Research
Identified Structural Bias in CMIP6 Sea Surface Temperature Simulations using Reanalysis Data. Presenting at the
Michigan Science Fair.
Climate models are important tools for predicting Earth's climate, but they are not
always accurate. CMIP6 models simulate Earth's climate by solving simplified physics equations on a
three-dimensional global grid. I hypothesize that CMIP6 climate models show larger errors at certain locations and
under specific physical conditions because small-scale processes, such as cloud formation and ocean mixing, must
be approximated in these models.
This research investigates where and when CMIP6 climate models differ from observed data by comparing model output
to ERA5 data, which combines real-world observations with physics-based modeling. The study focuses on sea surface
temperature and analyzes monthly data over a selected ocean region. Model error is calculated by measuring the
difference between CMIP6 predictions and ERA5 values.
Machine learning techniques are then used to identify patterns in the conditions associated with larger model
errors to uncover the specific factors associated with physical location and processes that cause climate models
to deviate from observed data.
