Lecture 7 - More on Scientific Method

Induction & Data-Driven Model

Just like with induction, data-driven models are completely based on observations, which are needed to derive the data to input into models.

Typically these observations are from experiments that generate data or observing the real world to collect data

The data will then inform model design, allowing us to analyse for patterns and make predictions (or just understand the system)

Deduction & Process-Based Models

Just like with deduction, we start with logic/theories/laws of nature which help us inform and formulate a question to address an objective.

From then, we can start with making our models, the perceptual part, conceptualising it, deciding equations, calibrating it, evaluating it, and so on and so forth.

Here, we could also feedback and continually improve the model or tweak our question/hypotheses.

Beach Cusps

These patterns of waves as they come onto sandy beaches where there is a consistent horn portrution formning a pattern of horns. We can use the different ways of thinking and different models to understand how beach cusps form, predict them, and so on

Beach Cusps: Data-Driven Model

First, we collected data on the observation of beach cusps in real life. We measure the wave height and plotted it against horn to horn spacing.

With this model, we can somewhat predict the cusp spacing based on wave height.

However, there tended to be inconsistencies within this method, it wasn’t a perfect or even a strong relationship as indicated by the low R² value.

Beach Cusps: Process-Based Model

First, we started with a hypothesis on why it was occuring. We looked through the theory and, in theory, as waves are shoring onto the beach slope and reflect, the development of edge waves are possible (although haven’t been observed irl yet).

These edge waves can then affect sediment transport which then could lead to beach cusps patterns

WIth this, an analytical model was used to validate this hypothesis. Various equations were decided for the model.

However, we haven’t been able to validate the equations and the model with real world observations

Beach Cusps: Physical Model

This would be replicating the beach system itself in order to produce beach cusp patterns in a laboratory setting. This way, we were able to easily control things such as parameters, variables, boundaries, and so on. However, the time and money for this infrastructure & machinery is very expensive.

In the end, it initially resulted with monochromatic forcing, where all the ‘waves’ had a uniform height, length, and so on (due to simplificaitons). Over time, more structure and pattern was observed in the ‘waves’. After 10 minutes, the theorised edge waves were observed producing some beach cusps-like patterns.

But, there are some issues with this physical model, mainly the lack of sediments, how the waves aren’t real (real waves are made through various factors like wind), the defined boundary don’t reflect how they are like in real life (and also caused instabilities).

Nontheless, the data from the physical model was used to inform predictions and was shown that it was somewhat similar to observed measured beach cusp spacing (with lots of scatter though)