Toward Large‐Scale Integrated Surface and Subsurface Modeling

Since the original publication of Freeze and Harlan (1969), both the hydrology and hydrogeology communities have made considerable progress regarding these three questions, particularly for local- to mid-scale applications (i.e., <1000 km2). Based on the large body of literature for surface water and groundwater models at these scales and the growing body of literature for fully-integrated hydrologic models, we believe that these three questions have been satisfactorily addressed at these scales. However, policy decisions are often concerned with much larger scales (e.g., political boundaries, or large watersheds and basins) requiring fully-integrated hydrologic models at increasingly larger scales (i.e., >100,000 km2). While some progress has been made in addressing Freeze and Harlans' (1969) questions at these larger scales, this is still very much an emerging field of research. Where are we now? One of the key benefits of integrated hydrologic models is that they are conceptually simpler to setup (e.g., the upper boundary of the model is driven by precipitation, and recharge is computed internally by the model). However, current computational resources prevent numerical models with meter scale resolution across 100,000 km2. So we either have to build coarse mesh models and rely on upscaling of parameters in the governing equations to compensate for the coarse mesh or use many linked high-resolution models covering the area of interest. Both approaches have their limitations, of course. However, as fast inexpensive computing resources continue to improve (i.e., graphics processing unit (GPU) solvers, model parallelization, central processing unit (CPU) clock-speed, and hyperthreaded chips), the deployment of these large models is becoming increasingly feasible. Data acquisition has also improved with government agencies compiling large harmonized open source data sets such as, for instance, digital elevation models (DEMs), land classification, vegetation mapping, soils maps, and increasingly also hydrostratigraphy. The availability of these datasets greatly facilitates the rapid generation of models for large regions of interest. Where are we going? Given the growth of fully-integrated model applications over the past 10–15 years, we believe that the use of such models to address real-world problems will continue to grow rapidly, both for academic and commercial applications. From our experience, one of the biggest challenges for commercial deployment is the lack of students who are trained to conceptualize and model the terrestrial water cycle in a holistic sense. Educational institutions need to rethink the classical treatment of surface water and groundwater systems as separate domains, both conceptually and numerically, and start teaching these as a single integrated continuum. These students, if supplemented with strong quantitative training, including gathering of datasets from diverse sources, numerical methods, and model construction, will be the leaders in the field of fully-integrated hydrologic modeling. Given the recent technological and scientific advancements that have occurred in this field, we believe that the future of integrated modeling will witness a shift toward living (i.e., real-time) models that are continuously running, driven by short- and medium-term weather forecasts while assimilating terrestrial and space-based sensor data. Many grand challenges remain, however, which have yet to be explored in process-based integrated models. Most if not all applications have yet to address surface and subsurface water quality issues at the basin scale, such as solute fate and transport taking in account multispecies chemical reactions as relevant to lake eutrophication due to agro-nutrient inputs. This is just one example, but many more challenges exist or will be uncovered as the water, food and energy sustainability concerns posed by society in the 21st century continue to grow.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.