Observed and modeled diazotroph nifH abundances in the global ocean

We supplemented the dataset presented in Luo et al. (2012) "Database of diazotrophs in global ocean: abundance, biomass and nitrogen fixation rates" with measurements of nifH genes using qPCR from 17 additional publications. This addition represents a 141% and a 118% increase in the number of depth-integrated and volumetric data points compared to the database of Luo et al. (2012). This new database includes 223 (1428), 253 (1575), 226 (1424) and 144 (958) observations of depth-integrated (volumetric) abundances of Trichodesmium, UCYN-A, UCYN-B and Richelia, respectively. Based on the updated dataset, we applied a machine learning algorithm-random forest to estimate the nifH abundances of Trichodesmium, UCYN-A, UCYN-B and Richelia in the global ocean. The updated dataset of observed diazotroph nifH abundances, machine learning algorithms, codes and the modeled diazotroph nifH abundances in the global ocean are deposited here.1. Diazotroph nifH abundances in the global ocean: observed diazotroph nifH abundances compiled from the literature2. Diazotrophs_RF_models: random forest models built in MATLAB to estimate diazotroph abundances in the global ocean including "model_Tricho_tree", "model_UCYNA_tree", "model_UCYNB_tree", and "model_Richelia_tree"3. Diazotrophs_RF_global_monthly: monthly average diazotroph abundances (nifH gene copies m-2) in the global ocean estimated by random forest models 4. diazotroph_abundances_prediction: MATLAB script file to predict diazotroph abundances using random forest models ("Diazotrophs_RF_models"), predictors ("predicotors_diazotrophs") and basin separation file ("basin_separation")5. predictors_diazotrophs: predictors at monthly scale [1.lat; 2.lon; 3.month; 4.solar-flux; 5.wind; 6.sst; 7.sss; 8.log10(nitrate); 9.log10(phosphate); 10.log10(p*); 11.oxygen; 12.par; 13.mld; 14.par_mld; 15.log10(chl_a)); 16.iron]6. basin_separation: a file used to denote and separate different ocean basins and land