Look at the skies. Are you enjoying the best air quality of your life? Hydrowonk is.
COVID-19 is a natural experiment of the impact of economic activity on the environment. Will the water industry take advantage of this generational opportunity? If so, how?
It is time to abandon how science is used in the water industry today and embrace science using 21st century technology. By integrating improved science into regulatory structures, our industry can improve the management of water resources and the environment.
Lessons from Daily White House Briefings
Hydrowonk is a fan of Dr. Deborah Birx. Clear and concise presentation and interpretation of data. Daily updates on how new data changes expectations and models going forward. She shares what scientists are learning today that they did not know yesterday. She talks routinely about how new data compares with earlier predictions. This is a blueprint.
Current Science Model
Current science of water resources and the environment involves a process of measure, model, prescribe, monitor and fight. Identify the area of concern (springflows, groundwater basins, river flows, water quality, species of concern, etc.). Model how the timing and volume of water use impacts the area of concern. Use the model for prescribing regulation. Monitor the area of concern. Stakeholders fight over outcomes. Political controversy and litigation may or may not provide useful feedback into the process. Under adaptative management, continue the process anew.
The water industry needs to test its models. Models should predict as well as prescribe. Monitoring outcomes versus predictions provides the opportunity for learning. Hydrowonk (reluctantly) keeps the fight phase in the process. Stakeholders in fight are too numerous. Perhaps learning can improve the outcomes from stakeholder fights. Learning should improve adaptative management.
Hydrowonk experiences almost daily the difficulty of making predictions. Difficulty is not an excuse for inaction.
Imagine the challenge of the scientists fighting COVID-19. They started with little data at the start of the pandemic. They are innovating on the fly on a virus with initially unknown characteristics. In a recent daily briefing, the phrase “artificial intelligence” was muttered.
A New Science Model Using 21st Century Information Technology
Hydrowonk recommends “The Machine Learning Primer” by Kimberly Nevada. (It is free). Machine learning studies data to detect patterns to:
- predict likely outcomes based on identified patterns
- identify unknown patterns and relationships
- detect anomalous or unexpected outcomes
Does this sound useful? How do predicted impacts of regulatory actions compare with outcomes? How is the environment affected by changes in actual water use? What do differences between predicted and actual outcomes say about our understanding of our environment and economy?
Machine learning uses algorithms (different algorithms learn in different ways). New data is used to improve algorithm performance. “Intelligence” increases over time.
Information technology has increased exponentially the ability to process data and learn. A key point is to test predictions against new data. When tested, does the algorithm accurately predict future events or result in desired outcomes? Can one put predictions into action? Machine learning experts anticipate future advances enabling algorithms to become smarter as new data becomes available.
Machine learning requires application of the scientific method with subject matter experts. Provide a clear statement of the problem or hypothesis to be explored. Apply appropriate scientific rigor. Don’t overcomplicate it. Engage stakeholders in validation.
Implications for the Water Industry
Take advantage of the COVID-19 natural experiment. Deploy resources to increase measurement of what is happening on the ground. Revisit the design of monitoring systems measuring water flows and environmental conditions. The water industry should enter a data building phase. 21st Century IT can handle big data for machine learning.
The water industry should enter an era of prediction, testing and learning. This would represent evolution. For example, the California State Water Project had a zero allocation in 2014, no water. The Department of Water Resources uses a model to assess the deliverability of SWP water every two years. The report before 2014 stated that the minimum SWP allocation would be 8%. That is, there was a zero probability that the actual zero allocation would occur. All updated reports after 2014 have a minimum possible yield above zero. Is this learning?
Improved comprehensive monitoring of actual conditions on the ground can also improve water resource management. The Edwards Aquifer Authority in Texas adopted a comprehensive groundwater permit program in the 1990s to protect springflows at Comal and San Marcos Springs, habitat for endangered species. The amount of allowed pumping under groundwater permits depends on defined triggers at two key wells and actual flows at Comal and San Marcos Springs. There are five stages of “critical management periods”, where the cutback in allowed pumping starts with 20% reduction in allowed pumping that increases to almost 50% under the most severe conditions.
Improved comprehensive monitoring of actual conditions can also be a tool for regulatory approvals of permit applications and projects. The mode is for proponents and opponents to offer “dueling” models. What would happen if a project proponent had to “back-up” their predictions with a mitigation plan of how the project changes, or even terminated, if actual conditions materially differed from predictions. The “soundness of expert opinions” will become part of project risk assessment.
Which impacts from the COVID-19’s natural experiment is expected and which unexpected? What can we learn? How do we use our “increased intelligence”? As with the economic consequences of COVID-19 for the water industry, there is a full agenda.