Data and code associated with "Marine heatwaves and iceberg melting in polar areas intensify phytoplankton blooms"

This record contains two sets of files related to the paper "Marine heatwaves and iceberg melting in polar areas intensify phytoplankton blooms";

The processed data (IPSL-CM6A-LR.zip, UKESM2-0-LL.zip, CESM2.zip, GISS-E2-1-G.zip, CNRM-ESM2-1.zip and Training Datasets.zip)  used in the training and evaluation of the proposed model (Figure 5a-5c in the manuscript).

The source code used in the training and analysis described in the paper.

These two sets of files are described further in the sections below.

1. The processed data generated by Earth system models

The seawater data are based on the last 5 years of monthly historical data generated by 5 different Earth system models (ESMs, source IDs: IPSL-CM6A-LR, CESM2, CNRM-ESM2-1, GISS-E2-1-G, IPSL-CM6A-LR, UKEMS1-0-LL; ESGF MetaGrid (llnl.gov)) from 2010 to 2014 within CMIP6, supplemented by the sea ice concentration (product name: HadISST1, Met Office Hadley Centre observations datasets), MHWs occurrence probability (Marine Heatwaves: NOAA Physical Sciences Laboratory), and two different datasets for model validation and uncertainty assessments (Ocean Productivity: Online VGPM Data (oregonstate.edu) and Global Ocean Colour (Copernicus-GlobColour), Bio-Geo-Chemical, L4 (monthly and interpolated) from Satellite Observations (1997–ongoing) | Copernicus Marine Service). The data processing methods are detailed in the paper.

The features of above processed data are as follows:

Spatial information

longitude (lon) and latitude (lat)

Climatic factors

partial pressure of carbon dioxide in surface seawater (spCO2)

Seawater conditions

seawater surface temperature (SST), hydrogen potential (pH); dissolved iron (Fe), dissolved oxygen (O2), silicate (SiO3), nitrate (NO3), phosphate (PO4), dissolved organic carbon (DOC), ammonium () levels and salinity (SAL)

Phytoplankton properties

chlorophyll a (Chla), iron limitation of diatom growth (LimFe), limitation of diatom growth due to solar irradiance (Limirr), and particulate organic matter concentration presented as the carbon content in seawater due to net primary productivity (NPP)

Additional indicators

sea ice concentration (SIC) and marine heatwave occurrence probability (mhw_probability)

Note that, NPP is the target variable for training, and the remaining seawater factors are feature variables for screening and training.

2. Methods and model

The source code includes data cleaning (Assembled Outlier Detection.py), feature contribution evaluation (Spearman-LASSO coefficient), optimized base learners (RF-HHO.py, XGB-HHO.py and CNN-HHO.py) and super learner (SuperLearner.py).