Abstract. Condensation trail (contrail) cirrus clouds cause a substantial fraction of aviation's climate impact. One proposed method for the mitigation of this impact involves modifying flight paths to avoid particular regions of the atmosphere that are conducive to the formation of persistent contrails, which can transform into contrail cirrus. Determining the success of such avoidance maneuvers can be achieved by ascertaining which flight formed each nearby contrail observed in satellite imagery. The same process can be used to assess the skill of contrail forecast models. The problem of contrail-to-flight attribution is complicated by several factors, such as the time required for a contrail to become visible in satellite imagery, high air traffic densities, and errors in wind data. Recent work has introduced automated algorithms for solving the attribution problem, but it lacks an evaluation against ground-truth data. In this work, we present a method for producing synthetic contrail detections with predetermined contrail-to-flight attributions that can be used to evaluate – or “benchmark” – and improve such attribution algorithms. The resulting performance metrics can be employed to understand the implications of using these observational data in downstream tasks, such as forecast model evaluation and the analysis of contrail avoidance trials, although the metrics do not directly quantify real-world performance. We also introduce a novel, highly scalable contrail-to-flight attribution algorithm that leverages the characteristic compounding of error induced by simulating contrail advection using numerical weather models. The benchmark shows an improvement of approximately 25 % in precision versus previous contrail-to-flight attribution algorithms, without compromising recall.

Abstract. Rain gauge measurements are one of the primary techniques used to estimate a precipitation field, but they require careful quality control. This paper describes a modified RainGaugeQC system, which is applied to real-time quality control of rain gauge measurements made every 10 min. This system works operationally at the national meteorological and hydrological service in Poland. The RainGaugeQC algorithms, which have been significantly modified, are described in detail. The modifications were made primarily to control data from non-professional measurement networks, which may be of lower quality than professional data, especially in the case of personal stations. Accordingly, the modifications went in the direction of performing more sophisticated data control, applying weather radar data, and taking into account various aspects of data quality, such as consistency analysis of data time series and bias detection. The effectiveness of the modified system was verified based on independent measurement data from manual rain gauges, which are considered one of the most accurate measurement instruments, although they mostly provide daily totals. In addition, an analysis of two case studies is presented. This highlights various issues involved in using non-professional data to generate multi-source estimates of the precipitation field.

Abstract. PurpleAir sensors offer a unique opportunity for a large-scale and densely populated array of sensors to study surface air quality. While PurpleAir sensors are inexpensive and abundant, they must be corrected to better agree with validated coincident measurements from more sophisticated instrumentation. Traditionally, this correction is performed using statistical methods. We propose a method to both correct PurpleAir PM2.5 measurements and allow for an estimate of the hygroscopic growth of aerosols, using a novel correction approach based on the optimal estimation method (OEM). The hygroscopic growth of aerosols can be retrieved using the sensitivity of the correction to water activity, which influences the measured size distribution of the aerosols. By employing the physically based correction using calibrated measurements from the nearby Ontario Ministry of the Environment, Conservation and Parks air quality site, the average daily mean absolute error (MAE) of the PurpleAir PM2.5 measurements is decreased from 5.58 to 1.68 µg m−3, and the average daily bias decreases from 4.75 to −0.23 µg m−3. This improvement in the correction is comparable to that seen using conventional statistical methodologies. Our OEM retrieval also allowed us to estimate seasonal bulk hygroscopicity values ranging from 0.33 to 0.40. These values are consistent with the accepted ranges of bulk hygroscopicity for atmospheric particulate matter (0.1 to 0.9) determined in previous studies using calibrated air quality measurement instruments, which suggests that our method allows a new aerosol product to be determined from a large sensor network.

Abstract. Particulate nitrate is a major component of ambient aerosol around the world, present in inorganic form, mainly as ammonium nitrate, and also as organic nitrate. It is of increasing importance to monitor ambient particulate nitrate, a reservoir of urban nitrogen oxides that can be transported downwind and harm ecosystems. The unit-mass-resolution time-of-flight aerosol chemical speciation monitor equipped with capture vaporizer (CV-UMR-ToF-ACSM) is designed to quantitatively monitor ambient PM2.5 composition. In this paper, we describe a method for separating the organic and ammonium nitrate components measured by CV-UMR-ToF-ACSM based on evaluating the NO2+/NO+ ratio (NOx+ ratio). This method includes modifying the ACSM fragmentation table, time averaging, and data filtering. By using the measured NOx+ ratio of NH4NO3 and a plausible range of NOx+ ratio for organic nitrate aerosol, the measured particulate nitrate can be split into inorganic and organic fractions. Data pre-treatment filters concentrations of particulate nitrate below 0.6–2.0 µg m−3, depending on the time averaging. The method detection limit, when considering ±10 % absolute uncertainty of organic nitrate fraction, is found to be 2 µg m−3 (120 min averaging) to 10 µg m−3 (10 min averaging) for total particulate nitrate concentration and 10 % (120 min) to 20 % (10 min) for organic nitrate fraction. We show that this method is able to distinguish periods with inorganic or organic nitrate as major components at a rural site in the Netherlands. A comparison to a high-resolution time-of-flight aerosol mass spectrometer equipped with a standard vaporizer (SV-HR-ToF-AMS) and positive matrix factorization (PMF) method shows similar response of increasing particulate organic nitrate fraction with uncertainties mainly from sensitivity to fragmentation table correction when obtaining the NO2+ signal. We propose that researchers use this NOx+ ratio method for CV-UMR-ToF-ACSM (adapting the appropriate fragmentation table and data pre-treatment for each specific application) to quantify the particulate organic nitrate fraction at existing monitoring sites in order to improve understanding of nitrate formation and speciation.

Abstract. Because of their high sensitivity to hydrometeors and high vertical resolutions, spaceborne radar observations are emerging as an undeniable asset for numerical weather prediction (NWP) applications. The EUMETSAT (European Organisation for the Exploitation of Meteorological Satellites) NWP SAF (Satellite Application Facility for Numerical Weather Prediction) released an active sensor module within version 13 of the RTTOV (Radiative Transfer for TIROS Operational Vertical Sounder) software with the goal of simulating both active and passive microwave instruments within a single framework using the same radiative transfer assumptions. This study provides an in-depth description of this radar simulator. In addition, this study proposes a revised version of the existing melting-layer parameterization scheme of Bauer (2001) within the RTTOV-SCATT v13.1 model to provide a better fit to observations below the freezing level. Simulations are performed with the revised and default schemes for the Dual-frequency Precipitation Radar (DPR) instrument on board the Global Precipitation Measurement (GPM) mission using the ARPEGE (Action de Recherche Petite Echelle Grande Echelle) global NWP model, operational at Météo-France for two different 1-month periods (June 2020 and January 2021). Results for a case study over the Atlantic Ocean show that the revised melting scheme produces more realistic simulations much closer to observations compared to the default scheme both at Ku (13.5 GHz) and Ka (35.5 GHz) frequencies. A statistical assessment covering several cases shows significant improvement of the first-guess departure statistics. As a step further, this study showcases the use of melting-layer simulations for the classification of precipitation (stratiform, convective, and transition) using the dual-frequency ratio (DFR) algorithm. The classification results reveal a significant overestimation of the rain reflectivities in both hemispheres, which could indicate either an overproduction of convective precipitation in ARPEGE or a misrepresentation of the convective precipitation fraction within the forward operator.

Abstract. The shapes of ice crystals play an important role in global precipitation formation and the radiation budget. Classifying ice crystal shapes can improve our understanding of in-cloud conditions and these processes. Existing classification methods rely on features such as the aspect ratio of ice crystals, environmental temperature, and so on, which bring high instability to the classification performance, or employ supervised machine learning algorithms that heavily rely on human labeling. This poses significant challenges, including human subjectivity in classification and a substantial labor cost in manual labeling. In addition, previous deep learning algorithms for ice crystal classification are often trained and evaluated on datasets with varying sizes and classification schemes, each with distinct criteria and a different number of categories, making it difficult to make a fair comparison of algorithm performance. To overcome these limitations, a contrastive semi-supervised learning (CSSL) algorithm for the classification of ice crystals is proposed. The algorithm consists of an upstream unsupervised learning network tasked with extracting meaningful representations from a large number of unlabeled ice crystal images, and a downstream supervised network is fine-tuned with a small subset of labeled images of the entire dataset to perform the classification task. To determine the minimum number of ice crystal images that require human labeling while maintaining the algorithm performance, the algorithm is trained and evaluated on datasets with varying sizes and numbers of categories. The ice crystal data used in this study were collected during the NASCENT campaign at Ny-Ålesund and CLOUDLAB project on the Swiss Plateau using a holographic imager mounted on a tethered balloon system. In general, the CSSL algorithm outperforms a purely supervised algorithm in classifying 19 categories. Approximately 154 h of manual labeling can be avoided using just 11 % (2048 images) of the training set for fine tuning, sacrificing only 3.8 % in overall precision compared to a fully supervised model trained on the entire dataset. In the four-category classification task, the CSSL algorithm also outperforms the purely supervised algorithm. When fine-tuned on just 2048 images (25 % of the dataset), it achieves an overall accuracy of 89.6 %, nearly matching the 91.0 % accuracy of the supervised algorithm trained on 8192 images. When tested on the unseen CLOUDLAB dataset, CSSL shows superior generalization, improving accuracy by an average of 2.19 %. Our analysis also reveals that both CSSL and purely supervised algorithms exhibit inherent instability when trained on small dataset sizes, and the performance difference between them converges as the training set size exceeds 2048 samples. These results highlight the strength and practical effectiveness of CSSL in comparison to purely supervised methods and the potential of the CSSL algorithm to perform well on datasets collected under different conditions.

Abstract. The aerosol layer height (ALH), from the Sentinel-5P/TROPOMI L2_AER_LH product, is based on an optimal estimation (OE) approach, fitting cloud-free measurements to synthetic reflectances in the strongest oxygen absorption band, provided by a neural network trained with high-resolution simulated reflectances. The ALH has been continuously improved since its release in 2019, focusing especially on (bright) land surfaces, over which the ALH product showed underestimated ALHs (biased towards the surface). This paper describes the latest updates of the ALH product, which includes first the introduction of the directional Lambertian-equivalent reflectance (DLER) climatology to improve the surface albedo characterization over land. Second, the paper describes a further improvement, adding the surface albedo in the feature vector of the OE inversion, using the DLER as prior information. Using this approach, the retrievals over land largely match the retrievals over ocean, which have shown a good comparison with validation data since its release, most notably with Cloud-Aerosol LIdar with Orthogonal Polarization (CALIOP) weighted extinction heights. The albedo is fitted for both land and ocean surfaces, but the implementation is different over land and ocean because of the large range of land surface albedos. Over land, the a priori surface albedo values are relaxed so the fitting procedure can incorporate the albedo effects in the retrieval over land. Over ocean, the retrievals are optimized by tuning the a priori error settings. The current implementation improves retrievals over land with approximately 1.5 times more converged results and decreases land–ocean contrasts in the ALH retrievals. The average difference between CALIOP weighted extinction height decreased for selected cases from approximately −1.9 to −0.9 km over land and from approximately −0.8 to +0.1 km over ocean. An independent verification with ATmospheric LIDar (ATLID) data from EarthCARE showed consistent behaviour between the new operational data and the test cases.

Abstract. Recent advancements in single-particle mass spectrometry (SPMS) have enabled the detection of aromatic hydrocarbons at the individual particle level in conjunction with inorganic/refractory particle components. However, the laser desorption (LD) of organic material from particles prior to their ionization in a two-step process necessitates pulsed infrared lasers with adequate pulse energy that can be irregularly triggered on detected particles. Pulsed CO2 lasers with a 10.6 µm wavelength have been traditionally utilized, yet these lasers are bulky and costly and require regular maintenance, including gas exchange or a continuous laser gas supply. In this study, we present the application of a prototype solid-state laser based on an erbium-doped yttrium aluminum garnet (Er:YAG) crystal, emitting long pulses of 200 µs at 3 µm wavelength as a compact, cost-effective, and user-friendly alternative for LD. We directly compared the new laser with a commonly used CO2 laser and found similar performance in LD for both laboratory particles and ambient air experiments. With the exception of slightly increased fragmentation observed with the CO2 laser due to its beam profile, no qualitative differences were noted in the resulting mass spectra. Additionally, we compared a novel two-step ionization (LD-REMPI–LDI) with the conventional single-step LDI regarding the potential to detect polycyclic aromatic hydrocarbons (PAHs) and inorganics in laboratory and field experiments. The combined methods demonstrated superior performance in the detection of PAHs, for both the CO2 and the new Er:YAG laser. In addition to its higher sensitivity and lower fragmentation for PAHs when compared to single-step LDI, it is less dependent on the particle matrix, sharing the benefits of traditional two-step methods but extending its capability to combine PAH measurements with the LDI-based detection of inorganic particle compounds.

Abstract. The Three-Dimensional Precipitation Particle Imager (3D-PPI) is presented as a new instrument for measuring the size, fall velocity, and three-dimensional shape of precipitation particles. The 3D-PPI consists of three high-resolution cameras with telecentric lenses and one high-speed camera with one non-telecentric lens. The former captures high-resolution images of falling particles from three angles, enabling three-dimensional (3D) shape reconstruction via a 3D reconstruction algorithm, while the large observation volume facilitates particle size distribution (PSD) analysis. The latter records the images of the falling precipitation particles with 200 frames per second, based on which the falling velocity of particles can be calculated. The field experiment of the 3D-PPI and OTT Parsivel disdrometer (OTT) was conducted in Tulihe, China, with more than 880 000 snowflakes recorded during a typical snowfall case lasting 13 h, and the results indicate strong agreement between the PSDs obtained by the 3D-PPI and those obtained by the OTT. This demonstrates its potential application in atmospheric science, polar research, and other fields.

Abstract. We used eight co-located global navigation satellite system (GNSS) stations with different antenna mounts to estimate atmospheric signal propagation delays in the zenith direction and linear horizontal gradients. The gradients are compared with the results from a water vapour radiometer (WVR). The water drops in the atmosphere have a negative influence on the retrieval accuracy of the WVR. Hence, we see better agreement using WVR data with a liquid water content (LWC) less than 0.05 mm compared to when LWC values of up to 0.7 mm are included. We used two different constraints when estimating the linear gradients from the GNSS data. Using a weak constraint enhances the GNSS estimates to track large gradients of short duration at the cost of increased formal errors. To mitigate random noise in the GNSS gradients, we adopted a fusion approach, averaging estimates from the GNSS stations. This resulted in significant improvements for the agreement with the WVR data, a maximum of 17 % increase in the correlation, and a 14 % reduction in the root mean square (RMS) difference for the east gradients. The corresponding values for the north gradients are both 25 %. Overall, no large differences in terms of quality are observed for the eight GNSS stations. However, one station shows slightly poorer agreement for the north gradients compared to the others. This is attributed to the station's proximity to a radio telescope, which causes data loss of observations at low-elevation angles in the south-south-west direction.