Precipitation observations (from CETEMPS dataset)

• Region: Italy
• Total of ~3700 stations
• Average of 1800 active stations: 1 station every 13x13km
• Max 2600 at one time (2012): 1 station every 11x11 km
• Min 500    at one time (2001): 1 station every 25x25 km
• Time resolution: hourly data (!!!) in mm/h
• ~2001-present, updating
• No useful metadata

Elevation characteristics

TIME

NUMBER OF STATIONS

Nespor and Sevruk, 1999

Undercatch

>30% ?

Common problems with

in-situ measurements

Temporal and spatial problems:

• Short timescale
• Missing periods
• Low station density
• Missing timesteps

Data quality problems:

• Breaks and inhomogeneities
• Manual measurement errors
• Equipment errors and failures
• Weather-related measurement errors

VERY FREQUENT OUTLIERS

The question is, as always:

HOW TO REMOVE OUTLIERS WITHOUT REMOVING HIGH PRECIPITATION EXTREMES?

FIRST-STAGE FILTERING

FLAGGING

FLAG CHECKING

CLEANED DATASET!

Possible first-stage filtering procedures

• Removal of extreme values > 200 mm/h
• Removal of isolated > 100 mm/h reports
• Removal of complete months with > 1800 mm (2.5 mm/h)
• Removal of continuous identical values
• ...?

Rauthe et al. 2013; A Central European precipitation climatology–Part I: Generation and validation of a high-resolution gridded daily data set (HYRAS)

Isotta et al. 2013; The climate of daily precipitation in the Alps: development and analysis of a high‐resolution grid dataset from pan‐Alpine rain‐gauge data

Perry et al., 2009; The generation of daily gridded datasets of temperature and rainfall for the UK

Hiebl et al., 2017; Daily precipitation grids for Austria since 1961—development and evaluation of a spatial dataset for hydroclimatic monitoring and modelling

Possible flagging procedures

• Mean + n*SD threshold
• Median + n*IQR threshold
• Peaks in the values distribution
• Isolated dry/wet event flagging
• Low correlation of close stations
• ...?

Rauthe et al. 2013; A Central European precipitation climatology–Part I: Generation and validation of a high-resolution gridded daily data set (HYRAS)

Isotta et al. 2013; The climate of daily precipitation in the Alps: development and analysis of a high‐resolution grid dataset from pan‐Alpine rain‐gauge data

Perry et al., 2009; The generation of daily gridded datasets of temperature and rainfall for the UK

Hiebl et al., 2017; Daily precipitation grids for Austria since 1961—development and evaluation of a spatial dataset for hydroclimatic monitoring and modelling

Possible flag-checking procedures

• Visual comparison of maps/videos
• Visual comparison of close timeseries
• Comparison with daily datasets (EURO4M-APGD, E-OBS, ...)
• Comparison with hourly datasets (PERSIANN, ...)
• ...?

Rauthe et al. 2013; A Central European precipitation climatology–Part I: Generation and validation of a high-resolution gridded daily data set (HYRAS)

Isotta et al. 2013; The climate of daily precipitation in the Alps: development and analysis of a high‐resolution grid dataset from pan‐Alpine rain‐gauge data

Perry et al., 2009; The generation of daily gridded datasets of temperature and rainfall for the UK

Hiebl et al., 2017; Daily precipitation grids for Austria since 1961—development and evaluation of a spatial dataset for hydroclimatic monitoring and modelling

Results so far...

(by applying first-stage filtering procedures only)

FILTERED

ORIGINAL

Monthly averages

However... flag checking madness: I still have ~60k flags to deal with!

At one flagged event per minute, that's 4 months of continuous, alienating work =

So I need either:

• Better first-stage filtering
• Faster flag checking

CETEMPS Hydrological Model

CHyM is a distributed (gridded) hydrological model from CETEMPS and University of L'Aquila.

• Model working on all test domains
• We can reproduce past results (Coppola et al. 2013) on our test domain (western Po basin)
• HR DEMs working (at low model res)

CETEMPS Hydrological Model

Model peculiarities:

• Can build DEM from various sources, smoothing by cellula automata algorithms
• Can use several kind of inputs, such as station observations, gridded model data, etc.
• Designed to work on any domain
• NetCDF output

CETEMPS Hydrological Model

CHyM-OP reproduced domains:

CETEMPS Hydrological Model

CHyM-OP reproduced domains:

Statistics

How to estimate hundred-years floods with only a few (~20) years worth of data?

The methodology is taken from Maione et al., 2003

LISFLOOD-FP hydraulic model

Widely-used flood inundation model from the University of Bristol (Bates et al., 2010)

SDH:

observed

CHyM

Real stations

CHyM stations

NO FLOOD!