Hasil untuk "Meteorology. Climatology"

B. Lloyd‐Hughes, M. Saunders

B. Ayarzagüena, A. H. Butler, P. Hitchcock et al.

<p>Several studies highlight the relevance of considering polar winter stratospheric information such as the occurrence of Sudden Stratospheric Warmings (SSWs) for skillful Subseasonal to Seasonal (S2S) surface climate predictions. However, current S2S forecast systems can only predict these events about two weeks in advance. A potential way of increasing their predictability is to improve the models' representation of the triggering mechanisms of SSWs. Traditional theories indicate that SSWs follow sustained wave dissipation in the stratosphere, but the relative role of tropospheric versus stratospheric conditions in the enhancement of stratospheric wave activity remains unclear.</p> <p>This study aims to quantify the role of the stratospheric state in wave activity preceding SSWs by analyzing three recent SSWs: the boreal SSWs of 2018 and 2019 and the austral minor SSW of 2019, using specific sets of S2S experiments. These ensembles follow the SNAPSI (Stratospheric Nudging And Predictable Surface Impacts) guidelines and include free-evolving atmospheric runs and nudged simulations, where the zonally-symmetric stratospheric state is nudged to either observations of a certain SSW or a climatological state. Our results show that the models struggle to capture the strong enhancement of wave activity preceding the 2018 SSW, limiting predictability beyond 10 d. In contrast, both SSWs of 2019 are better predicted, consistent with a more accurate simulation of the wave activity. Nudging the zonal mean stratospheric state does not drastically influence the upward wave activity flux or tropospheric circulation anomalies prior to these SSWs, but it has some impact on the stratospheric wave activity, although this modulation depends on the event characteristics. The boreal 2019 SSW appears to be primarily driven by tropospheric processes. In contrast, stratospheric contributions may have also played an important role in triggering the boreal 2018 SSW and the austral 2019 SSW. Understanding these variations is key to improving SSW predictability in S2S models.</p>

Pablo Rodríguez-Bocca, Guillermo Pereira, Diego Kiedanski et al.

In recent years, great progress has been made in the field of forecasting meteorological variables. Recently, deep learning architectures have made a major breakthrough in forecasting the daily average temperature over a ten-day horizon. However, advances in forecasting events related to the maximum temperature over short horizons remain a challenge for the community. A problem that is even more complex consists in making predictions of the maximum daily temperatures in the short, medium, and long term. In this work, we focus on forecasting events related to the maximum daily temperature over medium-term periods (90 days). Therefore, instead of addressing the problem from a meteorological point of view, this article tackles it from a climatological point of view. Due to the complexity of this problem, a common approach is to frame the study as a temporal classification problem with the classes: maximum temperature "above normal", "normal" or "below normal". From a practical point of view, we created a large historical dataset (from 1981 to 2018) collecting information from weather stations located in South America. In addition, we also integrated exogenous information from the Pacific, Atlantic, and Indian Ocean basins. We applied the AutoGluonTS platform to solve the above-mentioned problem. This AutoML tool shows competitive forecasting performance with respect to large operational platforms dedicated to tackling this climatological problem; but with a "relatively" low computational cost in terms of time and resources.

Sumanth Varambally, Marshall Fisher, Jas Thakker et al.

Foundation models for weather science are pre-trained on vast amounts of structured numerical data and outperform traditional weather forecasting systems. However, these models lack language-based reasoning capabilities, limiting their utility in interactive scientific workflows. Large language models (LLMs) excel at understanding and generating text but cannot reason about high-dimensional meteorological datasets. We bridge this gap by building the first agentic framework for weather science. Our framework includes a Python code-based environment for agents (ZephyrusWorld) to interact with weather data, featuring tools including a WeatherBench 2 dataset indexer, geolocator for geocoding from natural language, weather forecasting, climate simulation capabilities, and a climatology module for querying precomputed climatological statistics (e.g., means, extremes, and quantiles) across multiple timescales. We design Zephyrus, a multi-turn LLM-based weather agent that iteratively analyzes weather datasets, observes results, and refines its approach through conversational feedback loops. We accompany the agent with a new benchmark, ZephyrusBench, with a scalable data generation pipeline that constructs diverse question-answer pairs across weather-related tasks, from basic lookups to advanced forecasting, extreme event detection, and counterfactual reasoning. Experiments on this benchmark demonstrate the strong performance of Zephyrus agents over text-only baselines, outperforming them by up to 44 percentage points in correctness. However, the hard tasks are still difficult even with frontier LLMs, highlighting the challenging nature of our benchmark and suggesting room for future development. Our codebase and benchmark are available at https://github.com/Rose-STL-Lab/Zephyrus.

Koyena Ghosh, Suchismita Banerjee, Urna Basu et al.

Understanding pollutant meteorology interactions is essential for environmental risk assessment. This study develops an entropy-based statistical framework to analyze static and temporal dependencies between urban air pollutants and meteorological variables across multiple Indian cities. Dependence is quantified using complementary linear and nonlinear measures, including Pearson correlation, mutual information, and relative conditional entropy. A key methodological contribution is a PCA based composite indexing framework that integrates these heterogeneous metrics into a unified and interpretable correlation score. For each pollutant meteorological pair within a city, PCA is used to extract a joint variability index, while spatial variability is assessed by aggregating correlations across cities. These indices are further combined to derive a comprehensive city-level correlation score that represents overall pollutant meteorology coupling strength and enables classification of cities into distinct interaction regimes. Sensitivity analysis, performed by systematically excluding individual variable pairs, demonstrates the robustness of the framework, with no single pair exerting disproportionate influence. Temporal dependencies are examined using transfer entropy and time-delayed mutual information. Results indicate that relative humidity generally leads changes in pollutant concentrations, whereas ambient temperature tends to lag, highlighting contrasting causal influences. Mutual information peaks at zero lag and decays rapidly, indicating strong short term interactions with limited persistence. Overall, the proposed framework provides a unified and interpretable approach for assessing complex pollutant meteorology interactions across diverse locations and time.

Daniil Sukhorukov, Andrei Zakharov, Nikita Glazkov et al.

Weather forecasting is not only a predictive task but an interpretive scientific process requiring explanation, contextualization, and hypothesis generation. This paper introduces AI-Meteorologist, an explainable LLM-agent framework that converts raw numerical forecasts into scientifically grounded narrative reports with transparent reasoning steps. Unlike conventional forecast outputs presented as dense tables or unstructured time series, our system performs agent-based analysis across multiple meteorological variables, integrates historical climatological context, and generates structured explanations that identify weather fronts, anomalies, and localized dynamics. The architecture relies entirely on in-context prompting, without fine-tuning, demonstrating that interpretability can be achieved through reasoning rather than parameter updates. Through case studies on multi-location forecast data, we show how AI-Meteorologist not only communicates weather events but also reveals the underlying atmospheric drivers, offering a pathway toward AI systems that augment human meteorological expertise and support scientific discovery in climate analytics.

Nicholas S. Novella, W. Thiaw

Christopher D. Roberts, Martin Leutbecher

Long-range ensemble forecasts are typically verified as anomalies with respect to a lead-time dependent climatological mean to remove the influence of systematic biases. However, common methods for calculating anomalies result in statistical inconsistencies between forecast and verification anomalies, even for a perfectly reliable ensemble. It is important to account for these systematic effects when evaluating ensemble forecast systems, particularly when tuning a model to improve the reliability of forecast anomalies or when comparing spread-error diagnostics between systems with different reforecast periods. Here, we show that unbiased variances and spread-error ratios can be recovered by deriving estimators that are consistent with the values that would be achieved when calculating anomalies relative to the true, but unknown, climatological mean. An elegant alternative is to construct forecast climatologies separately for each member, which ensures that forecast and verification anomalies are defined relative to reference climatological means with the same sampling uncertainty. This alternative approach has no impact on forecast ensemble means but systematically modifies the total variance and ensemble spread of forecast anomalies in such a way that anomaly-based spread-error ratios are unbiased without any explicit correction for climatology sample size. Furthermore, the improved statistical consistency of forecast and verification anomalies means that probabilistic forecast skill is optimised when the underlying forecast is also perfectly reliable. Alternative methods for anomaly calculation can thus impact probabilistic forecast skill, especially when anomalies are defined relative to climatologies with a small sample size. Finally, we demonstrate the equivalence of anomalies calculated using different methods after applying an unbiased statistical calibration.

Grzegorz Urban, Michał K. Kowalewski, Jakub Sawicki et al.

At the Polish Institute of Meteorology and Water Management – National Research Institute (the Polish acronym: IMGW-PIB), experimental parallel measurements of daily precipitation totals were carried out from August 2021 – September 2022. Measurements were made with 5 types of amateur rain gauges (A – CliMET CM1016; B – TFA Dostmann 47.1008; C – TFA Dostmann 47.1000; D – Davis 6466 AeroCone Rain Collector; E – TFA Dostmann 47.3005.01) and a manual Hellmann rain gauge (H), which was taken as reference. The results showed that the most frequent differences in the rain gauges tested are very small differences (in daily rainfall totals Δ≤0.1 mm). They account for 63–86% of the days. On the other hand, large (1.0<Δ≤5.0 mm) and very large (Δ>5.0 mm) differences are most frequent in type D, respectively: 12% and 5% of days. This type, as the only one of the tested types, is characterised by the occurrence of days with very large differences with respect to rain gauge H. The tested rain gauges of the same type do not show the same sign of difference with respect to the reference rain gauge, as there are both negative and positive differences. The C type rain gauge shows the smallest average difference of daily precipitation totals against the H-type rain gauge (+0.04 mm). On the other hand, the average differences for types A and B are negative and amount to –0.1 and –0.2 mm, respectively. In contrast, the average difference for type D is +0.2 mm. Some amateur rain gauges perform better or similar to operationally used rain gauges of automatic networks. Types A, C and B can be used in voluntary networks with appropriate and continuous human supervision and their results used, as an additional source of information, for operational purposes at IMGW-PIB or other hydrological and meteorological services. The rain gauge D is useful only during warm season. Type E was not suitable for the conditions of the experiment for technical reasons.

Chanh Kieu

Do tropical cyclones (TC) possess chaotic dynamics at any stage of their development? This is an open yet important question in current TC research, as it sets a limit on how much one can further improve intensity forecast in the future. This study presents a novel use of machine learning (ML) to quantify TC intensity chaos. By treating TC scales as input features for different ML models, we show that TC dynamics displays a limited predictability range of ~3 hours at the maximum intensity (PI) state under a fixed environment, which confirms the existence of a chaotic regime in TC development. Using the minimum central pressure as a metric for TC intensity could extend the predictability range up to 9 hours, yet the low-dimensional chaos of TC intensity is still captured in all ML models. Additional sensitivity experiments with different ML model configurations, the number of input features, or sampling frequency all confirm the robustness of such limited predictability for TC intensity, thus supporting the existence of low-dimensional chaos at the PI limit. The existence of such intensity chaos has a profound implication that TCs must possess an intrinsic intensity variability even under an idealized condition. This internal variability dictates a lower bound for the absolute intensity error in TC models regardless of how perfect the TC models or initial condition will be.

Zohreh Javanshiri, Ebrahim Asadi Oskouei, Fatemeh Abbasi

Precipitation fluctuation is a critical factor in understanding climate, as it has far-reaching impacts on the environment, economy, and society. Excessive or insufficient precipitation can result in severe floods and droughts, causing significant consequences for both nature and humanity. For this reason, it is vital to understand the variability, behavioral changes, and extreme precipitation levels to successfully manage industries like agriculture, water resources, urban planning, and transportation. In this study, to investigate the extreme values of precipitation, the trends of four precipitation indices of 44 synoptic stations in Iran were calculated and compared for two recent climate normal periods of 1981-2010 and 1991-2020. Four precipitation indices, specifically PRCPTOT95, PRCPTOT99, R10MM, and R20MM, were investigated after undergoing quality control and data homogenization. An evaluation of the trend of heavy precipitation in two periods demonstrated a decrease in the eastern and northwestern parts of the country, while an increase was observed in the Zagros and northern regions. The increasing trend in the second period has become more significant in the northern part of the country, while the decreasing trend in the western part has decreased in the second period. The analysis of mean precipitation indices in two periods showed that heavy precipitation had increased in the second period on the coasts of the Caspian Sea, Zagros Mountains, and southern parts and decreased in the east of Iran.

Amit Kumar Maurya, Somil Swarnkar, Shivendra Prakash

The Indian Ganga basin (IGB) is one of the most valuable socioeconomic regions in the Indian subcontinent. The IGB supports more than half a billion people due to an abundant supply of freshwater for agro-industrial purposes, primarily through Indian Summer Monsoon (ISM) rainfall contributions (∼85%). Any alterations in ISM characteristics would significantly affect freshwater availability, and as a result, socioeconomic activities would be affected. Therefore, in this study, we have attempted to assess how the monsoon rain spell characteristics, i.e. peak, volume, and duration, altered historically between 1901 to 2019. We further analyzed the specific IGB regions where monsoon rain spell changes are more prominent and their hydrological implications. Our estimates reveal that short-duration high-magnitude rain spells have significantly increased across the major regions of the IGB after 1960, which implies the increased probabilities of flash flood hazards. At the same time, the rain spell volumes have been depleted across the IGB after 1960, especially in the eastern Indo-Gangetic plains and southern IGB regions, indicating increased drought frequencies. Further, Himalayan regions, i.e. upper Ganga, upper Yamuna, and upper Ghaghra, have demonstrated increasing magnitudes of rain spell peaks, volume, and duration post-1960. In addition, the continuous warming and anthropogenic alterations might further exaggerate the current situation. Thus, these inferences are helpful for river basin management strategies to deal with the extreme hydrological disasters in the IGB.

Zeyu Lyu, Kohei Ichikawa, Yongchao Cheng et al.

Historical weather records from Europe indicate that the Earth experienced substantial climate variability, which caused, for instance, the Little Ice Age and the global crisis in the period between the 14th and 19th centuries. However, it is still unclear how global this climate variability was because of the scarce meteorological data availability in other regions including East Asia, especially around the 17th century. In this context, Seungjeongwon Ilgi, a daily record of the Royal Secretariat of the Joseon Dynasty of Korea, is a precious source of historical meteorological records for the Korean Peninsula, as it covers 288 years of weather observations made during 1623-1910. We used the digital database of Seungjeongwon Ilgi to construct a machine-readable weather condition dataset. To this end, we extracted valid weather information from the original weather description text and compiled them into predefined weather categories. Additionally, we attempted to improve the usability of the dataset by converting the reported dates in the traditional calendar system to those in the Gregorian calendar. Finally, we outlined the promising implications of this dataset for meteorological and climatological studies, while describing the limitations of the dataset. Overall, future studies focusing on the climate and weather of the past could use this meteorological database for investigating long-term climate variability. Our datasets are publicly available at 10.5281/zenodo.8142701.

Zbyněk Engel, Kamil Láska, Jan Kavan et al.

The retreat rates of Triangular Glacier since 1979 and its mass changes during the period 2014/15–2019/20 indicate the sensitive response of small ice masses on the eastern side of the Antarctic Peninsula to air temperature evolution. This cirque glacier in the northern part of James Ross Island receded rapidly during the period of regional warming in the late 20th century, losing 30.8% of its surface area between 1979 and 2006 (−1.7% a−1). The retreat rate then dropped to −0.3% a−1 following the regional cooling trend, but started to accelerate again (−0.8 to −2.3% a−1) with increasing air temperature since the summer 2014/15. Since the glaciological year 2015/16, Triangular Glacier has experienced enhanced snow melt, wind scour and permanent mass loss with annual mass balance ranging from −0.08 ± 0.35 to −0.56 ± 0.25 m w.e. The largest mass loss was observed in the glaciological year 2019/20, which included the warmest summer of the observation period. The cumulative mass balance of −1.66 ± 0.83 m w.e. over the years 2014/15–2019/20 is consistent with the termination of the positive mass-balance period that occurred in the north-eastern Antarctic Peninsula from 2009/10 to 2014/15.

M. Bacci, O.A. Idrissa, C. Zini et al.

The increasing frequency of extreme events in West Africa, such as droughts and floods, has made populations that base their subsistence mostly on rainfed agriculture even more vulnerable to climate threats. Climate Services (CS) are largely acknowledged as effective tools for tackling climate risks in agriculture, particularly in semi-arid developing countries but evidences of their effectiveness are still jeopardized. In Niger a climate service (CS) has been set up in the regions of Dosso and Tillabéri by the National Meteorological Service (NMS) to provide salient information for smallholder farmers. The CS was built on a robust collaboration among NMS, local extension services and authorities and farmers in 8 municipalities. The case study shows that a large share of farmers receives throughout the cropping season climatic information and advice via roving seminars and various media, including instant messaging services and radio broadcasts. Nevertheless, the results indicate that access to CS alone doesn’t imply relevant positive impacts on crop yields while the training of farmers in the use of the information results to be a significant factor. Indeed, in 2020, yields of trained farmers are significantly higher by around 17% compared to those of non-trained ones. Training and iterative interaction between farmers and NMS could also have indirect effects on information uptake, contributing to building reciprocal trust and therefore stronger action by trained farmers. The study confirms the importance of the social learning process in CS co-development. Since the study is limited by the small sample and the dataset covering only one cropping season, further research is needed to deepen cost-benefit analysis and disentangle the relative contribution of the CS components to yield increase. Indeed, evidences of the positive impact of CS could represent a leverage for local governments and international funders to support CS co-development and related capacity building activities. Practical implication: Climate variability and a strong increase in extreme hydro meteorological events are affecting agriculture production and exacerbating food insecurity in West Africa. In Niger, the vulnerability of agricultural production systems is coupled to ecosystem fragility and soil degradation. In this area, the rural population is the most vulnerable to climate threats because they have a reduced capability to implement effective risk reduction and climate change adaptation strategies and national government has limited resources to invest in climate policies. The CS implemented for the regions of Dosso and Tillabéri in Niger demonstrates that it is possible to set up an effective network for disseminating agrometeorological information for smallholder farmers at the municipal level with the aim of reducing the impact of climate threats on agriculture production. The information produced by the National Meteorological Service (NMS) is spread through extension services and rural radios to reach farmers. At the same time the agrometeorological field data are collected by local farmers and sent to the national service, ensuring the continuous monitoring of the cropping season. Subsequently, the agrometeorological information is coupled with setting up roving seminars in each municipality to spread tailored advice for farmers concerning seasonal climate forecasts and to build capacities in the use of agrometeorological advices during the season. During these seminars, rain gauges are also distributed to farmers and their use explained. In this way, farmers become able to autonomously take some tactical decisions, such as better timing the sowing of crops or performing farming activities, basing these choices on direct observations. The present case study demonstrates that the mere receipt of the climate information is not clearly related to an increase in yields; contrariwise, farmers who received training on how to properly use the information, have significantly higher yields. Repeated capacity building and information distribution over the years represent an element of trust building for end users who are more prone to use these CS in their agricultural choices, integrating their traditional knowledge. The next challenge is to guarantee the sustainability of these networks over time, because, even if technology advances could reduce the costs of the production and distribution of climate services, the training activities and maintaining the rural observation network are challenging. A possible way to make it sustainable is to reinforce institutional collaboration. Moreover, the use of a participatory approach in co-designing the CS could be a key element in pursuing the active involvement of the local population and administrations and could increase their motivation in the data exchange process. Basing on obtained results, the authors recommend to pursue the development of tailored CS for smallholder farmers in similar rural contexts, since these services constitute a real contribution to climate change adaptation at the local level in rural areas and future experiences could ensure the fine tuning of the climate information products, reducing delivery costs and increasing benefits for stakeholders. Finally, it is also recommended to further assess the cost/benefit ratio of CS in order to leverage funds and ensure scaling up and sustainability.

Atef M. F. Mohammed, Inas A. Saleh, Hend R. Zahran et al.

This study was conducted in one of a diverse industrial area in Al-Akrasha, Egypt. Concentrations of select metals (Cu, Pb, Cr, Ni, Zn, Mn, Cd, Al, Ag, As, B, and Fe) were evaluated in ambient PM₁₀ and surface soils at nine sites. Random samples of fresh edible tilapia fish were collected from Ismailia Canal at two sites near the Al-Akrasha region. In addition, blood and hair samples were collected from workers and residents living in Al-Akrasha as biomarkers of contamination with these metals. The ecological and health risks of these metals to the workers and residents living in the Al-Akrasha region were assessed. The results showed that heavy metal levels in the ambient air (PM₁₀) of the Al-Akrasha region were higher than the national and international guidelines. There was a very high degree of contamination (CD > 32) of the surface soil in the Al-Akrasha area, which can be attributed to industrial activities emissions, mostly from smelters and the subsequent deposition on the surface soil. Ingestion was the dominant pathway for metals to enter the human body in the Al-Akrasha region. Adults have a higher daily intake and exposure risk than infants and children.

Annick Pouquet, Duane Rosenberg, Raffaele Marino et al.

Many issues pioneered by Jackson Herring deal with how nonlinear interactions shape atmospheric dynamics. In this context, we analyze new direct numerical simulations of rotating stratified flows with a large-scale forcing, which is either random or quasi-geostrophic (QG). Runs were performed at a moderate Reynolds number <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi>R</mi><mi>e</mi></mrow></semantics></math></inline-formula> and up to 1646 turn-over times in one case. We found intermittent fluctuations of the vertical velocity <i>w</i> and temperature <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mi>θ</mi></semantics></math></inline-formula> in a narrow domain of parameters as for decaying flows. Preliminary results indicate that parabolic relations between normalized third- and fourth-order moments of the buoyancy flux <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mo>∝</mo><mfenced separators="" open="⟨" close="⟩"><mi>w</mi><mi>θ</mi></mfenced></mrow></semantics></math></inline-formula> and of the energy dissipation emerge in this domain, including for passive and active scalars, with or without rotation. These are reminiscent of (but not identical to) previous findings for other variables and systems such as oceanic and atmospheric flows, climate re-analysis data, fusion plasmas, the Solar Wind, or galaxies. For QG forcing, sharp scaling transitions take place once the Ozmidov length scale <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msub><mi>ℓ</mi><mrow><mi>O</mi><mi>z</mi></mrow></msub></semantics></math></inline-formula> is resolved—<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msub><mi>ℓ</mi><mrow><mi>O</mi><mi>z</mi></mrow></msub></semantics></math></inline-formula> being the scale after which a turbulent Kolmogorov energy spectrum likely recovers at high <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi>R</mi><mi>e</mi></mrow></semantics></math></inline-formula>.

R. Seager, A. Hooks, A. P. Williams et al.

A. Turchi, E. Masciadri, C. Veillet

ALTA project has been active since 2016, providing, at LBT observatory site, forecasts of atmospheric parameters, such as temperature, wind speed and direction, relative humidity and precipitable water vapor, and optical turbulence parameters, such as seeing, wavefront coherence time and isoplanatic angle with the final goal to support nightly the science operation of the LBT. Besides to the forecasts, during the years ALTA has been collecting statistics on the atmospheric conditions which can be used to draw a very accurate characterization of the climatology of the telescope site located on top of Mount Graham, Arizona. Such characterization can be used both for the optimization and calibration of the forecast model and as a reference for a model validation. The climatology of these parameters is conceived to be a further output of ALTA that will be upgraded on the website with time and it will be able to put in evidence trends at short as well as long time scales. In this contribution we present a climatological description of all the atmospheric parameters relevant for ground-based astronomy in order to provide to the scientific community a robust reference of the bserving conditions at LBT. The study is performed using on-site measurements provided by DIMM and atmospheric sensors over several years and made available in the telescope telemetry data.

Rodrigo Alonso-Suárez, Daniel Aicardi, Franco Marchesoni-Acland

In this work we analyse a set of benchmark methods for solar irradiance forecasting based on the clear-sky index, namely, persistence, climatology, smart-persistence and convex combination (CC) of persistence and climatology. To assess the boundaries of the benchmarks, we include in the analysis a simple least squares regression technique that uses the last five past samples. For the methods that need data adjustment, we also analyze their variation with different training sets. This work is done for intra-day forecast of global solar horizontal irradiance with 10-minutes and hourly forecast horizons. Our results, obtained in the south-east region of South America known as Pampa Húmeda, confirm the benchmark recommendation, showing that the CC method is exigent for RMSD as it achieves almost the same performance of the more complex linear combination of past measurements. We also observe that the benchmarks have different behavior depending on the metric, as the most exigent benchmark for one metric is not necessarily the most exigent for another.