When rivers reach a standing body of water, such as a lake or ocean, the sediment (gravel, sand, clay) they are carrying is deposited. These sediment deposits hold clues to past environmental conditions. Motivated by sediment deposits in Utah and Sicily, Italy, this project took a multi-pronged approach to learning about how to decipher these sediment archives. One approach to understanding these deposits is to measure properties of different layers of sediment in field locations. Another approach is to use a computer model to understand how different environmental conditions may be interpreted from sediment deposits. Both approaches were used in this project. The Tulane University team led the computer modeling part of this study, in collaboration with researchers from two other universities who focused on interpreting the sediment deposits and landscapes in the two field areas.
Our aim with the computer modeling was to understand how landslide material moves through a river system to the river mouth and into the sediment archive. Geologically speaking, landslides happen in an instance. The side of a hill, or part of a mountain, slips into a river valley. The material from the landslide, known as the landslide deposit, is slowly eroded and moved downstream by the river (Figure 1). As the landslide deposit moves downstream, it gets smoothed out and reworked. In other words, some of the sediment moves faster and some moves slower. It may take days, years, or centuries for the landslide sediment to reach the river mouth, where it can be deposited into the sediment archive. Ultimately, we would like to know if periods of increased landsliding, due to tectonic or climatic conditions, can be identified in the sediment delivered from a river into a sediment deposit. This will help us identify time periods of increased and decreased landsliding in the past, which gives us more insight into past climatic conditions and potentially earthquake frequencies. This will also help us to understand how climate change could impact the infilling of reservoirs in some mountainous settings.
We used the open source Landlab Python Library to model a range of conditions that could occur naturally. For example, some models had very frequent landsliding (about every 10 years), while in others, landslides may occur about once every 10,000 years. We changed the shape and size of the river network that reworked the landslide deposits and delivered them to the sedimentary archive. We also changed the sediment properties of the landslide deposit (fine clay to large rocks). We chose these properties because we hypothesized they would impact whether landslide deposits arrived at the river mouth in one distinct pulse or got smoothed out through the river network. We ran hundreds of model scenarios. In all scenarios landslides occurred randomly. We could control how likely they were to occur but not exactly when or where they would happen. Figure 2A is an example of a modeled landscape and figure 2B is an example of how the sediment volume per time arriving at the river mouth changes over time. The spikes in Figure 2B are large volumes of sediment from landslide deposits that moved from somewhere in the river network to the river mouth. 
We ran each scenario for millions of years so that we could develop statistically sound results. We analyzed the data by calculating the frequency of large pulses of sediment transported to the river mouth. We specifically were interested in finding whether these volumes passed with regularity or if they were random. If sediment pulses were random, we would not expect to find any decipherable record of environmental signals in the sediment archive. However, if different conditions created different frequencies of sediment delivery, then this may also be the case in real-world landscapes where we can perhaps find similar environmental signals in real sediment archives.
We find that there are interpretable patterns in the sediment volume passing through the river mouth. The details of these patterns vary systematically among our model scenarios. In landscapes with more frequent landsliding, smaller volumes of sediment move through the river mouth more frequently. In contrast, infrequent landslides result in infrequent but larger volumes of sediment moving through the river mouth. We also find that the size of the landscape impacts how strong the periodic signal is. All these results can be used to improve our understanding of real sediment archives.
In the final part of this project, we developed new software to track sediment properties (for example, size, color, or mineral content) as the sediment moves off a hillslope and through a river network. We are still in the process of testing the software capabilities. As soon as it is fully tested it will be made available through the open source Landlab Python Library.

Last Modified: 08/16/2023
Modified by: Nicole Gasparini