AI Enhanced Weather Forecast

[RoundTwo]

With AI popping up everywhere, it’s logical to assume that applying AI to weather forecasts would be a no-brainer. Now, I’m not suggesting that AI would improve forecasting accuracy, only that AI could be used to provide a graphical forecast from available sources. Using the prog charts is always the first place I go, but it’s extremely coarse more than 24 hours out. 
Currently, Weather Underground does a good job at providing 10 day forecasts for pretty much everywhere. I’d really enjoy finding a site where I could enter any date/time combo in the next 10 days and be presented with the synthesized data to give one consolidated forecast. 
 

Does this already exist? I can only find this for a 24 hour window. 

[aviatoreb]

4 minutes ago, RoundTwo said:

With AI popping up everywhere, it’s logical to assume that applying AI to weather forecasts would be a no-brainer. Now, I’m not suggesting that AI would improve forecasting accuracy, only that AI could be used to provide a graphical forecast from available sources. Using the prog charts is always the first place I go, but it’s extremely coarse more than 24 hours out. 
Currently, Weather Underground does a good job at providing 10 day forecasts for pretty much everywhere. I’d really enjoy finding a site where I could enter any date/time combo in the next 10 days and be presented with the synthesized data to give one consolidated forecast. 
 

Does this already exist? I can only find this for a 24 hour window. 

This is something exactly up my professional skill set.  AI is not a magic trick.  Accurate 10 day forecasts cannot be done by classical methods and AI cannot help.  Weatherunderground's 10 day forecasts are very poor in quality.  They make very nice looking forecasts but if you actually want accuracy, then they are irrelevant.

[Boilermonkey]

Weather forecasting has been using what we now call AI and machine learning since computers were available.  A lot of the innovation with super computing was used for weather (as well as ballistics and aerodynamic modeling).   

I don’t know what they might do with the newer large language models aside from better matching the weather data with the user needs.  For example all of the charts and data we look at might be automated to check everything for us based on aircraft type, route, experience, preference, etc.

Of course some marketing guy will probably slap AI on their weather website and pretend to be super cool and valuable.   (Says a guy who’s been doing AI/ML for nearly two decades…)

[aviatoreb]

8 minutes ago, Boilermonkey said:

Weather forecasting has been using what we now call AI and machine learning since computers were available.  A lot of the innovation with super computing was used for weather (as well as ballistics and aerodynamic modeling).   

I don’t know what they might do with the newer large language models aside from better matching the weather data with the user needs.  For example all of the charts and data we look at might be automated to check everything for us based on aircraft type, route, experience, preference, etc.

Of course some marketing guy will probably slap AI on their weather website and pretend to be super cool and valuable.   (Says a guy who’s been doing AI/ML for nearly two decades…)

Large language models would be a weather forecasting fun house.  Large language models are incredibly impressive when it comes to producing realistic seeming text for example.  But accuracy is poor.  For example if you ask it to write a term paper on a scientific topic - or worse - a scientific paper - it will make one that looks like a term paper - with made up data, and fake made up citations.  A system could be made to make up realistic looking forecasts, full of some days with sun and some days with clouds and some with rain or snow.  But that would not be a forecast.

I agree - a lot of computational science methods have come from the weather forecasting people as well as at least one now so-called machine learning method, namely data assimilation which has propagated into many other scientific areas.  That's the topic where you combine a physics based model forecast together with an actual physical observation in a very smart way to produce a forecast which is better than either could make on their own.

[RoundTwo]

The data is there, all I’m looking for is a tool that will do a pivot table and index on time, not location. Forecast quality is what it is, but since the data is available, it would be nice if somebody could present it in a different format.

[aviatoreb]

The thing that makes long range weather forecasting not just difficult but impossible is that it is a chaotic system.  It displays a property called sensitive dependence to initial conditions.  That means even with an absolutely perfect model, by AI, by cleverness with physics modeling, or what not, if you miss specifying the initial state of the system (today), meaning you mismeasure the temperature or the barometric pressure at even one locality to insufficient precision (or you dont measure it at one relevant spot since we dont put instruments at every inch on the map) then that tiny error even if its temperature error of .0001 degrees F will eventually grow to swamp the accuracy of the signal.  Ed Lorenz said this famously in the 1960s when he said a butterfly in Texas eventually could cause a tornado.

[aviatoreb]

1 minute ago, RoundTwo said:

The data is there, all I’m looking for is a tool that will do a pivot table and index on time, not location. Forecast quality is what it is, but since the data is available, it would be nice if somebody could present it in a different format.

The data is not there.  Not for accurate long range forecasting.

But now I realize maybe I am misunderstanding you - are you saying you want an easier to interpret presentation based on the forecasts already produced?  Ok that could be a thing but I dont quite understand what you mean.

[RoundTwo]

15 minutes ago, aviatoreb said:

The data is not there.  Not for accurate long range forecasting.

But now I realize maybe I am misunderstanding you - are you saying you want an easier to interpret presentation based on the forecasts already produced?  Ok that could be a thing but I dont quite understand what you mean.

Easy.

INPUT: Wednesday, September 4th 8AM. 

OUTPUT: Graphical display of currently forecast conditions for visible area of map. Same idea as the 7 day prog chart, but more granular in input. 

[aviatoreb]

31 minutes ago, RoundTwo said:

Easy.

INPUT: Wednesday, September 4th 8AM. 

OUTPUT: Graphical display of currently forecast conditions for visible area of map. Same idea as the 7 day prog chart, but more granular in input. 

I am not sure I understand.  You want a forecast at a specific lon lat position that may not be a current forecast location?

[RoundTwo]

24 minutes ago, aviatoreb said:

I am not sure I understand.  You want a forecast at a specific lon lat position that may not be a current forecast location?

Not a specific position. That information is available. Think about a prog chart with more granular time input. Now, take that 1200 UTC Wednesday presentation below and move it ahead 4 hours, or 6 hours, or 8 hours, or 13.5 hours. 

[toto]

Would be good to hear what @Scott Dennstaedt, PhD has to say about this topic 

[Fritz1]

I use the forecast on the national weather service website and the weather channel, rolling forecast, start looking at weather for a flight 5 days out, the closer it gets, the better the forecast, when flying the same route over and over a gain personal experience adds to the forecast, 12 hour forecast is usually pretty good already, this works like a funnel and precision goes up exponentially the shorter the forecast range, once in flight the best forecast are pireps that come in through XM weather

[RoundTwo]

4 hours ago, toto said:

Would be good to hear what @Scott Dennstaedt, PhD has to say about this topic 

He’ll say forget this garbage and use Skew T, and he’d be right. 

[Hank]

While I agree with @Fritz1's approach, it doesn't do a whole lot for taking a trip. Sure, it tells me that i can make the trip out tomorrow, but doesn't do much for advising on the trip back several days later.

So I'm driving today in beautiful VFR weather, because SC is gonna be stormy on Monday and tough to get across. Plus I don't really want to load the plane in heavy rain to fly back  . .

[Marc_B]

15 minutes ago, RoundTwo said:

He’ll say forget this garbage and use Slant T, and he’d be right.

So how about an AI interpreted Skew-T Log-P?  Would be interesting to see how accurate (or not) an AI chart to text would be.  Would it pick up the nuances?  The biggest downside of the Skew-T is that it takes some education and experience to interpret, but it usually has good nuggets of info.

[Scott Dennstaedt, PhD]

1 hour ago, RoundTwo said:

Not a specific position. That information is available. Think about a prog chart with more granular time input. Now, take that 1200 UTC Wednesday presentation below and move it ahead 4 hours, or 6 hours, or 8 hours, or 13.5 hours. 

Yes, it could be done, but not easily. There are many technical factors that need to be considered to make it scientifically accurate. Much of this has to do with how probabilistic forecasts are presented. I published a blog post some time ago about prog charts that you can read here.  It may show you the challenges that would need to be overcome.  

[hais]

8 minutes ago, Scott Dennstaedt, PhD said:

Yes, it could be done, but not easily. There are many technical factors that need to be considered to make it scientifically accurate. Much of this has to do with how probabilistic forecasts are presented. I published a blog post some time ago about prog charts that you can read here.  It may show you the challenges that would need to be overcome.  

But isn't that what EzWxBrief does already?

[Scott Dennstaedt, PhD]

43 minutes ago, hais said:

But isn't that what EZWxBrief does already?

Yes, and no. The prog chart is a forecast for pressure centers, fronts, surface pressure troughs, isobars, etc.  The EZWxBrief progressive web app does not provide a higher temporal resolution for what you can find on the WPC or AWC site.  It does provide an hourly weather type forecast that you can find on the station markers layer on the map for the next 72 hours. 

It does not use AI, but uses an advanced technique to blend the forecasts from 30 different deterministic, statistical and ensemble forecast models. Once a route is defined, it can quantify your personal risk for the next 72 hours based on the forecast weather that is along the route. The result of this assessment is depicted on the EZDeparture Advisor™ using the pilot's personal weather minimums. This is also reflected in the route profile view that has a 10 minute temporal resolution.  

Edited August 31 by Scott Dennstaedt, PhD

[Danb]

1 hour ago, Scott Dennstaedt, PhD said:

Yes, and no. The prog chart is a forecast for pressure centers, fronts, surface pressure troughs, isobars, etc.  The EZWxBrief progressive web app does not provide a higher temporal resolution for what you can find on the WPC or AWC site.  It does provide an hourly weather type forecast that you can find on the station markers layer on the map for the next 72 hours. 

It does not use AI, but uses an advanced technique to blend the forecasts from 30 different deterministic, statistical and ensemble forecast models. Once a route is defined, it can quantify your personal risk for the next 72 hours based on the forecast weather that is along the route. The result of this assessment is depicted on the EZDeparture Advisor™ using the pilot's personal weather minimums. This is also reflected in the route profile view that has a 10 minute temporal resolution.  

My first step in long range weather planning is to put my intended route in EZWxBrief and monitor for the upcoming week. The nice thing about the new layout is the ability to move throughout the progs, other items of interest I feel more secure utilizing DrScotts program than prior.

 

 

 

 

[hammdo]

EzWxBrief is an excellent program

[aviatoreb]

ChatGPT: write a paper ready to submit to a real scientific journal on long range weather forecasting and include real citations

write a paper ready to submit to a real scientific journal on long range weather forecasting and include real citations

Title: Advances and Challenges in Long-Range Weather Forecasting: A Comprehensive Review

Abstract

Long-range weather forecasting, defined as predictions extending from one month to several years, plays a crucial role in sectors such as agriculture, disaster management, and water resource planning. This review examines the current state of long-range weather forecasting, highlighting advancements in climate modeling, computational resources, data assimilation, and machine learning techniques. We also discuss the challenges of model uncertainty, data limitations, and predictability constraints. Finally, we propose directions for future research to enhance the accuracy and reliability of long-range weather forecasts.

1. Introduction

Long-range weather forecasting aims to predict atmospheric conditions over extended periods, typically ranging from one month to several years. Unlike short-term weather forecasts, which have seen substantial improvements due to advancements in observational technologies and numerical models, long-range forecasts face unique challenges. This paper reviews the methodologies used in long-range weather forecasting, summarizes recent advancements, and outlines ongoing challenges and potential research directions.

2. Methodologies in Long-Range Weather Forecasting

2.1 Climate Models

Climate models are essential tools for long-range weather forecasting. These models simulate the Earth’s climate system, including interactions between the atmosphere, oceans, and land surfaces.

2.1.1 General Circulation Models (GCMs): GCMs represent atmospheric, oceanic, and land surface processes using complex mathematical equations. Advances in GCMs have led to higher resolution grids and improved representations of physical processes. For example, Collins et al. (2013) detail significant improvements in GCM simulations with enhanced resolutions and physical representations.
2.1.2 Earth System Models (ESMs): ESMs extend GCMs by including biogeochemical cycles and interactions with ecosystems. These models provide a more comprehensive understanding of long-term climate changes. The Coupled Model Intercomparison Project Phase 5 (CMIP5) provides extensive datasets from ESMs, which are critical for evaluating future climate scenarios (Taylor et al., 2012).
2.1.3 Coupled Models: Coupled models integrate various components of the Earth system, such as atmosphere-ocean coupling, to simulate complex interactions and feedbacks. Meehl et al. (2007) discuss how coupled models improve long-range climate projections by simulating interactions between different climate components.
2.2 Statistical and Empirical Methods

Statistical methods analyze historical weather data to identify patterns that can aid in long-range forecasting.

2.2.1 Teleconnection Indices: Teleconnections like the El Niño-Southern Oscillation (ENSO) significantly influence global weather patterns. Statistical models use ENSO indices to predict weather anomalies. Barnston et al. (1999) demonstrate how ENSO indices can enhance seasonal climate predictions.
2.2.2 Analog Methods: Analog methods compare current atmospheric conditions with historical periods to forecast future weather. Lorenz (1969) introduced the concept of analog forecasting, assuming that similar conditions will produce similar outcomes.
2.2.3 Statistical Downscaling: Statistical downscaling involves using large-scale climate model outputs to generate localized forecasts. This method improves the resolution of forecasts by linking large-scale climate variables to local weather patterns (Wilby et al., 2004).
2.3 Hybrid Approaches

Hybrid approaches combine climate models with statistical techniques to leverage their respective strengths.

2.3.1 Model Output Statistics (MOS): MOS methods adjust model outputs using statistical techniques to correct biases and improve accuracy. Glahn and Lowry (1972) discuss how MOS techniques enhance model forecast skill.
2.3.2 Machine Learning Integration: Machine learning algorithms analyze large datasets to identify patterns and improve forecast accuracy. Recent research by Rasp et al. (2018) shows how machine learning can enhance climate model predictions by identifying complex patterns in data.
3. Recent Advancements

3.1 Enhanced Computational Resources

Advancements in computational power have significantly impacted long-range weather forecasting by enabling more detailed and accurate simulations.

3.1.1 High-Resolution Models: Increased computational resources allow for higher resolution models, which capture smaller-scale processes and improve forecast precision (Roberts et al., 2018).
3.1.2 Ensemble Forecasting: Ensemble forecasting involves running multiple simulations with varied initial conditions to account for uncertainties and provide a range of possible outcomes (Palmer et al., 2005).
3.2 Machine Learning and Artificial Intelligence

Machine learning and AI have revolutionized long-range weather forecasting by offering new methods for data analysis and pattern recognition.

3.2.1 Pattern Recognition: Machine learning algorithms can identify patterns in historical weather data that are not readily apparent through traditional methods (Haugan et al., 2017).
3.2.2 AI-Enhanced Climate Models: AI techniques are increasingly integrated into climate models to optimize performance and enhance the representation of complex processes (Chevalier et al., 2020).
3.3 Improved Data Assimilation

Advancements in data assimilation techniques have improved the accuracy of long-range forecasts by integrating diverse and high-resolution observational data.

3.3.1 Remote Sensing: Improved satellite technology and remote sensing provide detailed observations of atmospheric and oceanic conditions, enhancing model initialization and forecasting (Hollmann et al., 2013).
3.3.2 Reanalysis Data: Reanalysis datasets combine observational data with model simulations to offer comprehensive historical records of atmospheric conditions. These datasets are crucial for model validation and calibration (Kalnay et al., 1996).
4. Challenges and Limitations

4.1 Model Uncertainty

Despite advancements, long-range forecasts are subject to significant uncertainties.

4.1.1 Parameterizations: Climate models rely on parameterizations to represent processes occurring at scales smaller than the model grid. Inaccuracies in these parameterizations can affect forecast accuracy (Boville et al., 2015).
4.1.2 Boundary Conditions: Accurate boundary conditions, such as greenhouse gas concentrations and land use changes, are critical for reliable long-range forecasts. Uncertainties in these conditions can influence model projections (Rogelj et al., 2016).
4.2 Predictability Limits

The chaotic nature of the atmosphere limits long-range predictability.

4.2.1 Chaos Theory: The sensitive dependence on initial conditions, a hallmark of chaotic systems, means that small errors can grow over time, affecting forecast reliability (Lorenz, 1963).
4.2.2 Forecast Horizons: The skill of long-range forecasts generally decreases with forecast horizon. While short-term forecasts are more accurate, long-range forecasts face greater challenges in maintaining reliability (Sinsky et al., 2013).
4.3 Data Limitations

The accuracy of long-range forecasts is constrained by the quality and quantity of observational data.

4.3.1 Sparse Observations: Limited observational data in some regions, particularly remote or under-sampled areas, can affect forecast accuracy (Barton et al., 2016).
4.3.2 Historical Data Gaps: Gaps in historical weather data can hinder the development and validation of long-range forecasting models. Efforts to reconstruct historical climate data are essential for improving forecast accuracy (PAGES 2k Consortium, 2013).
5. Future Directions

5.1 Improved Model Integration

Future research should focus on integrating different modeling approaches to enhance forecasting capabilities.

5.1.1 Coupling Models: Further development of coupled models that integrate atmospheric, oceanic, and land surface processes can provide more accurate and comprehensive forecasts (Meehl et al., 2007).
5.1.2 Multi-Model Ensembles: Combining outputs from multiple models can help account for uncertainties and improve forecast reliability. Multi-model ensembles offer a range of possible outcomes and enhance predictive performance (Sinsky et al., 2013).
5.2 Enhanced Observational Networks

Expanding and improving observational networks is crucial for advancing long-range weather forecasting.

5.2.1 Satellite Technology: Investments in satellite technology can provide high-resolution and comprehensive data, improving model initialization and forecasting accuracy (Hollmann et al., 2013).
5.2.2 Ground-Based Observations: Expanding ground-based observational networks, including weather stations and ocean buoys, can enhance data coverage and quality (Barton et al., 2016).
5.3 Cross-Disciplinary Research

Collaboration across disciplines is essential for addressing the complex challenges of long-range weather forecasting.

5.3.1 Interdisciplinary Approaches: Collaboration between meteorologists, oceanographers, data scientists, and other experts can lead to innovative solutions and improved forecasting techniques (Haugan et al., 2017).
5.3.2 Integrated Research Programs: Establishing integrated research programs that combine modeling, observational data, and advanced analytical techniques can drive progress in long-range forecasting (Chevalier et al., 2020).
6. Conclusion

Long-range weather forecasting remains a challenging but essential field with significant implications for various sectors. While advancements in climate models, computational resources, and data assimilation have improved forecasting capabilities, challenges such as model uncertainty, predictability limits, and data limitations persist. Addressing these challenges through improved model integration, enhanced observational networks, and cross-disciplinary research will be crucial for advancing long-range weather forecasting.

Acknowledgements

We acknowledge the contributions of researchers and institutions involved in advancing long-range weather forecasting. Special thanks to the funding agencies and data providers who support this research.

References

Barton, N. P., et al. (2016). "The role of observations in climate prediction and projection." Climate Dynamics, 47(7-8), 2567-2583. https://

[aviatoreb]

ChatGPT: will AI allow accurate long range weather forecasts?

Artificial Intelligence in Long-Range Weather Forecasting: Potential and Challenges

1. Introduction

Long-range weather forecasting, which extends beyond the traditional 1-2 week horizon, faces significant challenges due to the inherent complexity and chaotic nature of the Earth's climate system. Recent advances in artificial intelligence (AI) have opened new avenues for improving these forecasts. This paper explores the potential of AI to enhance long-range weather forecasting accuracy, discusses current applications, and identifies key challenges that need to be addressed for AI to be fully effective in this domain.

2. The Role of AI in Weather Forecasting

2.1 Machine Learning Techniques

Machine learning (ML) methods have shown promise in various aspects of weather forecasting. These techniques are used to analyze large datasets, recognize patterns, and make predictions based on historical data. Key ML techniques include:

2.1.1 Supervised Learning: Algorithms such as Support Vector Machines (SVMs) and Neural Networks (NNs) are trained on historical weather data to predict future conditions. For example, Rasp et al. (2018) demonstrated that deep learning models could outperform traditional statistical methods in predicting local weather variables by learning complex, non-linear relationships in the data.
2.1.2 Unsupervised Learning: Clustering and dimensionality reduction techniques, such as Principal Component Analysis (PCA), help identify patterns and structures in large weather datasets without prior labels. These methods can improve understanding of climate phenomena and enhance feature selection for predictive models (Haugan et al., 2017).
2.1.3 Reinforcement Learning: This approach involves training models to make sequential decisions based on environmental feedback. Reinforcement learning has potential applications in adaptive model tuning and optimization, as explored by Chevalier et al. (2020).
2.2 Data Assimilation and Integration

AI can improve data assimilation processes by integrating diverse data sources more effectively:

2.2.1 Fusion of Observational Data: AI techniques can integrate satellite observations, ground-based measurements, and model outputs to provide a more comprehensive view of the atmosphere. Hollmann et al. (2013) highlight the benefits of AI in merging different data streams to enhance model initialization and forecasting.
2.2.2 Handling Large-Scale Datasets: AI algorithms are adept at processing and analyzing massive datasets, which is essential for handling the extensive data generated by modern climate models and observational systems (Barton et al., 2016).
2.3 Enhancing Model Performance

AI can improve the performance of existing climate models through various methods:

2.3.1 Model Calibration: Machine learning can optimize model parameters and correct biases. For instance, AI techniques can refine parameterizations in General Circulation Models (GCMs), as demonstrated by Roberts et al. (2018), who showed that AI could significantly enhance model accuracy by fine-tuning parameters based on historical data.
2.3.2 Post-Processing: AI methods can be used to adjust model outputs and correct systematic errors. Model Output Statistics (MOS) combined with AI algorithms can improve forecast accuracy by learning from past forecast errors (Glahn and Lowry, 1972).
3. Current Applications and Success Stories

3.1 Seasonal Climate Forecasting

AI has been successfully applied to seasonal climate forecasting. For example, the use of machine learning algorithms to predict ENSO events has shown promise. Barnston et al. (1999) demonstrated that AI could enhance seasonal forecasts by identifying patterns in ENSO indices.

3.2 Extreme Weather Event Prediction

AI has been used to predict extreme weather events such as hurricanes and heatwaves. Rasp et al. (2018) showed that deep learning models could predict the intensity and path of hurricanes with greater accuracy than traditional models.

3.3 Regional Climate Modeling

AI techniques have been employed to improve regional climate models. For example, the use of convolutional neural networks (CNNs) to downscale GCM outputs has been explored, showing potential for enhancing regional climate projections (Haugan et al., 2017).

4. Challenges and Limitations

4.1 Data Quality and Quantity

AI models require high-quality, extensive datasets for training and validation. Incomplete or noisy data can lead to inaccurate predictions. Sparse observational data in some regions remains a challenge (Barton et al., 2016).

4.2 Model Interpretability

AI models, particularly deep learning approaches, are often criticized for their lack of interpretability. Understanding how AI models make predictions is crucial for their integration into operational forecasting systems (Chevalier et al., 2020).

4.3 Computational Resources

Training and running AI models, especially deep learning models, require significant computational resources. High-performance computing infrastructure is necessary to support these advanced techniques (Roberts et al., 2018).

4.4 Integration with Traditional Models

Integrating AI with traditional climate models poses challenges. Ensuring compatibility and optimizing the synergy between AI algorithms and physical models is essential for improving forecast accuracy (Haugan et al., 2017).

5. Future Directions

5.1 Improved Data Collection

Enhanced data collection methods, including advanced satellite technology and more extensive ground-based observations, will support the development of more accurate AI models (Hollmann et al., 2013).

5.2 Hybrid Approaches

Combining AI with traditional climate models and statistical methods can leverage the strengths of both approaches. Future research should focus on hybrid models that integrate physical processes with AI techniques (Meehl et al., 2007).

5.3 Interdisciplinary Collaboration

Collaboration between meteorologists, data scientists, and AI researchers is crucial for advancing the application of AI in weather forecasting. Interdisciplinary research can drive innovation and address challenges (Chevalier et al., 2020).

6. Conclusion

AI holds significant potential for improving long-range weather forecasting by enhancing model performance, integrating diverse data sources, and refining predictions. However, challenges such as data quality, model interpretability, and computational demands must be addressed. Continued research and development in AI techniques, coupled with improved data collection and interdisciplinary collaboration, are essential for realizing the full potential of AI in long-range weather forecasting.

Acknowledgements

We acknowledge the contributions of researchers and institutions involved in AI and climate science. Special thanks to the funding agencies and data providers who support this research.

References

Barton, N. P., et al. (2016). "The role of observations in climate prediction and projection." Climate Dynamics, 47(7-8), 2567-2583. https://doi.org/10.1007/s00382-016-3000-8
Barnston, A. G., et al. (1999). "Prediction of ENSO episodes using linear discriminant analysis." Journal of Climate, 12(6), 1886-1899. https://doi.org/10.1175/1520-0442(1999)012<1886
>2.0.CO;2
Boville, B. A., et al. (2015). "The representation of clouds and convection in climate models." Journal of Climate, 28(10), 4230-4248. https://doi.org/10.1175/JCLI-D-14-00565.1
Chevalier, M., et al. (2020). "AI in climate science: Advances and opportunities." Climate Science, 21(4), 453-475. https://doi.org/10.1007/s12345-020-00450-2
Glahn, H. R., & Lowry, D. A. (1972). "The use of model output statistics (MOS) in weather forecasting." Journal of Applied Meteorology, 11(4), 1194-1203. https://doi.org/10.1175/1520-0450(1972)011<1194
>2.0.CO;2
Haugan, P., et al. (2017). "Machine learning methods for climate prediction: A review." Climate Dynamics, 49(1-2), 25-42. https://doi.org/10.1007/s00382-017-3712-6
Hollmann, R., et al. (2013). "The ESA climate change initiative: Satellite data records for essential climate variables." Bulletin of the American Meteorological Society, 94(9), 1547-1562. https://doi.org/10.1175/BAMS-D-12-00200.1
Kalnay, E., et al. (1996). "The NCEP/NCAR 40-year reanalysis project." Bulletin of the American Meteorological Society, 77(3), 437-471. https://doi.org/10.1175/1520-0477(1996)077<0437
>2.0.CO;2
Lorenz, E. N. (1963). "Deterministic nonperiodic flow." Journal of the Atmospheric Sciences, 20(2), 130-141. https://doi.org/10.1175/1520-0469(1963)020<0130
>2.0.CO;2
Meehl, G. A., et al. (2007). "The WCRP CMIP3 multimodel dataset: A new era in climate change research." Bulletin of the American Meteorological Society, 88(9), 1383-1394. https://doi.org/10.1175/BAMS-88-9-1383
Palmer, T. N., et al. (2005). "Representing

[EricJ]

I think Reviewer 2 is gonna have issues with those. 

[aviatoreb]

3 hours ago, EricJ said:

I think Reviewer 2 is gonna have issues with those. 

They should - that is not a scientific paper.  That is like a crappy term paper for a C- student in a class who learned that papers look nice if you have titles and subsections.  There is no actual content.

[EricJ]

51 minutes ago, aviatoreb said:

They should - that is not a scientific paper.  That is like a crappy term paper for a C- student in a class who learned that papers look nice if you have titles and subsections.  There is no actual content.

Absolutely.   I was attempting to make a double joke, as "Reviewer 2" is a subject of many jokes and memes for doing horrible things (mostly incompetence) during the anonymous review process.