Hasil untuk "Meteorology. Climatology"

Xiannan Huang, Shen Fang, Shuhan Qiu et al.

Time series forecasting plays a critical role in domains such as transportation, energy, and meteorology. Despite their success, modern deep forecasting models are typically trained to minimize point-wise prediction loss without leveraging the rich information contained in past prediction residuals from rolling forecasts - residuals that reflect persistent biases, unmodeled patterns, or evolving dynamics. We propose TEFL (Temporal Error Feedback Learning), a unified learning framework that explicitly incorporates these historical residuals into the forecasting pipeline during both training and evaluation. To make this practical in deep multi-step settings, we address three key challenges: (1) selecting observable multi-step residuals under the partial observability of rolling forecasts, (2) integrating them through a lightweight low-rank adapter to preserve efficiency and prevent overfitting, and (3) designing a two-stage training procedure that jointly optimizes the base forecaster and error module. Extensive experiments across 10 real-world datasets and 5 backbone architectures show that TEFL consistently improves accuracy, reducing MAE by 5-10% on average. Moreover, it demonstrates strong robustness under abrupt changes and distribution shifts, with error reductions exceeding 10% (up to 19.5%) in challenging scenarios. By embedding residual-based feedback directly into the learning process, TEFL offers a simple, general, and effective enhancement to modern deep forecasting systems.

Caroline B. Alden, Doug Chipponeri, David Youngquist et al.

Methane emissions at oil and gas facilities can be measured in real time with continuous monitoring systems that alert operators of upset conditions, including fugitive emissions. We report on extensive operator field testing of a continuous laser monitoring system in ~year-long deployments at 46 oil and gas sites in two U.S. basins. The operator assessed periods of non-alerts with daily optical gas imaging sweeps to confirm emission status. Detection precision was 98% and false positive and negative rates were 3%. Quantification of challenge-controlled release tests at active oil and gas sites yielded a measured versus true emissions curve with slope = 1.2, R² = 0.90. Repeatability test measurements of four production facilities with two different laser systems showed 33.9% average quantification agreement. Separate third-party blind controlled release testing at two state-of-the-art test facilities yielded 100% true positive rate (0 false negatives). Combining the third-party blind tests with field tests, emission rate quantification uncertainty was +/−41% across five orders of magnitude. These varied evaluation approaches validate the measurement system and operator integration of data for measurement and monitoring of upstream oil and gas emissions and demonstrate a test regime for vetting of monitoring and measurement technologies in active oil and gas operations.

Weiming An, Jilong Tong, Lei Zhang et al.

Oil depots are continuous sources of volatile organic compounds (VOCs), which contribute to ground-level ozone (O₃) and secondary organic aerosol formation, posing threats to air quality and public health. This study investigated typical crude and refined oil depots in the Xigu District of Lanzhou by measuring VOC source profiles and establishing an emission inventory. The maximum incremental reactivity (MIR) method was applied to assess the chemical reactivity of VOCs; both the emission inventory and VOC profiles were incorporated into the WRF-CMAQ model for numerical simulations. Results showed that the average ambient VOC concentrations were 49.8 μg/m³ for the crude oil depot and 66.1 μg/m³ for the refined oil depot. The crude oil depot was dominated by alkanes (37.1%), aromatics (25.1%), and OVOCs (22.5%), while the refined oil depot was dominated by alkanes (57.3%) and OVOCs (16.7%), with isopentane identified as the most abundant species in both depots. The ozone formation potentials (OFPs) of the crude oil and refined oil depots were 153.1 μg/m³ and 178.3 μg/m³, respectively. Aromatics (47.0%) and OVOCs (29.0%) were the primary contributors at the crude oil depot, with isopentane, o-xylene, etc., as the dominant reactive species. In the refined oil depot, the main contributors were alkanes (27.8%), alkenes and alkynes (26.6%), OVOCs (24.5%), and aromatics (20.5%), among which isopentane, trans-2-butene, etc., were most prominent. In 2023, VOC emissions from the crude oil and refined oil depots were estimated at 1605.3 t and 1287.8 t, respectively, mainly from working loss (96.6%) in the crude oil depot and deck fitting loss (60.7%) and working loss (31.3%) in the refined oil depot. Numerical simulations indicated that oil depot emissions could increase regional MDA8 O₃ concentrations by up to 40.0 μg/m³. At the nearby Lanlian Hotel site, emissions contributed 15.1% of the MDA8 O₃, equivalent to a 6.1 μg/m³ increase, while the citywide average was 1.7 μg/m³. This study enriches the VOC source profile database for oil depots, reveals their significant role in regional O₃ formation, and provides a scientific basis for precise O₃ control and differentiated emission reduction strategies in Northwest China.

Saori Miyake, Jonathan Rispler, Sven Teske

Australia is positioning itself to become a ‘renewable energy superpower’ and achieve net-zero emissions by 2050. A GIS-based spatial analysis was conducted to assess the country’s renewable energy potential relative to projected electricity demand in 2050. The results highlight that Australia is exceptionally well-placed to lead the global renewable energy transition. Over 5.1 million km ^2 of land was identified as potential for solar energy development, and 4.8 million km ^2 for onshore wind energy- capable of generating electricity 256 and 132 times greater, respectively, than the projected 2050 demand. This suggests that utilising only 0.4% of the solar potential areas or 0.8% of the onshore wind potential area could meet the country’s electricity demand in 2050. Additionally, 347,578 km ^2 of offshore wind potential area (at water depths ≤50 m) was identified, with the capacity to generate electricity 11 times greater than the projected 2050 demand. Beyond energy generation, renewable energy development could deliver substantial benefits for remote and regional communities in Australia, including enhanced energy security, reliability, independence, and socio-economic development. However, challenges remain, particularly community concern and oppositions related to land-use competition from large-scale renewable energy projects and associated infrastructure in regional areas. Successful energy transition can be enabled through a combination of approaches: developing and promoting effective planning and community engagement processes, adopting emerging technologies to reduce competition for land and the potential socio-economic and environmental impacts, and leveraging existing support for renewable energy. In this context, the finer resolution of spatial analysis and mapping considering local contexts could also play a significant role in initiating conversations with local communities, supporting the engagement process, enabling local input, and guiding informed decision-making in the energy transition of regional areas.

Anuvab Sen, Udayon Sen, Mayukhi Paul et al.

Accurate weather forecasting holds significant importance, serving as a crucial tool for decision-making in various industrial sectors. The limitations of statistical models, assuming independence among data points, highlight the need for advanced methodologies. The correlation between meteorological variables necessitate models capable of capturing complex dependencies. This research highlights the practical efficacy of employing advanced machine learning techniques proposing GenHybQLSTM and BO-QEnsemble architecture based on adaptive weight adjustment strategy. Through comprehensive hyper-parameter optimization using hybrid quantum genetic particle swarm optimisation algorithm and Bayesian Optimization, our model demonstrates a substantial improvement in the accuracy and reliability of meteorological predictions through the assessment of performance metrics such as MSE (Mean Squared Error) and MAPE (Mean Absolute Percentage Prediction Error). The paper highlights the importance of optimized ensemble techniques to improve the performance the given weather forecasting task.

Reza Nematirad, Anil Pahwa, Balasubramaniam Natarajan

Time series forecasting is an important application in various domains such as energy management, traffic planning, financial markets, meteorology, and medicine. However, real-time series data often present intricate temporal variability and sharp fluctuations, which pose significant challenges for time series forecasting. Previous models that rely on 1D time series representations usually struggle with complex temporal variations. To address the limitations of 1D time series, this study introduces the Times2D method that transforms the 1D time series into 2D space. Times2D consists of three main parts: first, a Periodic Decomposition Block (PDB) that captures temporal variations within a period and between the same periods by converting the time series into a 2D tensor in the frequency domain. Second, the First and Second Derivative Heatmaps (FSDH) capture sharp changes and turning points, respectively. Finally, an Aggregation Forecasting Block (AFB) integrates the output tensors from PDB and FSDH for accurate forecasting. This 2D transformation enables the utilization of 2D convolutional operations to effectively capture long and short characteristics of the time series. Comprehensive experimental results across large-scale data in the literature demonstrate that the proposed Times2D model achieves state-of-the-art performance in both short-term and long-term forecasting. The code is available in this repository: https://github.com/Tims2D/Times2D.

Xuanyi Zhao, Jiawen Ding, Xueting Huang et al.

With the rapid development of electricity markets, price volatility has significantly increased, making accurate forecasting crucial for power system operations and market decisions. Traditional linear models cannot capture the complex nonlinear characteristics of electricity pricing, necessitating advanced machine learning approaches. This study compares eight machine learning models using Spanish electricity market data, integrating consumption, generation, and meteorological variables. The models evaluated include linear regression, ridge regression, decision tree, KNN, random forest, gradient boosting, SVR, and XGBoost. Results show that KNN achieves the best performance with R^2 of 0.865, MAE of 3.556, and RMSE of 5.240. To enhance interpretability, LIME analysis reveals that meteorological factors and supply-demand indicators significantly influence price fluctuations through nonlinear relationships. This work demonstrates the effectiveness of machine learning models in electricity price forecasting while improving decision transparency through interpretability analysis.

David Alumbaugh, Erika Gasperikova, Dustin Crandall et al.

Abstract We present a synthetic multi‐scale, multi‐physics dataset constructed from the Kimberlina 1.2 CO2 reservoir model based on a potential CO2 storage site in the Southern San Joaquin Basin of California. Among 300 models, one selected reservoir‐simulation scenario produces hydrologic‐state models at the onset and after 20 years of CO2 injection. Subsequently, these models were transformed into geophysical properties, including P‐ and S‐wave seismic velocities, saturated density where the saturating fluid can be a combination of brine and supercritical CO2, and electrical resistivity using established empirical petrophysical relationships. From these 3D distributions of geophysical properties, we have generated synthetic time‐lapse seismic, gravity and electromagnetic responses with acquisition geometries that mimic realistic monitoring surveys and are achievable in actual field situations. We have also created a series of synthetic well logs of CO2 saturation, acoustic velocity, density and induction resistivity in the injection well and three monitoring wells. These were constructed by combining the low‐frequency trend of the geophysical models with the high‐frequency variations of actual well logs collected at the potential storage site. In addition, to better calibrate our datasets, measurements of permeability and pore connectivity have been made on cores of Vedder Sandstone, which forms the primary reservoir unit. These measurements provide the range of scales in the otherwise synthetic dataset to be as close to a real‐world situation as possible. This dataset consisting of the reservoir models, geophysical models, simulated time‐lapse geophysical responses and well logs forms a multi‐scale, multi‐physics testbed for designing and testing geophysical CO2 monitoring systems as well as for imaging and characterization algorithms. The suite of numerical models and data have been made publicly available for downloading on the National Energy Technology Laboratory's (NETL) Energy Data Exchange (EDX) website.

V. Jayachandran, T. Narayana Rao

India is experiencing a rapid urban growth in recent decades modifying the regional air quality around urban agglomerations. Hyderabad, the capital city of Telangana state in India, has been experiencing significant urbanization of about 17 % growth in urban agglomeration over the past two decades. We investigated the long-term pollution characteristics along with the meteorology in and around Hyderabad (300 km × 300 km) using satellite-based remote sensing, and reanalysis data. Columnar aerosol loading was highest during the Spring while the positive trend was more during the Winter. The northeastern and southeastern parts of the study domain experienced higher aerosol loading. A significant increasing linear trend in AOD and PM2.5 is observed over the urban region as well as the northern and eastern parts. The NO2 and SO2 columnar concentrations showed considerable enhancement over the northeast sub-region where numerous thermal power plants are located, and over the urban centre. The SO2 concentration and SSA values were higher during the Autumn, while the NO2 values peaked along with lower SSA values during the Spring. The observed spatio-temporal features in air pollutants are further investigated using rainfall information, transport pathways, vegetation index, and fire events. Higher surface temperature and the polluted northeasterlies caused the comparative enhancement of NO2 concentration during Spring. The investigation on the NDVI and the fire events in different sub-regions points to the possibility of enhanced human settlement, and thereby the associated anthropogenic activities are notable over the West and South parts of Hyderabad. However, the presence of thermal power plants in the northeast and natural gas plants along the coast act as persistent regional sources for aerosols and pollutant gases irrespective of the wet removal.

Wen Lai, Xiumian Hu, Xiaolong Dong et al.

Abstract As a hot topic in Earth sciences, the Qinghai‐Tibet Plateau has accumulated a large amount of sedimentary‐related data. We constructed a dataset of detrital components for Qinghai‐Tibet Plateau from 63 peer‐reviewed publications. The dataset thus comprises 1813 Late Proterozoic to Pleistocene sandstones from 84 stratigraphic units. For each sample, we present details on reference, detrital composition, GPS, geographic location, depositional age, tectonic setting and depositional environment. It becomes a high‐quality dataset after the information on each sandstone sample was standardized and reviewed by sedimentary experts. The dataset can be used for regional geoscience studies, exploring the general laws of the source‐to‐sink process. The dataset may also be useful in the field of utilities, such as assisting in finding suitable building stones, helping oil and gas and mineral exploration, and so forth.

Patrick Martineau, Hisashi Nakamura, Yu Kosaka et al.

Abstract Temperature variability has substantial socioeconomic impacts through its association with the frequency and severity of heat extremes. Under anthropogenic influence, climate models project seasonally-dependent amplifications of near-surface temperature variability over some sectors of the Southern Hemisphere, and robust positive trends have already been observed in recent decades. Here we show that the amplification of subweekly temperature variability simulated by the multi-model ensemble mean of the sixth phase of the Coupled Model Intercomparison Project (CMIP6) over South Africa, Australia, and South America is often substantially smaller than in reanalyses in recent decades, reaching a similar amplification only at the end of the 21st century due to a weaker amplification of subweekly variance generation efficiency. Analysis of a large model ensemble indicates that this discrepancy may be due to internal climatic variability suggesting that the recent rapid amplification seen in reanalyses may slow down or even temporarily reverse in the near future.

Junzhe Yin, Cristian Meo, Ankush Roy et al.

Nowcasting leverages real-time atmospheric conditions to forecast weather over short periods. State-of-the-art models, including PySTEPS, encounter difficulties in accurately forecasting extreme weather events because of their unpredictable distribution patterns. In this study, we design a physics-informed neural network to perform precipitation nowcasting using the precipitation and meteorological data from the Royal Netherlands Meteorological Institute (KNMI). This model draws inspiration from the novel Physics-Informed Discriminator GAN (PID-GAN) formulation, directly integrating physics-based supervision within the adversarial learning framework. The proposed model adopts a GAN structure, featuring a Vector Quantization Generative Adversarial Network (VQ-GAN) and a Transformer as the generator, with a temporal discriminator serving as the discriminator. Our findings demonstrate that the PID-GAN model outperforms numerical and SOTA deep generative models in terms of precipitation nowcasting downstream metrics.

Xiaoxu Tian, Daniel Holdaway, Daryl Kleist

The use of machine learning (ML) models in meteorology has attracted significant attention for their potential to improve weather forecasting efficiency and accuracy. GraphCast and NeuralGCM, two promising ML-based weather models, are at the forefront of this innovation. However, their suitability for data assimilation (DA) systems, particularly for four-dimensional variational (4DVar) DA, remains under-explored. This study evaluates the tangent linear (TL) and adjoint (AD) models of both GraphCast and NeuralGCM to assess their viability for integration into a DA framework. We compare the TL/AD results of GraphCast and NeuralGCM with those of the Model for Prediction Across Scales - Atmosphere (MPAS-A), a well-established numerical weather prediction (NWP) model. The comparison focuses on the physical consistency and reliability of TL/AD responses to perturbations. While the adjoint results of both GraphCast and NeuralGCM show some similarity to those of MPAS-A, they also exhibit unphysical noise at various vertical levels, raising concerns about their robustness for operational DA systems. The implications of this study extend beyond 4DVar applications. Unphysical behavior and noise in ML-derived TL/AD models could lead to inaccurate error covariances and unreliable ensemble forecasts, potentially degrading the overall performance of ensemble-based DA systems, as well. Addressing these challenges is critical to ensuring that ML models, such as GraphCast and NeuralGCM, can be effectively integrated into operational DA systems, paving the way for more accurate and efficient weather predictions.

Finn Felix Duill, Florian Schulz, Aman Jain et al.

This study focuses on the effect of portable and large filter-based air cleaners (HEPA filters), which became popular indoors during the COVID-19 pandemic, and their suitability for classrooms (here 186 m³). The decay rates of the particle number concentration (PNC) were measured simultaneously at up to four positions in the room. It was found that the different air outlet configurations of the units have an effect on the actual PNC removal in the room when operated at the same volume flow rates. This effect of the airflow efficiency of the air cleaners (AP) in a classroom is quantified with an introduced Air Cleaning Efficiency Factor in this study to identify beneficial airflows. In this context, the effect of filter loading in long-term operation on the cleaning effect is also investigated. The emitted sound pressure levels of the APs are given special attention as this is a critical factor for use in schools, as well as power consumption. A total of six different devices were tested—two portable APs and four large APs. In order to achieve the necessary volume flow rates, three or four of the portable units were used simultaneously in one room, while only one of the large units was used per room. When used at the same air circulation rates in the room, the portable APs exhibit higher sound pressure levels compared to the large APs. At air circulation rates of 4–5 h⁻¹, the portable APs exceeded a value of 45 dB(A). Two of the four large units reach sound pressure levels below 40 dB(A) at air circulation rates of 4–5 h⁻¹, whereby both large units, which are positioned on the rear wall, realize a homogeneous dilution of the room air. This is achieved by an air outlet directed horizontally at a height above 2 m or diagonally towards the ceiling, which points into the room and partly to the sides. On the other hand, an air outlet directed exclusively to the sides or horizontally into the room at floor level to all sides achieves lower particle decay rates. To investigate the influence of the filter loading, three large APs were operated in a school for a period of one year (190 days with 8 h each). For the three APs, long-term operation leads to different changes in PNC reduction efficiency, ranging from −3% to −34%. It is found that not only the size of the prefilter and main filter has a significant influence, but also whether there is a prefilter bypass that negatively affects the loading level of the main filter. At the same time, it was shown that one type of AP, measuring the pressure drop across the filters and readjusting the fan, kept the circulation rate almost constant (up to −3%) over a year.

Jin Ding, Guoping Zhang, Shudong Wang et al.

Based on the Quick Access Recorder (QAR) data covering over 9000 routes in China, the monthly and intra-day distribution characteristics of aircraft bumpiness at different levels were analyzed, and the relationships between the eddy dissipation rate (EDR) and other aircraft flight status elements during bumpiness occurrence were also analyzed. Afterward, aircraft bumpiness routes were constructed using 19 machine learning models. The analyses show that (1) aircraft bumpiness was mainly concentrated between 0:00 a.m. and 17:00 p.m. Severe aircraft bumpiness occurred more frequently in the early morning in January, especially between 5:00 a.m. and 6:00 a.m., and moderate bumpiness always occurred from 3:00 a.m. to 11:00 a.m. (2) The relationship between the left and right attack angles and aircraft bumpiness on the routes was more symmetrical, with a center at 0 degrees, unlike in the approach area where the hotspots were mainly concentrated in the range of −5 to 0 degrees. In the approach area, the larger the Mach number, the more severe the bumpiness. (3) The performances of the Automatic Relevance Determination Regression (ARD), Partial Least Squares Regression (PLS), Elastic-Net Regression (ENR), Classification and Regression Tree (CART), Passive Aggressive Regression (PAR), Random Forest (RF), Stochastic Gradient Descent Regression (SGD), and Tweedie Regression (TWD) based models were relatively good, while the performances of the Huber Regression (HUB), Least Angle Regression (LAR), Polynomial Regression (PLN), and Ridge Regressor (RR) based models were very poor. The aircraft bumpiness prediction models performed best over the approach area of ZBDT (airport in Datong), ZULS (airport in Lhasa), ZPPP (airport in Kunming), and ZLQY (airport in Qingyang). The model performed best in predicting the ZLLL-ZBDT air route (flight routes for Lanzhou to Datong) with different prediction times.

Christoph Sauter, Hayley J. Fowler, Seth Westra et al.

The potential compounding behaviour of heatwaves and extreme rainfall have important implications for a range of hazards, including wildfires and flooding, yet remain poorly understood. In this global study, we analyse the likelihood of extreme 1-hr rainfall immediately following a heatwave, and identify climate zones where this phenomenon is most pronounced. We find the strongest compounding heatwave-extreme rainfall relationships in central Europe and Japan, where the likelihood of extreme rainfall after a heatwave is increased by approximately four times compared to climatology. Significant compounding is found mainly in temperate or colder climates, provided these areas receive ample moisture. As both heatwaves and extreme rainfall are expected to become more frequent in the future, our results indicate that the potential impacts from compounding heatwave-extreme rainfall events might significantly increase as well.

Ariyanto Puji, Anfasa Imam Tanthawi, Sinambela Marzuki et al.

The island of Borneo has relatively low seismic activity. However, the plan to relocate the capital city to the East Borneo region could potentially increase the population, making the area more vulnerable to earthquake occurrences. Therefore, this research aims to determine the depth of the Mohorovičić discontinuity layer, which will provide insights into the thickness of the Earth's crust, and model the local P and S wave velocities using the Inversion, Migration, and Stacking H-k methods. The data used consists of earthquake events with magnitudes greater than 6, located within a distance of 30° to 90° from 6 BMKG stations around the new capital of Indonesia. The research results indicate that the depth of the Mohorovičić discontinuity layer varies between 28 and 43 km. The model of P-wave velocities varies between 1.8 km/s to 9.1 km/s, while the model of S-wave velocities ranges from 1.0 km/s to 5.1 km/s.

Rushil Vohra, Ali Rajaei, Jochen L. Cremer

With the increasing penetration of renewable power sources such as wind and solar, accurate short-term, nowcasting renewable power prediction is becoming increasingly important. This paper investigates the multi-modal (MM) learning and end-to-end (E2E) learning for nowcasting renewable power as an intermediate to energy management systems. MM combines features from all-sky imagery and meteorological sensor data as two modalities to predict renewable power generation that otherwise could not be combined effectively. The combined, predicted values are then input to a differentiable optimal power flow (OPF) formulation simulating the energy management. For the first time, MM is combined with E2E training of the model that minimises the expected total system cost. The case study tests the proposed methodology on the real sky and meteorological data from the Netherlands. In our study, the proposed MM-E2E model reduced system cost by 30% compared to uni-modal baselines.

Minghao He, Shaobo Zhang, Xianyu Yang et al.

Cloud microphysical processes significantly impact the time variation and intensity of precipitation. However, due to the high altitude of the Tibetan Plateau (TP) and the lack of observational data, the understanding of cloud microphysical processes on the TP is relatively insufficient, affecting the accuracy of precipitation simulations around the TP. To further reveal the characteristics of convective precipitation and cloud microphysical structure over the TP, the mesoscale numerical model, WRF, and various observational data were used to simulate and evaluate typical convective precipitation in the Yushu area, which was recorded from 11 to 12 August 2020. The results showed that the combination of the Lin scheme in the WRF model could effectively reproduce this case’s characteristics and evolution process. In the simulation process, the particles of each phase were distributed at different altitudes, and their mass and density over time reflected the characteristics of surface precipitation changes. Among the particles mentioned above, rainwater contributed the most to the initiation and growth of graupel particles. Further research established that the initiation of graupel was mainly affected by the freezing effect of rainwater and cloud ice, while the growth of graupel was influenced primarily by the collision of graupel particles and rainwater. On the whole, from the evolution characteristics of microphysical processes over time, it was found that the ice phase process plays an essential role in this typical convective precipitation.

Ziying Hu, Guangpeng Wang, Yong Liu et al.

Wind regime is one of the main natural factors controlling the evolution and distribution of aeolian sand landforms, and sand drift potential (DP) is usually used to study the capacity of aeolian sand transport. The Badain Jaran Desert (BJD) is located where polar cold air frequently enters China. Based on wind data of eight nearby meteorological stations, this research is intended to explore the temporal variation and spatial distribution features of wind speed and DP using linear regression and cumulative anomaly method, and reveal the relationship between atmospheric circulation and wind speed with correlation analysis. We found that the wind speed and frequency of sand-blowing wind in the BJD decreased significantly during 1971–2016, and the wind speed obviously mutated in 1987. The regional wind speed change was affected by the Asian polar vortex, the northern hemisphere polar vortex and the Tibet Plateau circulation. The wind rose of the annual sand-blowing wind in this region was the “acute bimodal” type. Most of the annual wind directions clustered into the W-NW, and the prevailing wind direction was WNW. During 1971–2016, the annual DP, the resultant drift potential (RDP) and the directional variability (PDP/DP) in the desert showed an obvious downtrend, with a “cliff-like” decline in the 1980s and relative stable fluctuation thereafter. The BJD was under a low-energy wind environment with the acute bimodal wind regime. Wind speed, sand-blowing wind frequency and DP were high in the northeast and low in the southwest.