Validation Discussion

Find a modeling article that you can critique for verification and validation. This could be a new article, or one from a list of provided articles in other online discussions. Find the lists of suggested articles by going to Content > Articles for post-session discussion. Article critiques: Create a new thread and post your response. Your critique should be at least 300 words and should address the following questions:

Describe (or remind us) what the model was, and what research question it was used to investigate.

How did the researchers go about verifying the model?

How did the researchers go about validating the model?

What else could be done (by these researchers or other researchers) to make sure the model is credible?

Include a link or full citation for your source material.

Reply posts: Next, write substantive, thoughtful replies to at least two of your peers' posts. Reply posts should address the following prompts:

Do you agree with the assessment in the original post? If you do agree, indicate what convinced you. If you don't agree, explain why.

What other methods of verification and/or validation could the researchers have used?

In your opinion, what are the biggest challenges for verification and validation of scientific models?

Post one to reply:

This was probably my favorite Model and study from the course. The study was all about the use of computer simulating and finding a method of planting maize efficiently. They investigated many questions, such as "What will the influence of field seedling emergence rate on the yield be like?", "Which planting methods is the best?", "Planting density will influence the yield, then how much will the influence be?", and "Different planting density can lead to the yield compensation because of the sparse breaks up, then how much will the compensation be?"

To Verify their model essentially they only needed to make sure their software was working, and was working as intended. They output several visualizations to verify that their simulation is working as intended.

To Validate, the researchers used mathematical depictions based on previous field experiments and studies of maize planting. By basing their model on real world data they can to a certain degree replicate the natural world.

This is only a model of course, and to completely validate their model for the natural world would involve many more factors to be considered. The researchers mention such aspects as seed quality and variety, soil factors such as acidity, fertilizers, basic things like sunshine and weather or moisture where much of their mathematics they utilized will have be changed to more accurately the germination and growing of maize.

The study can be found here.

Second Post to reply:

Describe (or remind us) what the model was, and what research question it was used to investigate.

The model is intended to research Sea surface temperature patterns that occur with global warming. The interest in investigating these patterns is due to information about how the temperature in the Arctic Ocean have increased without negative feedback loops from ice melt. It has generally been understood that the water warms when ice melts from global warming, and there is less mass to reflect solar radiation. Since the information indicates that warming is occurring without the accepted theory, researchers were interested in learning why this is the case.

How did the researchers go about verifying the model?

In this model the researchers aggregated 17 prior models with verified information about longwave radiation. Another study had pointed to the data being similar in warming patterns so that seemed to indicate there may be information that could be inferred from studying the data. They also worked with two twenty year data sets of historical climate and projected climate information in what they referred to as a multimodel ensemble analysis approach in order to reduce internal noise.

How did the researchers go about validating the model?

So for one, they applied some fancy math that I don’t understand, and that helped them to find patterns in the data. Then they compared the data from their model to the models from their aggregated lists average data.

What else could be done (by these researchers or other researchers) to make sure the model is credible?

A main factor in the results of this model being credible have to do with the information from the previous studies being accurate, and the metrics for measuring short and longwave radiation being available. Only 13 of the 17 models had this data available in the way that the researchers needed, so I think if there were a way to aggregate more data related to the short and longwave radiation that was used to calculate the results, and if after that the results we're able to be repeated, then the model would be more credible.

Include a link or full citation for your source materia

https://journals.ametsoc.or

less