Improved observation of

real time events:

Satellite based precipitation

estimationEstelle de Coning

Chief Scientist: Nowcasting and very short range forecasting

1

Measurement of precipitation

• Rain gauges:

– provide a direct measurement of rainfall,

– networks are far too coarse to capture all the rainfall, especially at smaller scales

– unevenly distributed and,

– most importantly, they provide point source data and not a representation of a

spatial domain

• Radars rainfall:

– an indirect measurement of rainfall,

– radars need to cover the entire area of interest,

– Well correlated and

– have a good radar rainfall relationship.

• Satellite based estimates of rainfall:

– not as accurate as gauges or radar,

– major advantage is the high temporal resolution and coverage, even over

oceans, in mountainous regions and sparsely populated areas where rainfall is

not measured.

2

• Fourth meeting of the CBS-SWFDP Steering Group in Geneva,

February 2012:

– Challenge for the SWFDP: “the need for very short-range

forecasting tools, to address especially the rapid onset of

localized severe thunderstorms which can produce heavy

precipitation and strong wind, given the absence of adequate

real-time observational networks, especially weather radar

coverage.”

– The usefulness of EUMETSAT satellite based instability products,

such as the Global Instability Index, for nowcasting purposes was

recognized

– Also agreed that real time satellite rainfall estimates have proven

particularly useful in regions where rain gauges and radar coverage

is sparse.

• Surface-based precipitation measurement systems in DC and LDC still

needed to accurately measure and monitor precipitation amounts on

the ground, to be incorporated into hydrological runoff models as well

as aid in validation of other (satellite and model based) methodologies. 3

Sources of satellite QPE

• Satellite rainfall can be derived from low earth orbiting (LEO) or

geostationary satellites (GEO).

• LEO satellites:

– advantage of high spatial resolution,

– disadvantage is that the spatial coverage is in the form of narrow

swaths at times which are not always coinciding with

precipitating weather systems.

– not useful for operational forecasting purposes, although the

data can be utilized for other purposes, including data

assimilation for numerical weather prediction (NWP).

– the main advantage of LEO satellites is that the microwave

sensors carried by these low orbiting satellites have a more

direct link to precipitation than the instruments on board GEO

satellites.

4

• GEO satellites:

• are located much higher above the earth surface than

the LEO satellites and thus have a coarser resolution.

• major advantage that the data and images are

available in near real-time for the entire footprint of

the satellite.

• operational forecasting requires frequent updates and

near real-time availability. Forecasters who have to

monitor and forecast the movement and changes in

intensity of precipitating weather features, find GEO

most useful

5

The Hydroestimator

• The National Environmental Satellite, Data and Information Service

(NESDIS) developed an automated SPE algorithm for high-intensity

rainfall called the Autoestimator (AE) using the GOES satellite

(GEO)

• The original AE, developed by Vicente et al. (1998), computes rain

rates from 10.7 µm brightness temperatures based on the

comparison with radar rainfall

• The dependence of the initial AE on radar was a significant problem,

because one of the advertised strengths of satellite QPE

(Quantitative Precipitation Estimation) is its usefulness in regions for

which radar and/or rain gauge coverage is unavailable.

• Another version of the AE, called the Hydroestimator (HE) has been

developed which can be used outside of regions of radar coverage

without compromising accuracy.

6

Cold Cloud Tops and Rain -

Principle

200 250 290K

Tb=230 K

Tb=224 K

Tb=212 K

Tb=200 K

7

Exceptions to the Rule

200 250 290K

Ns:Tb=235 K

Cb:Tb=200 K

Ci:Tb=205 K

8

Hydroestimator

• Simple use of 10.8-µm brightness temperatures leads to missing of warm,

stratiform rain and incorrect designation of cold cirrus as raining clouds

• Improvements made to the HE considered the temperature relative to the

surrounding pixels:

– Pixels colder than their surroundings are assumed to be convective

updrafts and hence producing rainfall

– Pixels as warm as or warmer than their surroundings are presumed to be

convectively inactive

9

Hydroestimator

• Satellite imagery alone does not contain all the information needed for evaluating rainfall. Numerous processes occur below clouds e.g.:– evaporation of raindrops in dry air,

– reduction or increase by terrain features

• Improvements to the HE considers some correction figures taken from numerical weather prediction models:

– PW: rain is enhanced/reduced in high/low PW

– RH: reduces rain in dry lower atmosphere

– Actual level of neutral buoyancy is accounted for

– 700 hPa wind field is interleaved with the topography

10

Hydroestimator vs IR108 on 29 Jan 2013

11

Hydroestimator

IR108 colour enhanced

Global HE http://www.star.nesdis.noaa.gov/smcd/emb/ff/index.php

12

Local HE

• Code for HE was installed in SA in 2007,

using NWP input from Unified Model, run

in SA and MSG input (IR108)

13

Hydroestimator – operationally

available every 15 minutes

14

Applications of HE

• Tracking

• Accumulations

• Dissemination to SADC region via RSMC

website

• SAFFG

15

HE Tracking

• HE tracking done of the 10mm (~40 dBz

core)

• Available on RSMC website: http://rsmc.weathersa.co.za/RSMC/login.php

• Tracks storm direction for 30min and

60min forecasts.

• Growth and decay accounted for

16

HE Tracking

17

HE Tracking

18

Blue is current and red is predicted 30/60 minutes

19

HE accumulations• 1 hour, 3

hour, 6

hours, 24

hours

• 10 days

• 30 days

• NEW! 32

day

archive of

daily totals20

21

SARFFG for southern Africa

• WMO - similar flash flood guidance system over

seven southern African countries

• The SADC SARFFG system covers the rest of

South Africa and six other countries where there are

no radar coverage at the coarser 200km2 resolution.

• SARFFG depends primarily on satellite QPE as

precipitation input for modelling soil moisture and

flash flood guidance over large parts of southern

Africa.

• The Global Hydroestimator is used for this purpose

22

Example Tropical Cyclone

Dando 18 January 2012

24

Summary• Accurate satellite rainfall estimation is crucial for

good guidance in data sparse areas over Africa

• GEO satellite based QPE – great spatial and

temporal coverage

• Global and local Hydroestimator – available

every 15 minutes

• Local Hydroe - accumulation and tracking

products

• SARFFG – guidance to forecasters for flash

flood forecasting using Global Hydroestimator

and gauges (where possible)25

Shortcomings of the HE

• The HE works best for convective events,

but stratiform events (warm rain) might be

underestimated or missed.

• Convective rainfall intensity is often over

estimated

What if we had another

method?• Still relying on MSG (GEO satellite)

• Still good temporal and spatial

coverage….

27

• Satellite Application

Facilities (SAFs) are

centres for processing

satellite data

• The Nowcasting SAF

started in February

1997 aiming to produce

the software to deal

with the Nowcasting

and Very Short Range

Forecasting using the

characteristics of the

MSG SEVIRI data and

the NOAA and EPS

AVHRR data

(EUMETSAT

Satellites).

Nowcasting SAF products

website

29

Cloud Mask

Cloud type

Cloud top temperature and height

Precipitation Clouds

Convective Rainfall Rate

Precipitation with microphysics

TPW

Layer PW

RDT

Stability indices

High Resolution wind

Satellite Image interpretation

CRR – theoretical background

• Convective Rainfall Rate (CRR) product developed in the SAF NWC

context, is a Nowcasting tool that provides information on

convective, and stratiform associated to convection from MSG-

SEVIRI channels.

• CRR uses either 2 or 3 of the MSG SEVIRI channels:

– IR108 and (IR108 – WV062) (24hours)

– IR108, (IR108 – WV062) and VIS006 (day time only)

• The empirical relationship than the higher and thicker are the clouds

the higher is the probability of occurrence and the intensity of

precipitation is used in the CRR algorithm.

CRR theoretical background

• Information about cloud top height and about cloud

thickness can be obtained, respectively, from the

infrared brightness temperature (IR) and from the visible

reflectances (VIS)

• IR-WV brightness temperature difference is a useful

parameter for extracting deep convective cloud with

heavy rainfall. Negatives values of the IR-WV brightness

temperature difference have been shown to correspond

with convective cloud tops that are at or above the

tropopause

CRR theoretical background• To take into account the influence of environmental and

orographic effects on the precipitation distribution, some

corrections can be applied to the basic CRR value, based on

input from numerical weather prediction models (ECMWF):

– the moisture correction,

– the cloud top growth/decaying rates or evolution correction

– the cloud top temperature gradient correction

– the parallax correction

– the orographic correction

• At the end of the process CRR product produces information

on the instantaneous rain rate in mm/h in each pixel of the

image.

Convective Rainfall Rate Product v2012

Slide courtesy Cecilia Marcos, Nowcasting SAF product developer in Spain

6 Oct 2014 0600-1100 UTC

34

Example: Daily rainfall CRR vs Hydroe vs

raingauges for 23 Feb 2010

Hydroestimator

Using 1 MSG channel

CRR using 3 MSG channels

Rain gauges

35

Summary of 14 summer cases

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

9.00

10.00

RMSE MAE BIAS 1mm BIAS 10mm BIAS 20mm

RMSE, MAE and BIAS for 1, 10 and 20mm for all 14 cases

HE CRR

Better Bias scores for 1, 10 and 20mm

Smaller errors

36

CRR daily total for 30 Oct

Example: Daily rainfall CRR and Hydroe vs TRMM

for 2 Jan 2014

Hydroestimator

Using 1 MSG channelCRR using 3 MSG channels

HE overestimates

0

5

10

15

20

25

30

Case 1 Case 2 Case 3 Case 4 Case 5 Case 6 Case 7 Case 8 Case 9 Case 10

RMSE and MAE for all 10 cases

HE RMSE

CRR RMSE

HE MAE

CRR MAE

International Precipitation

Working Group (IPWG)• IPWG objectives:

– Set standards for satellite precipitation measurements

and

– verification of the products

– Data exchange

– Stimulate international research and development

• IPWG validates and verifies precipitation products in

standardized format

• Allows product developers and users to gain insight into

satellite observations for precipitation estimations

• NWP models and satellite algorithms compared

against surface reference data sets derived from

rain gauges and/or radar observations

• 0.25°grid inter-comparisons

• Data comparison results available on internet

• Displays include daily images of the products

and validations including scatterplots and

various statistics.

http://rsmc.weathersa.co.za/IPWG/ipwgsa_qlooks.html

Gauges NWP Satellite QPE GEO Sat

HE in IPWG

44

CRR since 15 Sep 2014 also

part of the IPWG

45

CRR summary

• CRR (three MSG channels) compares better to

daily rainfall totals from rain gauges than the

Hydroestimator (single MSG channel).

• RMSE, MAE and Bias scores of the CRR were

better then those of the Hydroestimator for all cases

considered.

• CRR has been part of IPWG since 15 Sep and

during this time has been performing better than the

HE on several occasions.

• Due to the lack of adequate rain gauges over data

sparse regions such as Africa, CRR can provide

good estimates of rainfall in real time.

46

Back to 16 October 2014…CRR SADC from 0600 – 2345 UTC

47

CRR for SA from 0600 – 2345 UTC

48

Hydroestimator 0600 – 2300 UTC

(hourly)

49

Daily totals for 16 Oct

50

51

Overestimation?

IPWG comparison

52

Hydroe:

POD 0.615

FAR 0.143

HSS 0.167

RMSE 13.4

Correlation 0.596

IPWG comparison

53

CRR:

POD 0.558

FAR 0.174

HSS 0.265

RMSE 8.0

Correlation 0.671

Summary• GEO satellite QPE can provide good

estimates of rainfall intensity in real time

• Satellite QPE are especially useful where

we don’t have rain gauges (or radar

rainfall)

• CRR is now operational for SADC region

and seems to be outperforming Hydroe

• Future work – Continuous testing/perhaps

replacing Hydroe with CRR in some

places/products54