Baltic hydro-climatic data: a regional data synthesis for the baltic sea drainage basin

Understanding and predicting water flows and storage changes on land is crucial for addressing waterrelated scientific and practical challenges across disciplines and geographical regions and scales. The degree to which various hydrological, environmental, ecological, biogeochemical, geological, atmospheric, and climate sciences consistently and realistically capture and represent analogous parts of the terrestrial water system and its spatio-temporal dynamics remains largely unknown and uncertain. Coherently addressing the terrestrial water and related knowledge gaps is essential for tackling the key scientific and practical questions (Zarei such measurements are available through stream discharge monitoring that integrates total runoff across a whole catchment and yields catchmentaverage R by division of the measured discharge with the contributing catchment area. With regard to the vertical water fluxes precipitation (P) and evapotranspiration (ET), precipitation data are available from relatively widespread meteorological monitoring stations, enabling measurement-based interpolation or extrapolation of catchment-average P. In contrast, ET measurements are much scarcer, often requiring model-based estimations due to the lack of sufficient direct flux measurements to cover a whole hydrological catchments. Different datasets are available for addressing the questions and deciphering the interactions involved in the hydrological processes, fluxes, storages, and their changes. Each discipline and sector may then use their preferred dataset, based on some selected combination of ground measurements, satellite observations, and model-based data, to represent, investigate, and predict changes in the terrestrial water system at the scale and world region in focus (Yang et al., 2019;Zhang et al., 2016). Ideally, the different datasets should be consistent, but major discrepancies often emerge (Bring et al., 2015;Ghajarnia et al., 2021;Zarei Destouni et al., 2017). Such data are needed to decipher and quantify the key drivers affecting the nutrient and pollutant loading from land to the coastal waters, as well as the variations and changes in the freshwater flows, quality, and inputs to the sea across the BSDB. The data are vital for the development, design, and evaluation of national and international management plans for water quality and ecosystem health improvements in the Baltic inland, coastal and marine waters, as well as for needed scientific knowledge advancements on crucial water resource dynamics in the changing regional climate and other environmental and societal conditions.The BSDB covers ~1.7 million km² land, and is influenced by large spatial and temporal (seasonal and long-term) variations in hydro-climatic (e.g., precipitation and snowmelt), water quality, and anthropogenic conditions across the multiple countries that are encompassed within it (Hannerz Andersson et al., 2015). To meet the challenge of comprehensive open accessibility to relevant, quality-checked and harmonized hydro-climatic data across these diverse conditions, we here present such a regional dataset synthesis, entitled the Baltic Hydro-Climatic Data (BHCD), for 69 main non-overlapping hydrological catchments with continuous data time series availability over the 30-year period 1980-2020 within the BSDB (Figure 1). The Baltic catchments with relevant data availability included in the BHCD collectively cover approximately 722,235 km² of land, accounting for nearly half of the total BSDB land area. This synthesis is derived by extraction of data for the Baltic catchments represented within the Global Hydro-Climatic Data (GHCD) compilation provided by Zarei Gudmundsson et al., 2018a) and precipitation (P) (Schneider et al., 2016) and modelled data for the associated average annual evapotranspiration (ET), based on the simple model ET≈P-R assuming negligible average annual water storage change (DS≈0); and, for direct comparison with this simple ET model and DS assumption, (ii) Mixed, which synthesizes the same observational R and P data as Obs but differs in the model used for ET, which for Mixed is the global model GLEAM (Martens et al., 2017;Miralles et al., 2011), based on which the implied water storage change can be calculated as DS=P-ET-R; and corresponding terrestrial water data extracted from the global reanalysis products (iii) GLDAS (Beaudoing Rodell et al., 2004) and (iv) ERA5 (Hersbach et al., 2017) that each provide a complete set of model-based data for P, ET and R, from which the implied storage change also can be calculated as DS=P-ET-R.By comparing these different datasets, the BHCD enables a comprehensive assessment of important hydro-climatic dataset consistencies, inconsistencies, and uncertainties for the BSDB region. This can support hydrological, coastal-marine, and climate studies, revealing important hydro-climatic and societal relationships, impacts and feedbacks across the catchments included in BHCD. Use of the BHCD can facilitate identification of spatial and temporal patterns, and key data gaps that need to be bridged within the BSDB, and distinction of consistency, realism, and accuracy among the comparative datasets. The BHCD is also a resource for calibration and validation of hydrological, climate, and related coastal-marine models regarding the multiple catchments included in this data synthesis across the Baltic region. The selection of catchments in the BHCD -as that globally in the GHCD -was based on strict criteria to ensure comprehensive and harmonized spatiotemporal data coverage, open accessibility, and direct comparability across several comparative datasets included in the synthesis. The main catchment Formatted: Font:Formatted: Heading 2, Space Before: 0 pt selection criteria were: (i) a minimum of 300 non-missing monthly runoff values over the period of 30 years for all datasets, (ii) complete dataset consistency, and (iii) the largest spatial coverage as possible with non-overlapping catchments. Areas within the BSDB that did not meet these criteria across all datasets were excluded from this synthesis. While hydrological studies often consider nested catchments (i.e., including smaller sub-catchments withinand, as such, partly overlapping withlarger catchments), the requirement (iii) of non-overlapping catchments with tThe largest spatial coverage requirement was used for the BHCD to aimed to prevent redundancy in data representation and avoid over-representing the same hydrological signals for small sub-catchments inside the larger catchments in the aggregated statistics. Additionally, priority was given to selecting the largest possible catchments that met all other inclusion criteria, thereby maximizing the spatial data coverage while maintaining data integrity and consistency.The BHCDas also globally the GHCD -includes four comparative datasets for the same hydro-climatic variables. The datasets are: (i) Obs, which synthesizes in-situ observational data for runoff (Do et al., 2018a;Gudmundsson et al., 2018a) and precipitation (Schneider et al., 2016) and modelled data for the associated average annual evapotranspiration, based on the simple model ET≈P-R assuming negligible average annual water storage change (DS=P-ET-R≈0); and, for direct comparison with this simple ET model and DS assumption, (ii) Mixed, which synthesizes the same observational R and P data as Obs but differs in the model used for ET, which for Mixed is the global model GLEAM (Martens et al., 2017;Miralles et al., 2011), based on which the implied water storage change can be calculated as DS=P-ET-R; and corresponding terrestrial water data extracted from the global reanalysis products (iii) GLDAS (Beaudoing Rodell et al., 2004) and (iv) ERA5 (Hersbach et al., 2017) that each provide a complete set of model-based data for P, ET and R, from which the implied storage change also can be calculated as DS=P-ET-R.For each of these comparative datasets, tThe BHCD integrates the multiple observational and model-based data for these the main water flux (P, R, ET) and storage-change (DS) variables catchment-wise to ensure consistent and comprehensive variable and spatial and -temporal hydroclimatic coverage between its comparativethe datasets. Main The data sources in Obs and Mixed include observational P data from the "Global Precipitation Climatology Centre (GPCC-V7)" (Schneider et al., 2016) and R data from the "Global Streamflow Indices and Metadata (GSIM)" (Do et al., 2018a;Gudmundsson et al., 2018a). Mixed also includes model-based data for ET and soil moisture (SM) from the "Global Land Evaporation Amsterdam Model (GLEAM)" (Martens et al., 2017;Miralles et al., 2011), which combines satellite observations with its model algorithms. The observational R data from GSIM in Obs and Mixed define the contributing catchments, which are included consistently in all comparative datasets of the BHCD. Additionally, data for air temperature (T) are also included in Obs and Mixed from GHCN-CAMS (Fan Rodell et al., 2004), and on global climate modeling from the "ECMWF Reanalysis 5th Generation (ERA5)" (Hersbach et al., 2017), respectively. A main point of including the different comparative datasets for the same hydro-climatic variables in the BHCD is to facilitate assessment of uncertainty ranges, confidence intervals, and sensitivity analysis in further studies that use the data. That is, for the specific research purposes and catchment areas considered in each study, the results and implications of the different datasets in the BHCD can be directly compared, the consistency/divergence and uncertainty ranges between datasets can be determined, and the result/implication dependence on and sensitivity to dataset choices can be assessed.The BHCD uses cCatchment-wise water balance closure (DS=P-ET-R) as is a fundamental key (Lehmann et al., 2022;Berghuijs et al., 2019;Bring et al., 2015) to facilitate investigating further investigation of the terrestrial water system state and the consistency/ or divergence, uncertainty ranges, and realism of the differentthe comparative datasets around the global land area (Lehmann et Formatted: Space Before: 0 pt Formatted: Font: English (United Kingdom) al., 2022; Berghuijs et al., 2019;Bring et al., 2015). While However, otbtaining consistent, highresolution DS data across numerous catchments of various scales around the world still remains a major challenge.In the BHCD, the water balance closure in Obs is assumed to yield negligible storage change (DS ≈ 0), with ET thereby determined as ET ≈ P -R, while the other comparative datasets include calculated DS=P-ET-R based on their available more elaborately modelled ET data. Obtaining consistent, high-resolution DS data across numerous catchments of various scales around the world is generally a major challenge. If For some areas, however, a reliable combination of ground-based and remote sensing observation datas may be available of for DS were widely available, such, it could instead be that ET that wasis alternatively estimated from catchment-wise water balance as ET=P-R-DS (Bhattarai et al., 2019). For such areas, both the independent DS data and the associated calculated ET data can be compared with the corresponding DS and ET data in the BHCD for further consistency/divergence, uncertainty range, and realism assessment across datasets. However, obtaining consistent, high-resolution DS data across numerous catchments of various scales around the world still remains a major challenge.AdditionallyIn the BHCD, the temporal changes in SM (DSM) are also calculated for each dataset, in order to so that further studies using the data can compare compare and check the internal dataset consistency for thein change directions of DS and DSM, as the latter is also a component of the former. That is, DSM is an integral part of and should be expected to change in the same direction as the total water storage change DS in each catchment (Destouni and Verrot, 2014). For Obs, however, in which DS=0 by assumption, comparison with DSM is not meaningful. There is also no related set of ground-measured SM data to consistently include in the Obs dataset.The measured stream discharge data from GSIM define the 69 non-overlapping catchments consistently included in all BHCD datasets. They also determine catchment-average R in the Obs and Mixed datasets and enable ET calculation in Obs (ET = P -R). Global gridded datasets provide the remaining variables, extracted within each catchment's hydrological boundaries (i.e., water divides). Spatial interpolation generated aggregated catchment value, with an area-weighted averaging approach applied to grid cells intersecting catchment boundaries. Data were processed at the finest consistently available temporal resolution (monthly) to produce catchment-average time series. Additionally, annual and long-term averages were derived to support analysis of longer-term change trends.To compare total average DS, as derived from water balance closure, with DSM, we separately calculated DSM for each catchment, expressed in fractional units per year, consistent with the relative area-normalized SM values (e.g., mm/mm). The calculation was based on average SM over a moving 3-year window, with DSM quantified as the change from one 3-year window to the The of a 3-year window for a assessment of variations while also longer-term in DSM over the 30-year data as needed to facilitate with the corresponding in total The comparison between DS and DSM as an independent of the internal consistency and realism of the DS results implied by each This comparison is relevant and important soil moisture and its changes are part of the water system and are directly to the storage dynamics (Destouni & Verrot, 2014). that the largest of freshwater on & and the entire land surface area of each catchment the the storage changes total DS compared to the surface water storage changes in and that over a of the land surface in an of the data involved in the BHCD dataset, the approach used to this Baltic and synthesis (Zarei & Destouni, goes comparative datasets in the BHCD provide time series of monthly and annual average values for each included with their long-term averages over the 30-year period the Obs dataset not include monthly ET and DS time as its of ET ≈ P and DS ≈ 0 are over at a full or on a monthly The period the that 30 years of data should be used to represent climatic conditions the BHCD can be further as more data available and climatic variables in the datasets are provided as catchment-average values and, in to the catchment-average water flux and total storage-change variables (P, ET, R, the datasets also catchment-average T, and DSM data, with data for catchment-average water change over the total 30-year and the long-term is the fully and flux of long-term and data facilitate important of realism in the DS implications of the different datasets for the various Obs not include data, its DS ≈ 0 = and this specific DS is instead by comparison with the DS implications that emerge as from the other datasets. The calculated DS data and their comparison with the corresponding DSM data in the BHCD can be used to possible important internal within a dataset of different storage change directions implied for DS in as well as storage change implications and uncertainty ranges between the comparative datasets. on the average DS obtained from each dataset, the the BHCD a calculated corresponding average example has calculated as an of dataset The is calculated as the of average DS and an average example of for the that the the entire land surface area of each Zarei and Destouni the of such particularly especially for the Mixed and ERA5 datasets, for which values emerge as with catchmentaverage or by and for some catchments around the global land particularly for ERA5, large and unrealistic water is for catchments in the a of is here to from DS, values of can both within and between catchments on The calculated for a of in the BHCD is is not a specific but a comparative for of assessment of DS realism across the datasets, an also that Obs not include data, its DS ≈ 0 = this specific DS is instead by comparison with the DS implications that emerge as from the other the between datasets for DS and DSM which BHCD is to through of BHCD can modeling to further quantify and uncertainty of the also further of dataset this for the various catchments in et al., is a widely used approach to terrestrial water conditions et al., & et al., et al., Zarei and Destouni that the Mixed and ERA5 datasets yield average that is considerably Formatted: Font: the of long-term average 1 for many catchments around the This an unrealistic water balance closure in these datasets, that a of that provided by P the part to R, is needed to the large modelled ET this water then from continuous water storage average which is also the Mixed and datasets for some catchments around the and in ERA5 particularly so for catchments (Zarei and Destouni, 1 in the BHCD the catchment which were used to data from global datasets and over each Baltic catchment to associated catchment-average variable time series. catchment were from GSIM (Do et al., et al., and to with the used in the BHCD A in 1 the BHCD catchment their corresponding in the where the catchment is and the catchment area in as in GSIM (Do et al., et al., further catchment-average monthly and annual time series for the variables P, ET, R, DS, and across the 69 non-overlapping catchments within BSDB for each dataset in Obs, Mixed, and The Obs dataset includes annual time series and long-term average values for ET and DS, and no associated ground SM data. The data are provided as separately for each variable and Each dataset (i) an that includes annual time series data, (ii) a that includes monthly time series (iii) a that a of all data for the 69 non-overlapping catchments across the comparative datasets, including catchment-wise long-term average values for P, R, ET, DS, DSM, T, and the relative and and the average for all catchments, and associated catchment including catchment of the catchment and of the catchment and catchment areas and (iv) a and that comprehensive the variables in the and including their data units of and in the time series For the SM the also the for the soil moisture used in the associated datasets Mixed on ERA5, and of the BHCD are to the the variable time series in the and BHCD synthesis assessment of important and and associated uncertainty ranges in the catchment-wise water balance closure and hydro-climatic conditions of implied by the different comparative datasets. Understanding the the dataset and and uncertainties can research on the freshwater flux and storage change conditions contributing to and related changes for the Baltic should that the comparative datasets are not fully but some data and with key that can further studies the of For the Obs and Mixed datasets in their ET while GLDAS and ERA5 variations from the BHCD enables further research into the and of the dataset and Formatted: Heading Formatted: uncertainties for the water fluxes, storage and their in the BSDB and the associated implications for the Baltic further to determine the of the comparative datasets for specific catchments and scales of and thereby that validation independent data the water balance and derived measures in the BHCD can be further using available ground-measured and satellite data for independent comparative of catchment-average DS, and Such can confidence in a dataset or areas with need of further The catchments in the Baltic Hydro-Climatic Data synthesis and their around the Baltic Drainage the of the 69 non-overlapping catchments and the Baltic Drainage Basin boundaries derived from GSIM (Do et al., et al., by the using of the and for the Baltic Hydro-Climatic Data synthesis.