1. Altai flood’s 2014 and model precipitationg forecasts
Author Zdereva Marina , Doctor of Geographical Sciences,
* Siberian Regional Hydrometeorological Research Institute
City Sovetskaya street 30 , Novosibirsk , Russian Federation

2. The potential of using model precipitation forecasts to predict floods on the example of the south of Western Siberia
Zdereva M.*, Sannikova S.*, Khluchina N.*

3. In late May 2014, the highland south of Western Siberia become a disaster area. Wave cyclones actively moving along the periphery of the anticyclone from the west to the east caused heavy rains which occurred in the period of intensive snow melting in the mountains. This process resulted in a sharp water level rise up to critical elevations in rivers of the Altai Republic and the Altai Territory. Maximum levels were observed in the period May 27 - June 1, 2014.

4. Dozens of settlements were within the flooding zone

5. Was it possible use model precipitation forecasts to foresee flooding?
The problem of prediction flash floods will not be solved without meteorological forecasts, especially in case when the rise of water in the river channel occurs within a few minutes after the start of precipitation. There are very sparse meteorological stations and hydrological observation stations in the Altai high-mountain, and so it’s impossible to get reliable statistical decisions for spatial-time arrangement of precipitation.
Was it possible use model precipitation forecasts to foresee flooding?

6. Base models: Computing Center COSMO SLAV
SGI Altix 4700
West Siberian
Computing Center
(Novosibirsk))
Base models:
Semi-Lagrangian
Absolute Vorticity
global model
COSMO
step=14kmP
non-hydrostatic
mesomodel
SLAV
step=50km
Technological complexes COSMO and SLAV are introduced in the Computer SGI Altix 4700 of the West Siberian Computing Center (Novosibirsk) with well-run initial and boundary data aquisition system, launch of calculation units on models, and resulting output in map, text, and GRIB formats.
The spatial step in the modification of COSMO being run in the West Siberian Computing Center (SOSMO-SIB14) is 14 km or degrees.
The step in the SLAV model is about 50 km.

7. Models and actual precipitation from May 24, 2014
date
hour+
stations
Кызыл-Озек
Шебалино
Чемал
Турочак
Яйлю
Усть-Кан
Усть-Кокса
Онгудай
Катанда
Кош-Агач
Кара-Тюрек
24 mai
fact 12
slav 1 2 3 4
cosmo 24 9 15 13 8 14 6 10 20 17 5 36 7 48 60 27 21 72 22 11 33
Model precipitation forecasts showed a surprisingly good results for areas with complicated topography. For every station modeled precipitation were presented the quantity from nearest gridbox.
The COSMO generated a forecast with heavy rainfall warning already at May, 24 with maximum for this model lead time of 72 hours, that is, for May, 27.
Precipitation forecast for Chemal was 33mm (actual quantity 17 mm), for Yaylyu - 24 mm (33 mm), and for Kyzyl-Ozek — 10 mm (24 mm), respectively.

8. Models and actual precipitation from May 27, 2014
date
hour+
stations
Кызыл-Озек
Шебалино
Чемал
Турочак
Яйлю
Усть-Кан
Усть-Кокса
Онгудай
Катанда
Кош-Агач
Кара-Тюрек
27 mai
fact 12 7 5 20 3 9 2 1 16
slav 6 10 4 13
cosmo 19 8 24 36 11 14 35 48 22 31 15 23 18 17 26 21 27 60 28 72
Further predictions were confirmed with a relatively low rate of "false alarms" at individual stations.

9. Dynamic actual and modeled precipitation for 24h in the Katun river basin
Results of two models were different. This is evidenced by dynamic actual and modeled precipitation sums for the Katun river basin.

10. The Actual-Forecast contingency table and estimations of POD(%) and FOH(%):
Event – precipitation ≥ 10 мм
Forecast
Actual
All
No. «yes»
No. «no» a
(hits «yes») b
(false alarm)
a+b
(all forecasts «yes») c
(gaps) d
(hits «no»)
c+d
(all forecasts «no»)
a+c
(all fact «yes»)
b+d
(all fact «no»)
ALL:
N=a+b+c+d
PОD= a/(a+b) *100
FOH=a/(a+c) *100

11. The variant of coupling: maximum from models forecasts
The flood forecasting requires accounting for water catchment in river basins. Analysis of precipitation forecast for Biya and Katun river basins showed an even greater coincidence with actual values
The efficiency of using precipitation forecasts to predict floods may be improved with application of simple coupling algorithm: a maximum of modeled precipitation (complex)
Estimations to 72 h (Katun river basin )

12. Estimation of modeled precipitation for Biya river basin
Estimation POD%
Estimation FOH%

13. Let's take a look at the behavior of the modeled precipitation for high-mountain stations at the example of 2015 summer with the addition of NCEP and UKMO models. Representations of output fields have different scales. Estimations (POD and FOH) were executed for every station per nearest grid box, then were averaged all round.

14. Estimation of modeled precipitation for high-mountaing stations of Altai Republic. June-July 2015.
Estimation POD%
Estimation FOH%
POD+FOH%
COSMO and NCEP are most similar in the quality of forecasts. SLAV model overestimates precipitation, so it has the highest warning estimation. In this case the forecast is improved by averaging of model-simulated precipitation.

15. Conclusions :
Behavior of modeled precipitation from different hydrodynamical approaches differs in space and in forecast advancement. The cyclonic or frontal precipitation are the better forecasting.
In general, model forecasts of precipitation in riverheads may be used in hydrological schemes for forecasting rainfall floods. This requires researches for optimize the interpolation of expected precipitation on river basin areas.
Improvement of the quality of hydrodynamic precipitation forecasts may be achieved by coupling of the results of different models. Coupling algorithm may be either the average or the maximum value.