NewsDrop-April-2024

NOTHING ARTIFICIAL ABOUT THIS COLLABORATION

By: Paul Bertetti , EAA Senior Director, Aquifer Science Research & Modeling

OF THE FIRST 2 YEARS LED TO MORE LOFTY GOALS AS THE COLLABORATION MATURED.

SUCCESSES

THE EDWARDS AQUIFER AUTHORITY (EAA) MUST EVALUATE THE EFFECTS OF CLIMATE CHANGE ON THE AQUIFER SYSTEM TO SUPPORT THE EDWARDS AQUIFER HABITAT CONSERVATION PLAN INCIDENTAL TAKE PERMIT RENEWAL PROCESS & TO PROVIDE INSIGHT ON THE SUSTAINABILITY OF THE AQUIFER.

• Development of XAI models to estimate water levels and spring flow in the Edwards Aquifer under future climates. • Developing XAI tools to characterize and eval uate the effectiveness of aquifer management mitigation strategies. The future of the EAA/UTSA partnership looks bright. EAA is currently funding a longer-term project to advance the modeling of the Aquifer to include additional data, estimating historical pump ing, and incorporating weather data on a broader spatial scale. Leveraging the talent from a nearby university is a desire for many organizations. Often, universities conduct leading edge research, but there is a dis connect in the practical application of that research. Hakan and Debaditya have led the way in setting an example of the advantages of beneficial coopera tion and support between our two organizations. Not only will the technical results be important, but the examples provided via the collaboration effort may be just as beneficial.

links the models use produce projections from training data—often these models are referred to as “black box” models. With the XAI framework, how ever, Dr. Chakraborty can provide quantitative mea sures of the relative importance of model inputs and reveal how the input parameters are related to each other within the model. These explanatory results help investigators understand the important compo nents of the model and aid in interpreting some of the physical processes driving the results. The EAA/UTSA collaboration began modestly with funding of Dr. Chakraborty’s time during the summer months. Still, the partnership produced important results right away. Debaditya and Hakan developed ways to estimate evapotranspiration, per haps the most important part of the water balance in our semi-arid environment, using limited weather station data. They also used climate model results to evaluate the effects of climate change on the energy requirements needed to cool office buildings in the future. The successes of the first two years led to more lofty goals as the collaboration matured. Hakan and Debaditya identified a range of Aquifer system top ics and questions that could be addressed. And they

have since developed multiple projects and worked with several other research teams to advance the science and application of AI/ML. In fact, funding from EAA combined with successful funding grants from other agencies enabled Dr. Chakraborty to hire two post-doctoral fellows to expand UTSA’s capabil ities. Dr. Chetan Sharma is a researcher specializing in climate change impact assessment and an expert in AI/ML modeling, and Dr. Icen Yoosefdoost spe cializes in water resources engineering with a strong focus on environmental sustainability. Both of these excellent researchers have been engaged in EAA-re lated projects for the past couple of years. Recent results and published papers demonstrate the high quality of work accomplished by Dr. Chakraborty’s team and EAA modelers. The list of topics addressed through our modeling efforts is too vast to cover here, but some of the most important contributions include: • Using XAI models, historical weather data, and climate models to characterize and project droughts in semi-arid regions. • Developing applications for predicting soil moisture.

Global climate models can provide projections of future climate, but integrating those projections into the EAA’s computer-based models of the Aquifer is quite a challenge. One of the ways in which we might address this challenge is to use alternative modeling approaches that correlate spring flows and water levels directly to climate parameters like tempera ture and precipitation. One promising alternative approach includes the use of machine learning (ML) models, which are a subset of artificial intel ligence (AI) models. Of course, AI models have been all the rage in news cycles of late, but the

AI/ML models of interest to EAA represent a dif ferent class of models. These have been shown to have great utility in representing time-series data like daily temperature and springflow. As the leader of the modeling group at EAA, Dr. Hakan Başağaoğlu has made it a priority to engage with regional entities to create partner ships in research. When he was hired back in 2019, one of Hakan’s main tasks was to deter mine if computer-based modeling techniques, using AI/ML methods, could help EAA assess the impacts of future climates. A fruitful outcome of Hakan’s efforts over the past few years is the

development of a working relationship with AI/ ML modeling experts at the University of Texas at San Antonio. Starting in 2019, EAA has worked with Dr. Debaditya Chakraborty to develop approaches to modeling the Edwards Aquifer system. Debadi tya is an assistant professor at the University of Texas at San Antonio and an expert in applied AI/ ML research. One innovation of Debaditya and his group of researchers is the development of eXplainable AI/ML models (XAI models). In the past, use of AI/ML has been criticized because of a lack of transparency in the calculations and

Chetan Sharma, Ph.D., is a post doctoral researcher specializing in climate change impact assessment, artificial intelligence, and hydrology. Dr. Sharma has developed innovative methodologies to analyze large-scale geospatial data and develop predictive models for climate extremes and groundwater sustainability.

Icen Yoosefdoost, Ph.D., is a postdoctoral researcher specializing in water resources engineering with a strong focus on environmental sustainability. Her research covers a broad spectrum of topics, from enhancing water distribution and agricultural planning using advanced programming techniques to investigating the impacts of climate change on agriculture and water reserves.

Debaditya Chakraborty, Ph.D., is an Assistant Professor at UTSA’s School of Civil and Environmental Engineering, and Construction Management. He is at the forefront of applied Artificial Intelligence research and is spearheading research in developing an innovative eXplainable and counterfactual AI/ML framework.

[Figure 1] Some of the collaborative effort has produced quantifiable ways to compare historical droughts with each other and to use that information to characterize future droughts. Here, drought intensity of recent droughts is shown relative to the 1950’s drought.

[Figure 2] Spring discharge at Comal Springs can be effectively modeled using only temperature, precipitation, and historical data using XAI modeling approaches. Advances to improve the models and their predictive capability are underway.

10

11