Wednesday, October 7, 2026

Daily Forcings Datasets (Rainfall and Temperature) for hydrological Purposes over the Po river district

If you want to run a hydrological model in the Po basin, or anywhere in northern Italy, the first question is which meteorological forcing to use. A few years ago the choice was short. Today there are at least a dozen gridded products, regional and European, built with different networks, methods and purposes. This post is a review of what is available, written from the point of view of someone who needs daily forcing for water budgets with GEOframe or a similar model.

A disclosure: with Hossein Salehi and colleagues I have a dataset of my own in this list (Salehi et al., preprint, currently in open discussion at ESSD). I have tried to judge it with the same yardstick as the others.


What a hydrologist needs from a forcing dataset

Not every good climatological product is a good forcing. For water budgets, these are the requirements I use:

Requirement Why it matters
Daily time step, with the day aligned to discharge records (00–24) Snow, soil moisture and routing run daily; misaligned days shift peaks
Precipitation and temperature from the same network and method Rain–snow partition, melt and ET depend on both; mixing products mixes biases
Tmin and Tmax, or radiation Needed by Hargreaves or Priestley–Taylor ET and by melt and frost thresholds
Resolution close to the model units, about 1 km in Alpine sub-basins Aggregating is exact; disaggregating needs assumptions
Coverage of the whole hydrological district, Swiss and French headwaters included Budgets must close over the basin, not over administrative borders
Uncertainty for each cell and day So that forcing error can be carried into budgets and calibration
Station data and metadata So that users can check, re-interpolate or bias-correct
Documented behaviour with elevation and undercatch High-elevation input errors dominate Alpine budgets
Open licence, persistent identifier, standard format Reuse, citation, reproducibility (the FAIR principles)
Updates after 2020 Operational water-balance and drought monitoring

No product meets all of them.

The datasets

Dataset Variables Period, step Grid Method
ARCIS precipitation (Pavan et al., 2019) P 1961–present, daily (9-to-9 local day) ~5 km Modified Shepard with topographic distance, on 1,048 homogenized series
ARCIS temperature (Pavan et al., 2026) Tmin, Tmax 1991–present, daily, monthly updates ~5 km Piecewise lapse rate, urban and water fraction, in 28 sub-areas; Shepard on residuals
IH-GAR (Manara et al., 2026) P 1951–2023, daily and monthly ~800 m Anomaly method: elevation-based normals, monthly anomalies, daily fractions
APGD (Isotta et al., 2014) P 1971–2008; extended version to 2019 5 km Anomaly method with PRISM-type normals; Alpine part only
Crespi et al. (2018) P normals 1961–1990 climatology ~800 m Local weighted regression on elevation
Crespi et al. (2021) P, mean T 1980–2018, daily 250 m Anomaly method; Trentino–South Tyrol only
Salehi et al. (preprint) P, mean T 1991–2020, daily 1 km Kriging with daily variograms; daily lapse rate for temperature
EMO-1 (Salamon et al., preprint) P, Tmin, Tmax, T, wind, radiation, vapour pressure 1990–2024, daily and 6-hourly ~1.5 km Angular distance weighting
E-OBS P, T, Tmin, Tmax and more 1950–present, daily ~11 km Ensemble interpolation
CERRA-Land P and land variables 1984–present 5.5 km Regional reanalysis
EEAR-Clim (Bongiovanni et al., 2025) Station P, T, Tmin, Tmax to 2020, daily stations Quality-controlled, homogenized series, about 9,000 stations
BIGBANG (ISPRA) Water-balance outputs 1951–2024, monthly 1 km National monthly water-balance model

A few notes on each family.

ARCIS is the work of the regional meteorological services of north-central Italy. Its strengths are care for the data (homogeneity tests, synchronization of 9-to-9 readings) and the fact that it is updated operationally; its temperature companion (Pavan et al., 2026) handles Po Valley inversions with sub-regional, two-slope lapse rates. Its limits for Alpine hydrology are the 5 km grid and the absence of an elevation model for precipitation: Manara et al. (2026) find ARCIS lower than other products above about 750 m.

IH-GAR (Manara et al., 2026) uses the densest network (7,417 stations in its calibration domain, about 2,000 active each year) and the anomaly method, so its normals capture the increase of precipitation with elevation where stations are missing. Its daily values are built from monthly anomalies and interpolated daily fractions; the paper validates normals and monthly anomalies, but I could not find a validation of the daily fields themselves. Temperature has to come from another product.

A caution that applies to all products built from normals. IH-GAR, APGD and Crespi et al. (2021) obtain daily values by scaling station anomalies with elevation-aware normals. A normal is the number of wet days times their intensity, so this scaling can turn a higher frequency of rain at altitude into heavier storms. Monthly totals stay correct; daily wet-day frequency and intensity at altitude are worth checking before using these products as daily forcing.

APGD (Isotta et al., 2014) remains the Alpine reference for daily precipitation and shows the sharpest valley gradients. It covers only the Alpine part of the basin and is not updated.

EMO (Thiemig et al., 2022, and EMO-1) is the most complete in variables, open and updated. It is worth knowing how it is built in the Alps: EMO-5 imports about 5,000 APGD grid points as virtual stations, and its direct Italian station input is small. In the Alpine part of the Po basin it largely re-grids APGD.

E-OBS and CERRA-Land are useful for large-scale consistency, but too coarse, or too model-dependent, for Alpine sub-basins.

Our dataset was designed as hydrological forcing: precipitation and temperature from the same network (1,583 precipitation and 1,555 temperature stations, from the Po River District Authority and EEAR-Clim), on a 1 km grid, over the whole district plus a 20 km buffer. The review it received was useful and sometimes hard. The precipitation field is smoother than ARCIS, IH-GAR and APGD in some Alpine valleys, Valtellina above all. The lapse rate is one per day for the whole domain, so it misses local inversions. The record stops in 2020, and ordinary kriging of daily totals spreads small amounts across the wet/dry boundary. In the revision we are testing, on the gauges and within single events, whether elevation acts on the number of wet days or on storm intensity, and we will build precipitation from occurrence and intensity separately. What it adds is the daily variograms, the daily lapse rates and the per-cell uncertainty, which we will release with the fields.

How many stations, in the same area?

Comparing station totals is misleading, because each product covers a different domain. What matters is how many stations support the field inside the area you model. For the Po River District with a 20 km buffer (about 90,000 km²):

Dataset Stations in the district Basis
Salehi et al., precipitation 472 (1991–2000), 954 (2001–2010), 1,108 (2011–2020) active per day Exact
ARCIS precipitation roughly 500–580 Estimate, scaled by area from about 1,000 series per year
ARCIS temperature roughly 570 (1991) to 930 (after 2013) Estimate, scaled by area from their Fig. 1a
IH-GAR roughly 1,100–1,400 active per year Estimate, from their station density
EMO few direct Italian stations; APGD points in the Alps Thiemig et al. (2022)

The estimates assume uniform station density and should be replaced by real counts. They still show a point that is easy to miss: station support changes a lot over time. Our network more than doubles after 2000, and any network that grows this much makes trends from the gridded series suspect unless the series are homogenized and the network is controlled, as ARCIS and IH-GAR do.

Fitness for water-budget modelling

Dataset P and T, same network Tmin/Tmax ≤ 1.5 km Whole district Per-cell uncertainty P with elevation Homogenized After 2020
ARCIS P + T same agencies, different methods yes no Italian part no no yes yes
IH-GAR P only no yes yes no normals yes to 2023
APGD P only no no Alps only ensemble version on request normals no to 2019
EMO-1 yes yes yes yes yes (EMO-5) limited no yes
E-OBS yes yes no yes ensemble spread no no yes
Salehi et al. yes mean T only yes yes yes, kriging variance and cross-validation tested and rejected no no

My reading, by use:

  • Long-term climate and trends: ARCIS and IH-GAR, which are homogenized and designed for it.
  • Precipitation totals in Alpine basins: IH-GAR, or APGD up to 2019; check ARCIS above 750 m.
  • Operational monitoring: ARCIS precipitation and temperature, updated monthly.
  • Daily forcing for a distributed model of a Po sub-basin: EMO-1 if you need many variables and recent years; our dataset if you want precipitation and temperature from one regional network with their uncertainty, with the caveats above; and in any case a check of the totals against IH-GAR at altitude.

Data availability and FAIR principles

Dataset Identifier and access Licence Station data
ARCIS arcis.it, no DOI check the site no
IH-GAR UNIMI Dataverse, doi:10.13130/RD_UNIMI/CSYYAP check the repository no
APGD MeteoSwiss, doi:10.18751/Climate/Griddata/APGD/1.0, registration non-commercial use no
Crespi et al. (2018) ISAC-CNR CC BY-NC-ND 4.0 no
Crespi et al. (2021) PANGAEA, doi:10.1594/PANGAEA.924502 CC BY 4.0 partly
EMO-1 JRC Data Catalogue CC BY 4.0 no
E-OBS, CERRA-Land Copernicus Climate Data Store open partly (E-OBS)
EEAR-Clim Zenodo, doi:10.5281/zenodo.10951609 CC BY 4.0 yes
Salehi et al. Zenodo, doi:10.5281/zenodo.19207256 CC BY 4.0 no (sources: Po District Authority, EEAR-Clim)

The weakest point is common to all regional grids: none releases the input station series, because they belong to the regional services. EEAR-Clim is the exception and a real step forward. A second gap is code: among these products, only EMO and ours (GEOframe Krigings, GPL-3) point to open interpolation code.

What is still missing

  • A daily product that brings in the elevation dependence of precipitation where it acts, on wet-day frequency, on intensity or both, with an explicit uncertainty for every day and cell.
  • Validation of daily fields at the daily scale, including wet-day frequency and extremes, for products built from monthly components.
  • Independent checks of precipitation at altitude, through water budgets closed on gauged basins with storage and evapotranspiration estimated independently. Snow undercatch cannot be seen by comparing gauges with gauges.
  • Open station data. Without it, every dataset is a black box that can only be compared with other black boxes.

References

  • Bongiovanni, G., et al. (2025). EEAR-Clim: a high-density observational dataset of daily precipitation and air temperature for the extended European Alpine region. Earth Syst. Sci. Data, 17, 1367–1391. doi:10.5194/essd-17-1367-2025
  • Crespi, A., Brunetti, M., Lentini, G., Maugeri, M. (2018). 1961–1990 high-resolution monthly precipitation climatologies for Italy. Int. J. Climatol., 38, 878–895. doi:10.1002/joc.5217
  • Crespi, A., Matiu, M., Bertoldi, G., Petitta, M., Zebisch, M. (2021). A high-resolution gridded dataset of daily temperature and precipitation records (1980–2018) for Trentino-South Tyrol. Earth Syst. Sci. Data, 13, 2801–2818. doi:10.5194/essd-13-2801-2021
  • Isotta, F. A., et al. (2014). The climate of daily precipitation in the Alps. Int. J. Climatol., 34, 1657–1675. doi:10.1002/joc.3794
  • Manara, V., et al. (2026). A new daily high-resolution gridded precipitation dataset for the Italian Greater Alpine Region (1951–2023). J. Hydrol.: Reg. Stud., 67, 103776. doi:10.1016/j.ejrh.2026.103776
  • Pavan, V., et al. (2019). High resolution climate precipitation analysis for north-central Italy, 1961–2015. Clim. Dyn., 52, 3435–3453. doi:10.1007/s00382-018-4337-6
  • Pavan, V., et al. (2026). A new operational dataset of gridded minimum and maximum temperature over north-central Italy 1991 to present. Climate Services, 43, 100689. doi:10.1016/j.cliser.2026.100689
  • Salamon, P., et al. EMO-1: an improved version of the high-resolution multi-variable gridded meteorological dataset for Europe. Earth Syst. Sci. Data Discuss., essd-2025-723
  • Salehi, H., et al. (2026). A 30-year 1-km daily precipitation and air temperature dataset for the Po River District (Italy). Earth Syst. Sci. Data Discuss. doi:10.5194/essd-2026-467
  • Thiemig, V., et al. (2022). EMO-5: a high-resolution multi-variable gridded meteorological dataset for Europe. Earth Syst. Sci. Data, 14, 3249–3272. doi:10.5194/essd-14-3249-2022
  • Copernicus Climate Data Store: E-OBS and CERRA-Land
  • ISPRA, BIGBANG national water balance: workshop presentation, March 2026

No comments:

Post a Comment