Agreement, opposition, and dataset influence in global evapotranspiration trends

ET Datasets included - reanalysis products: ERA5-Land, JRA-55, and MERRA-2 - remote sensing-based products: BESS v2, ETmonitor, GLEAM v4.1a, and MODIS 16A2 - land surface model-based products: FLDAS, GLDAS-CLSM v2.1, GLDAS-NOAH v2.1, GLDAS-VIC v2.1 and TerraClimate and - composite products: CAMELE and SynthesizedET, ## Global gridded results All spatial data are provided on a regular geographic latitude–longitude grid (EPSG:4326) at 0.25° spatial resolution. 1. Trend estimates by product for each grid Data: global_grid_trends_all_datasets.csv Contains: - Dataset: ET input dataset - lon, lat: Coordinates at center point of 0.25x0.25 grid - p-value, slope, lower and upper: Output from openair package. Theil–Sen trend estimate and lower/upper 95 % bootstrap confidence bounds in mm yr$^{-2}$. 2. Trend indices Data: data_fig_2_grid_trend_stats.csv Contains: - lon, lat: Coordinates at center point of 0.25x0.25 grid - DCI_all_brk: Dataset Concurrence Index for all trends irrespective of p-value - more-sig_trends: p-value categorized according to the smallest significance threshold met (1, 0.1, 0.05, 0.01) at which the majority of trends becomes significant. - p_val_opposing: p-value categorized according to the smallest significance threshold met (1, 0.1, 0.05, 0.01) at which significant trends have different direction 3. Topology score by product for p-value of 0.05 Data: grid_dataset_trend_topology.csv Contains: - lon, lat: Coordinates at center point of 0.25x0.25 grid - dataset: ET input dataset - area: grid cell area - positive_signal: 0 (false) and 1 (true) if positive signal category is met, i.e. trend significant and positive - negative_signal: 0 (false) and 1 (true) if negative signal category is met, i.e. trend significant and negative - dampener: 0 (false) and 1 (true) if dampening signal category is met, i.e. trend is nonsignificant - contributing_opposition: 0 (false) and 1 (true) if dataset is the only dataset contributing to opposition - opposing_trend: 0 (false) and 1 (true) if dataset has a different sign than majority trend direction estimated by Dataset Concurrence Index - opposing_significance : 0 (false) and 1 (true) if dataset has different significance than the majority of datasets ## Data used to create figures Figure 1: Examples of time series illustrating topology categories. The time series highlighted in red represents the dataset that fulfills the definition of the respective topology category, while the remaining ensemble members are shown in gray. The cases shown here represent illustrative examples and do not restrict the sign of the trend for opposition-based categories. In the examples shown, the positive signal booster exhibits a positive statistically significant trend, the negative signal booster exhibits a negative statistically significant trend, and the signal dampener exhibits a non-significant trend. The trend opposer exhibits a negative trend while the majority of ensemble members show a positive trend direction. The significance opposer exhibits a statistically significant trend while most ensemble members show non-significant trends. The opposition contributor exhibits a statistically significant negative trend while the remaining ensemble members show either positive significant or non-significant trends. Dataset name: data_fig1_topology_methods.csv Column names: - lon, lat: Coordinates at center point of 0.25x0.25 grid - year - dataset: ET input dataset - evap: Annual ET value for given year - topology category - highlight (true or false) to highlight dataset Figure 2: Variation in ET trends. (a) Variation of global annual trends across data sets including trend significance. The black line indicates the upper (95 \%) and lower (5 \%) trend bounds from block-bootstrap for BESS v2, CAMELE, ERA5-Land, ETMonitor, FLDAS, GLDAS-CLSM v2.1, GLDAS-NOAH v2.1, GLDAS-VIC v2.1, GLEAM v4.1a, JRA-55, MERRA-2, MOD16A2, SynthesizedET, and TerraClimate. (b) Map of majority trend direction of all trends irrespective of their significance. Red indicates a positive direction and blue indicates a negative direction. Gray indicates an equal number of positive and negative trends.(c) Map of quartile uncertainty in magnitude and direction of global annual trends. (d) Map of p-value indicating onset of directional opposition of significant trends. Data: a. data_fig_2_global_evap_trend.csv Contains: - dataset: ET input dataset - slope, lower and upper: Output from openair package. Theil–Sen trend estimate and lower/upper 95 % bootstrap confidence bounds in mm yr$^{-2}$. - trend_significance: categorical significance level based on p-value threshold. b and d: data_fig_2_grid_trend_stats.csv Contains: - lon, lat: Coordinates at center point of 0.25x0.25 grid - DCI_all_brk: Dataset Concurrence Index for all trends irrespective of p-value - more-sig_trends: p-value ( grouped as 1, 0.1, 0.05 or 0.01) at which the majority of trends becomes significant. - p_val_opposing: p-value (grouped as 1, 0.1, 0.05 or 0.01) at which significant trends have different direction c: data_fig_2_grid_quartile_stats.csv Contains: - lon, lat: Coordinates at center point of 0.25x0.25 grid - min: Minimum trend in ensemble - max: Maximum trend in ensemble - Q25: Lower quartile Q25 of ensemble trend - Q75: Upper quartile Q75 of ensemble trend - fold_brk: Symmetric ratio group to define quartile uncertainty in magnitude - fold_brk_detailed: Detailed symmetric ratio group - sign: If signs of Q25 and Q75 are the same or different to define quartile uncertainty in direction - problem: quartile uncertainty in magnitude, direction, both or none Figure 3: ET trends and quartile uncertainty across IPCC reference regions . (a) Annual ET trend estimates in mm yr$^{-2}$ were calculated from annual data covering 2000-2019 for the ensemble and for BESS v2, CAMELE, ERA5-land, ETMonitor, FLDAS, GLDAS-CLSM v2.1, GLDAS-NOAH v2.1, GLDAS-VIC v2.1, GLEAM v4.1a, JRA-55, MERRA-2, MOD16A2, SynthesizedET, and TerraClimate. Ensemble trends were estimated based on the time series of the ensemble mean where all products were weighted equally. The data products are ordered by mean trend magnitude, with the dataset with the highest positive trend shown at the top. (b) Area fraction of grid-scale quartile uncertainty in trend direction, magnitude, both or none. The IPCC reference regions in a and b are ordered by their area fraction of quartile uncertainty, with the IPCC reference region with highest area fraction listed first. (c) Hexagon map of IPCC reference region with quartile uncertainty using aggregated trend estimates from a. Central American IPCC reference regions were grouped with North America. a. data_fig_IPCC_ref_regions_trends_by_product.csv Contains: - dataset: ET input dataset - IPCC_ref_region: Reference region defined by the intergovernmental panel on climate change AR 6. - p, slope, lower, upper: Output from openair package. Theil–Sen trend estimate and lower/upper 95 % bootstrap confidence bounds in mm yr$^{-2}$. - trend_direction_detailed: trend direction and significance in words - region: Larger continental region b. data_fig_IPCC_ref_regions_problem_area_fraction.csv Contains: - problem: quartile uncertainty in magnitude and direction - IPCC_ref_region: Reference region defined by the intergovernmental panel on climate change AR 6. - problem_area: Area fraction - ipcc_area: Area fraction of IPCC reference region (with complete dataset coverage) - ipcc_fraction: problem_area divided by ipcc_area to obtain area fraction of fraction of the IPCC reference-region area affected by the specified uncertainty category. - region: Larger continental region c. data_fig_IPCC_ref_regions_problem_aggregated.csv Contains: - IPCC_ref_region: Reference region defined by the intergovernmental panel on climate change AR 6. - Q25: Lower quartile Q25 of ensemble trend - Q75: Upper quartile Q75 of ensemble trend - fold: Symmetric ratio of Q25 and Q75 - sign: Different or same sign of Q25 and Q75 - problem: quartile uncertainty in magnitude, direction, both or none - region: Larger continental region Figure 4: Global topology of trend signatures. Darker color and larger circles indicate a stronger signature of a given role. Datasets are ordered according to their rank as "Trend opposers", highlighting products that most frequently deviate from the majority trend direction across the ensemble Data: global_dataset_trend_topology.csv Contains: - dataset: ET input dataset - rank_trend_opposer: Global dataset rank of dataset with different sign than majority trend direction estimated by Dataset Concurrence Index - rank_opposition_contributor: Global dataset rank of dataset contributing to opposition - rank_significance_opposer: Global dataset rank of dataset with different significance as majority estimated by significant dataset count - rank_dampener: Global dataset rank of dampener category - rank_pos_signal: Global dataset rank of positive signal category, i.e. trend significant and positive - rank_neg_signal: Global dataset rank of negative signal category, i.e. trend significant and negative - p-value: p-value threshold used for significance definition Figure 5: Topology of trend signatures for three IPCC reference regions South American Monsoon (SAM), Tibetan Plateau (TIB) and West-Central Europe (WCE). Dark color and larger radius indicate a stronger trend signature for a given category. Datasets are ordered according to trend opposer of the global topology. Data: ipcc_ref_regions_dataset_trend_topology.csv Contains: - dataset: ET input dataset - IPCC_ref_region: Reference region defined by the intergovernmental panel on climate change AR 6. - rank_trend_opposer: Global dataset rank of dataset with different sign than majority trend direction estimated by Dataset Concurrence Index - rank_opposition_contributor: Global dataset rank of dataset contributing to opposition - rank_significance_opposer: Global dataset rank of dataset with different significance as majority estimated by significant dataset count - rank_dampener: Global dataset rank of dampener category - rank_pos_signal: Global dataset rank of positive signal category, i.e. trend significant and positive - rank_neg_signal: Global dataset rank of negative signal category, i.e. trend significant and negative Figure 6: Where does GLEAM 4.1a oppose majority trend direction the most? Data: ipcc_ref_regions_dataset_trend_topology.csv - dataset: ET input dataset - IPCC_ref_region: Reference region defined by the intergovernmental panel on climate change AR 6. - rank_trend_opposer: Global dataset rank of dataset with different sign than majority trend direction estimated by Dataset Concurrence Index - rank_opposition_contributor: Global dataset rank of dataset contributing to opposition - rank_significance_opposer: Global dataset rank of dataset with different significance than majority estimated by significant dataset count - rank_dampener: Global dataset rank of dampener category - rank_pos_signal: Global dataset rank of positive signal category, i.e. trend significant and positive - rank_neg_signal: Global dataset rank of negative signal category, i.e. trend significant and negative