CW3E Publication Notice: Enhancing Regional Climate Downscaling Through Advances in Machine Learning

CW3E Publication Notice

Enhancing Regional Climate Downscaling Through Advances in Machine Learning

April 4, 2024

CW3E postdoctoral researcher Jorge Baño-Medina recently published a co-authored article titled “Enhancing regional climate downscaling through advances in machine learning” in Artificial Intelligence for the Earth Systems. This review article was led by researchers from the National Institute of Water and Atmospheric Research (New Zealand), along with researchers from the Santander Meteorology Group (Spain), the Climate Change Research Centre and ARC Centre of Excellence for Climate Extremes (Australia), the Faculty of Geosciences and Environment (Switzerland), the National Center for Atmospheric Research (USA), and the Fraunhofer ITWM (Germany). The work aligns with CW3E’s 2019-2024 Strategic Plan to support Monitoring and Projections of Climate Variability and Change using Emerging Technologies.

Despite the sophistication of Global Climate Models (GCMs), their coarse spatial resolution limits their ability to resolve important aspects of climate variability and change at the local scale. Statistical downscaling aims to bridge the gap between the global and the regional-to-local scales by learning an empirical function using large records of data. In the last few years, deep learning (DL) topologies have emerged as potential statistical downscaling tools, due to their ability to learn complex patterns from data. This study summarizes the existing literature about statistical downscaling of climate change scenarios by means of DL, offering a thorough exploration of the current applications, and training strategies that can circumvent existing limitations of statistical downscaling models (e.g., generalization to out-of-distribution spaces or extrapolation to climate change scenarios).

Figure 1. below illustrates the structure of the manuscript and the topics covered in the review. First, we provide an overview of empirical downscaling and we introduce computer-vision based climate downscaling algorithms. Then, we discuss recent machine learning (ML) advances in observational downscaling and Regional Climate Model (RCM) emulation, together with strategies to enhance the out-of-distribution performance of the models. Finally, we outline various future research directions and present an evaluation framework for ML-based empirical downscaling algorithms.

Figure 1: Adapted from Figure 2 in Rampal et al. (2024) An overview of the topics concerning climate downscaling that are discussed in this review article.

Rampal, N., Hobeichi, S., Gibson, P. B., Baño-Medina, J., Abramowitz, G., Beucler, T., … & Gutiérrez, J. M. (2024). Enhancing Regional Climate Downscaling Through Advances in Machine Learning. Artificial Intelligence for the Earth Systems. https://doi.org/10.1175/AIES-D-23-0066.1

CW3E Subseasonal Outlook: 22 March 2024

CW3E Subseasonal Outlook: 22 March 2024

March 22, 2024

Click here for a pdf of this information.


 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Summary provided by C. Castellano, J. Wang, Z. Yang, M. DeFlorio, and J. Kalansky; 22 March 2024

To sign up for email alerts when CW3E post new AR updates click here.

*Outlook products are considered experimental

CW3E Hydrologists Ming Pan and Taylor Dixon Participate in an Exchange with DWR

CW3E Hydrologists Ming Pan and Taylor Dixon Participate in an Exchange with DWR

March 19, 2024

On Tuesday, March 5th, DWR’s Hydrology Section hosted CW3E employees, Ming Pan and Taylor Dixon, to get better acquainted with each other’s hydrologic forecasting work. Even though the two teams have already developed a strong partnership in improving hydrologic runoff models together, the in-person meeting proved to be extremely beneficial.

DWR’s staff shared with CW3E how the legacy system of forecasting water supply runoff volumes for the Bulletin 120 have been produced. This included the technical details of the various multi-linear equations and other methods that have historically been used to forecast April through July runoff volumes for nearly two dozen watersheds in California. DWR further discussed how the 80 percent confidence range is calculated and what innovations the DWR team has produced over the past few years to help improve the median forecasts and the confidence range. DWR staff highlighted which innovations have come from their work with CW3E which includes advancements in the West-WRF forecasts, S2S forecasts, and CW3E’s development of WRF-Hydro models. Through rich back and forth discussions, the CW3E representatives learned more about the long-standing but evolving statistical forecast procedures and gained a deep appreciation for the institutional dedication to and value of the DWR forecasting platform. The CW3E representatives asked many questions and helped identify potential future collaboration opportunities between the two groups.

The discussion later was focused on the WRF-Hydro model where CW3E gave an update on the model products produced the same week and a deeper dive into the modeling framework. This included discussions on fundamental concepts such as model pre-processing, spatial/temporal resolution on the hydrometeorological data used, modeling calibration methods, and post processing algorithms. The discussions were also expanded to include how to leverage the benefit of both physical (process-based) and statistical (data-driven) modeling, as well as the rich collection of observation records, in the context of Machine Learning and how that may apply to problems like data assimilation/fusion and impact of climate change and landcover change (e.g. wildfire). The teams went through the WRF-Hydro “sandbox” established by CW3E on DWR’s Linux workstation, including the JupyterHub/JupyterLab platform, WRF-Hydro model, and hydrometeorological forcing data in order to help DWR implement innovations made by the CW3E team into water supply forecasts. The meeting was extremely valuable, it brought the two teams’ relationship closer together, and both teams look forward to the next opportunity.

CW3E Subseasonal Outlook: 15 March 2024

CW3E Subseasonal Outlook: 15 March 2024

March 15, 2024

Click here for a pdf of this information.


 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Summary provided by Z. Yang, C. Castellano, J. Wang, M. DeFlorio, and J. Kalansky; 15 March 2024

To sign up for email alerts when CW3E post new AR updates click here.

*Outlook products are considered experimental

CW3E Publication Notice: Synthetic forecast ensembles for evaluating Forecast Informed Reservoir Operations

CW3E Publication Notice

Synthetic forecast ensembles for evaluating Forecast Informed Reservoir Operations

March 11, 2024

A new article titled “Synthetic forecast ensembles for evaluating Forecast Informed Reservoir Operations” by Zach Brodeur (Cornell University), Chris Delaney (Sonoma Water), Brett Whitin (NOAA/NWS), and Scott Steinschneider (Cornell University) was recently published in the American Geophysical Union’s Water Resources Research. The work supports an ongoing, combined effort between Cornell University, CW3E, NOAA/NWS, USACE, and local water agencies to integrate synthetic forecasts into the development of robust and resilient Forecast Informed Reservoir Operations (FIRO). Prior CW3E-led FIRO viability assessments have identified the development of synthetic forecasts as a key effort to enable the thorough design and testing of FIRO policies, ensuring they are safe and effective across the widest possible range of events.

In the featured work, the authors develop a statistical methodology to generate synthetic ensemble forecasts that emulate the skill and behavior of ensemble forecasts from the NOAA/NWS Hydrologic Ensemble Forecasting System (HEFS). The synthetic forecast approach models the ensemble forecast uncertainty as a function of streamflow observations, enabling the generation of new sets of plausible ensemble forecast sequences for any series of streamflow observations in the historical record (see Figure 1). Since HEFS hindcasts are limited to the period from the early 1980s to present (because of limitations to the satellite record), this capability has the potential to expand the availability of forecast data to the 1950s or earlier in many locations. Moreover, the technique enables generation of plausible ensemble forecasts for any streamflow sequence (e.g. stochastic streamflow models, scaled design events, climate change simulations, etc.), allowing performance assessments of FIRO strategies under a wide range of potential scenarios. The methodology is demonstrated at Lake Mendocino, CA, the pilot site for the FIRO viability assessment process.

Figure 1: Adapted from Figure S3.1 in Brodeur et al. (2024) for inflows at Lake Mendocino, CA. a) HEFS forecast issued 5 days prior to the February 18, 1986 flood, b-d) 3x synthetic HEFS (sHEFS) ensembles for the same event and lead time.

The synthetic forecasting technique presented in this work overcomes earlier limitations in synthetic forecast methodologies that were too simplified to capture the complexities of forecast uncertainty across multiple lead times, sites, and ensemble members at a daily resolution. The evaluation of the synthetic forecast approach consists of two primary efforts: 1) ensemble forecast verification techniques to ensure parity between the synthetic forecasts (syn-HEFS) and actual forecasts (HEFS); and 2) operational validation of the synthetic forecasts using the Ensemble Forecast Operations (EFO) model of Delaney et al. (2020).

The operational validation results for the hindcast period (1986-present) are highlighted below in Figure 2, with particular emphasis on the February 18, 1986 flood. The synthetic HEFS forecasts emulate operational behavior of the single hindcast sequence of the actual HEFS, while also revealing operational vulnerabilities (i.e. potential emergency spillway usage) to alternative ensemble forecasting sequences during extreme events.

Figure 2: Adapted from Figure 9 in Brodeur et al. (2024). a) Lake Mendocino storage time series for HEFS, perfect forecast operations (PFO), and 100x samples of synthetic forecasts; b) as in (a), but zoomed in to the February 18, 1986 event; c) as in (b) but for the reservoir release time series.

Figure 3 demonstrates the capability to extend synthetic HEFS forecasts to the pre-hindcast period (prior to 1985) and assess operational policy performance outside the available hindcast record. This demonstration shows that the synthetic forecast approach is an effective risk analysis tool and underscores their potential for out-of-sample FIRO policy evaluation.

Figure 3: Adapted from Figure 10 in Brodeur et al. (2024). As in Figure 2, but for the pre-hindcast period (1948-1985) for (a) and zoomed in to the December 23, 1964 event for (b-c). Importantly, there are no available HEFS forecasts to compare in this period.

Overall, the study showcases an extremely important ‘first-of-its-kind’ effort towards the development of a synthetic ensemble forecasting methodology that operates at the requisite resolution needed for operationally relevant FIRO risk analysis. The synthetic forecast work is a continuing collaboration between Cornell University and CW3E and opens the door to a wide variety of potential future studies and exciting potential to integrate with ongoing FIRO activities, including forthcoming viability assessments.

Brodeur, Z. P., Delaney, C. , Whitin, B., & Steinschneider, S. (2024). Synthetic forecast ensembles for evaluating Forecast Informed Reservoir Operations. Water Resources Research, 60(2) e2023WR034898. https://doi.org/10.1029/2023WR034898

Delaney, C. J., Hartman, R. K., Mendoza, J., Dettinger, M., Delle Monache, L., Jasperse, J., et al. (2020). Forecast Informed Reservoir Operations using ensemble streamflow predictions for a multipurpose reservoir in Northern California. Water Resources Research, 56(9). https://doi.org/10.1029/2019WR026604

CW3E Subseasonal Outlook: 8 March 2024

CW3E Subseasonal Outlook: 8 March 2024

March 8, 2024

Click here for a pdf of this information.


 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Summary provided by J. Wang, C. Castellano, Z. Yang, M. DeFlorio, and J. Kalansky; 8 March 2024

To sign up for email alerts when CW3E post new AR updates click here.

*Outlook products are considered experimental

CW3E Event Summary: 28 February – 3 March 2024

CW3E Event Summary: 28 February – 3 March 2024

8 March 2024

Click here for a pdf of this information.

Atmospheric River and Winter Storm Produce Heavy Snow Across Western US

  • An atmospheric river (AR) and a slow-moving mid-level trough fueled a long-duration precipitation event over the Western US during 28 Feb – 3 Mar.

The AR:

  • The AR made landfall over the Pacific Northwest late on 27 Feb, producing AR2 conditions (based on the Ralph et al. 2019 AR Scale) along the coast of southern Washington, Oregon, and far Northern California.
  • After the initial landfalling AR weakened, a second pulse of moisture transport brought another period of AR conditions to Northern and Central California.
  • The initial AR landfall and the second pulse of moisture transport combined to produce AR1–2 conditions in the foothills of the Sierra Nevada and played a key role in supporting very heavy snowfall accumulations.

Impacts:

  • The heaviest precipitation (> 6 inches liquid equivalent) was observed in the Pacific Coast Ranges and Sierra Nevada.
  • An estimated 4–10 feet of snow fell in the Northern and Central Sierra Nevada, with the highest totals near Lake Tahoe.
  • About 2–4 feet of snow fell in the Olympic Mountains, Cascades, Klamath Mountains, and Southern Sierra Nevada.
  • Low freezing levels also facilitated significant snowfall accumulations (> 12 inches) in the Willapa Hills, Oregon Coast Ranges, and Northern California Coast Ranges.
  • This event provided a substantial boost to seasonal snowpack in the Sierra Nevada, with many stations reporting snow water equivalent (SWE) increases of 8–12 inches (~20–30% of the typical peak SWE) over a 5-day period.
  • Heavy snowfall and high winds created extremely dangerous travel conditions, resulting in closures of I-80 and US-395.

Click images to see loops of GFS 500-hPa Geopotential Height/Vorticity and IVT analyses

Valid: 1200 UTC 27 February – 1200 UTC 4 March 2024

GOES West GEOCOLOR Composite: NOAA/NESDIS/STAR
Valid: 0000 UTC 28 February – 0000 UTC 2 March 2024


 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Summary provided by C. Castellano, S. Bartlett, P. Iniguez, J. Kalansky, and G. Lewis; 8 Mar 2024

To sign up for email alerts when CW3E post new AR updates click here.

CW3E Subseasonal Outlook: 1 March 2024

CW3E Subseasonal Outlook: 1 March 2024

March 1, 2024

Click here for a pdf of this information.


 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Summary provided by Z. Yang, C. Castellano, J. Wang, M. DeFlorio, and J. Kalansky; 1 March 2024

To sign up for email alerts when CW3E post new AR updates click here.

*Outlook products are considered experimental