Written by Ethan Lee
Glaciers across the tropical Andes are useful water towers for downstream communities1. During periods of drought, the melt water of these glaciers can buffer against the effects.
Due to recent climate change, these glaciers are shrinking, reducing their water holding capacity2. This has massive implications in a changing world, where climate is set to reduce overall yearly rainfall across the Andes, leading to reduced run off and stream flow3. Coupled with the reduction in glacial area and volume, the effect of future climate extremes4 cannot be buffered against.
This leads to one of the overarching questions of the Deplete and Retreat project; how will future changes in glacial extents impact water resources by 2150 AD?
One aspect of the Deplete and Retreat project that shall aid in answering this question, is the use of glacial numerical modelling techniques.
Numerical modelling glaciers within the Andes
Glacial numerical modelling, while not a new method, has in recent years become an important technique in understanding the change of glaciers under current and projected future climate change. However, many glaciers across the Andes are valley or ice cap glaciers that require fine resolution mapping to ensure glacial and topographical interactions are modelled accurately. This has meant that numerical modelling across the Andes has been limited to single glacial valleys.
The improvement of computers, along with the use of high performance computing (HPC; Figure 1), has meant we are now able to numerical model entire regions at much finer resolutions, down to 100 m or less across large areas. This allows us to overcome past obstacles of coarse resolution modelling and begin to conduct regional glacial numerical modelling. For this project, we selected the Parallel Ice Sheet Model (PISM) to model our selected catchments.
Within PISM, “Parallel” means it is able to scale well within a HPC environment, using more CPU cores to compute mathematical calculations faster. PISM has also been used extensively across valley based glacial regions providing us with the confidence that it will be useful for our catchment regions that incur high topographical variations, which should reconcile their frontal positions.
Overcoming numerical model uncertainties
Uniquely for the Deplete and Retreat project, the chosen catchments cover 4 key climatological regions within the Andes (Figure 2). These have been discussed in detail in the Andean mountain hydrology. The number of differing climatic zones means that one of the most important factors of glacial modelling is representing the correct climate. While the PISM model includes a basic climate model (a Positive Degree Day (PDD) there is little information of the pass climate in the Andes. If we do not know what the climate was, we cannot accurately model glaciers in the past.
To overcome these limitations, the Deplete and Retreat project is using a weather model in (in WP2), to model the past, present, and future climate over the Andes using robust global climate models (GCMs). We will also use a mass balance model to generate yearly mass balance fields between 1850AD and 2150AD. Figure 3 details how these two models are used to generate a continuous model input for PISM. Hopefully, if we are able to model climate somewhat realistically in the past and present day, it should be representative of the future.
Our approach to numerical modelling glaciers of the Andes
1. Sensitivity analysis of the PISM model
Sensitivity analysis is how the outputs of the model are influenced by the inputs. This allows us to understand how varying model inputs parameters impact the outputted glacial extents (area and volume).
Sensitivity analyses are important to determine which parameters are the least important and can be set to fixed parameters and discounted, while the most impactful can be varied to best parametrise them for subsequent work. The most impactful parameters shall be varied within the next stage of the modelling.
2. Spin-up runs to Little Ice Age (Neoglacial; 1850 AD) extents
When running numerical models into the future, you want to choose a period of time when there were no glaciers or when they were at their most extensive position. This period is where the ice will be grown to before running the model forward in time, this is the starting point, or the ‘spin-up’ run. WP1, who are doing fieldwork and mapping of the catchments, shall detail the Little Ice Age (LIA) positions of glaciers in the catchments, the last time glaciers advance. These outlines shall be used to grow ice in PISM to their LIA extents.
To spin the model up to the LIA positions, a number of models runs (or an ensample), with different parameter values shall be used. The output of model extents will be compared to the LIA outlines. The best determined models that fits the LIA extents (also known as the ‘best-fit’) shall the be used for the next model runs.
3. Past to present model runs
With the determined best-fit spin up outputs, the model shall be run forward to present-day. These are to ensure the model is accurately representing the glacial extents both in the past and the present. Models that do not match the present-day extents shall be discounted, while the model configuration that best represents the present-day extents shall then be used to run forward into the future.
With this we can have high confidence that the model configuration, and the ice generated within it, are going to best represent future ice extents.
4. Present to future
Much like the past to present model runs, a single model run that best represents the past and present glacial extents shall be used to force the model into the future. These will have the entire glacial thermal history within the model, having been spun up at the LIA, and forced forward to the present day. It is known that glacial ice incurs its own ‘memory’ from its past extents, and thus, having that history of the ice from the model ensures that the glacial dynamics within the model best represent the real world.
5. Glaciers and climate extremes
While understanding glacial changes to future climate change is important, climate extremes that occur throughout the time period may hasten or dull the potential demise of these glaciers. Such climate extremes are those of heatwaves (e.g., 2023 South American heatwave), increased precipitation events, and if ENSO events are made more extreme in the future.
Where this fits in for Deplete and Retreat
After the glacial modelling has been conducted, the story does not end. The output from the glacial numerical modelling shall be used in the hydrological modelling aspect of the Deplete and Retreat project (in WP4). This is to understand has glaciers have receded in a warming climate, how much runoff is expected from these glaciers, and how will this change the downstream runoff and overall discharge of the catchment. This will have important ramifications for adaptation strategies and local mountain region water security.
About the Author
Ethan is a PDRA based at Sheffield University from January 2024, focusing on numerical modelling of glaciers (WP 3). In D&R he will be leading the numerical modelling and evaluation of ensemble members against the observational datasets produced in WP 1.