35 %) and elevated temperature (∼ 30 °C) on 7 October 2023 further support favourable nucleation conditions. Similar meteorological features may also occur on non-event days; however, NPF occurrence reflects the combined influence of favourable meteorology, precursor availability, and sufficiently reduced sink conditions, no single factor being independently sufficient. On 7 October, the daily median CS was 0.0158 s-1 (IQR: 0.0078–0.0199 s-1), and during 06:00–12:00 LT it was 0.0093 s-1 (IQR: 0.0056–0.0185 s-1). Thus, the event occurred during a low-CS morning period relative to surrounding polluted conditions.

The combined PNSD from NAIS and SMPS, together with sink parameters, show a temporal correspondence between particle population dynamics and sink strength. Periods with enhanced nucleation-mode particle number concentrations in the NAIS spectrogram consistently coincide with reduced CS values. In contrast, days with persistently elevated sink values show no corresponding clear event-specific growth features in the nucleation-mode range, indicating conditions unfavourable for particle formation and early growth.

To place observations from the CAS-AO in a broader urban pollution context, the CS values obtained in this study are interpreted alongside reported environmental ranges and process-based thresholds governing NPF. As established in Sect. 3.2, event-day CS values during 06:00–12:00 LT (∼ 0.0095 s-1) were lower than non-event values (∼ 0.0292 s-1), indicating a reduction in vapour scavenging during periods of particle formation. This reduction enhances the survival probability of freshly nucleated clusters by increasing vapour lifetime and reducing coagulation losses.

Reported values from polluted urban environments, such as those presented by Gani et al. (2020); Laakso et al. (2006), span a wide range depending on pollution level and time of day, with CS values reaching ∼ 0.14–0.17 s-1 under highly polluted winter evening conditions, ∼ 0.035 s-1 during relatively cleaner spring daytime periods, and as low as ∼ 0.004 s-1 during monsoon afternoons. In this framework, the CAS-AO site represents an intermediate regime: event-day CS values (∼ 0.0095 s-1) are comparable to cleaner urban daytime conditions, while non-event values (∼ 0.0292 s-1) approach moderately polluted regimes.

Further observations from Delhi show that NPF events are typically observed when CS remains below ∼ 0.06 s-1, despite an overall range of ∼ 0.004–0.8 s-1, indicating that only periods with sufficiently reduced sink conditions allow particle formation and sustained growth (Ali et al., 2025). The event-day CS values in this study fall well within this favourable regime, whereas non-event conditions approach this threshold, indicating a transition between conditions that permit and inhibit NPF.

Importantly, the growth rates in this study exhibit a clear size dependence. The event-wise dataset gives median values of 10.6 and 14.7 nmh-1 in the nominal 3–7 nm range using the appearance-time and Gaussian/mode-peak methods, respectively, and 16.8 and 19.0 nmh-1 in the nominal 7–20 nm range. These values are comparable to the range of 5.7–30 nmh-1 reported for Delhi in previous studies under highly polluted conditions (Aliaga et al., 2023; Mönkkönen et al., 2004; Sarangi et al., 2015).

The reported CS and growth rate values at CAS-AO suggest that there exist sufficient sources of vapours that can form clusters and enable rapid growth, maintaining a sufficiently high survival probability despite the polluted background. Variable wind fields, radiation, and sink strength can disrupt the balance between vapour production and particle losses, leading to fragmented or suppressed nucleation signatures. While nucleation may still occur intermittently, the combined effect of rapid vapour removal and efficient particle scavenging strongly limits particle survival and growth. In contrast, the comparatively lower sink environment at Sonipat permits temporal windows favourable for particle survival and intermittent growth under suitable meteorological conditions.