Supporting data for Tradeoff of CO2 and CH4 emissions from global peatlands under water-table drawdown

data_revision1.csv collects flux measurements from water table manipulation experiments listed in Reference.doc. site_chara_filled_revision1.csv collects site characteristics, soil, climate, management and topographic information. data_in_shape_permu_revision1.r is the script that combines previous...

Full description

Bibliographic Details
Main Authors: Yuanyuan Huang (8769083), Phillipe Ciais (7680905), Yiqi Luo (248188), Dan Zhu (139675), Yingping Wang (8343481), Chunjing Qiu (9560398), Daniel S. Goll (8540778), Bertrand Guenet (8023223), david makowski (4677643), Inge De Graaf (9560400), Jens Leifeld (8132049), Min-Jung Kwon (414708), Jing Hu (41899), Laiye Qu (8042660)
Format: Dataset
Language:unknown
Published: 2020
Subjects:
Online Access:https://doi.org/10.6084/m9.figshare.13139906.v3
Description
Summary:data_revision1.csv collects flux measurements from water table manipulation experiments listed in Reference.doc. site_chara_filled_revision1.csv collects site characteristics, soil, climate, management and topographic information. data_in_shape_permu_revision1.r is the script that combines previous two csv datasets and reorganize it into high-low water table pair for each flux. data_in_shape_permu_revision1.r generate files XXXX_ final_data_revision1.csv that are used in analysing and plotting. pred_nee_future_map_rev2_full.py and pred_ch4_future_map_rev2_long_full.py generate gridded global predictions of future CO 2 and CH 4 in response to water table drawdown through machine learning models (random forest) trained by the collected water table manipulation experimental datasets. Code in the uncertainty folder combines bootstrap resampling and ensemble predictions to generate 95% confidence intervals. For the response of CO 2 , the code randomly sampled 80% of observation samples to build one random forest model. This random model was then used to make future predictions with different combinations of predictor datasets. This bootstrap resampling, random forest model building and future prediction was repeated 200 times. In total, the code generated 25200 (200 x 21 CO 2,initial datasets x 2 WTD initial datasets x 3 Climate datasets) ensemble members and estimated the 95% interval as the indicator of prediction uncertainty. For the response of CH4, the code generated 8400 ensemble members through 200 times bootstrap resampling, 7 CH 4,initial datasets, 2 WTD initial datasets and 3 climate datasets. Scripts in the “Figures” folder are used to generate figures in the main text and the supplementary information with, FigureS referring to figures in supplementary information. delta_rcp85_figshare.nc stores means and the 95% confidence intervals of future CO 2 and CH 4 emissions in response to water table drawdown across the globe under RCP8.5 climate conditions, in unit of mg CO 2 -eq m -2 h -1 . delta_rcp26_figshare.nc stores means and the 95% confidence intervals of the response under RCP2.6 conditions. We assume CH 4 has a global warming potential that is 25 times CO 2 over a 100-year time horizon. Note for site_chara_filled_revision1.csv WT: >0 below soil surface; <0 above soil surface Flux: positive sign = release gas to the atmosphere Manipulation length: s: within 1 year; m, 1-10 years; l, >10 years Arctic: North of 66.5N; Boreal, 50-66.5; Temperate, 30-50; tropical, -30 – 30 CH 4 , unit, µg m -2 h -1 CO 2 , unit, mg m -2 h -1