Using the NWC SAF products over South Africa

Using the NWC SAF products over South Africa and southern Africa to enhance nowcasting capabilities in data sparse regions E de Coning, Morne Gijben, Louis van Hemert, Bathobile Maseko South African Weather Service

Content • Background • CRR examples for SA and SADC • CRR vs Hydroestimator vs rain gauge data for Oct-Dec 2014 • RDT examples for SA and SADC • RDT validated against lightning data for Dec ’ 14 and Jan ‘ 15 • Conclusions and future work

Background • Nowcasting is needed to warn the public of severe weather events • Data sparse regions – very few lightning/radar networks • Satellite and NWP blended products (WMO SWFDP) • In SA: using local version of UKMO Unified Model (12 km) • NWCSAF v 2012 tested with case studies over SA/SADC • NWCSAF v 2013 installed operationally in SA during 2014 • Southern African domain since Oct 2014. • CRR and RDT, no other products yet

CRR example: Daily rainfall CRR vs Hydroe vs raingauges for 23 Feb 2010 (using v 2012) Rain gauges Hydroestimator Using 1 MSG channel CRR using 3 MSG channels 4

CRR example: Daily rainfall CRR and Hydroe vs TRMM for 2 Jan 2014 (using v 2012) Hydroestimator Using 1 MSG channel CRR using 3 MSG channels

CRR validation against daily rainfall totals over South Africa (using v 2013) • • • International Preciptiation Working Group (IPWG) Different rainfall algorithms (NWP and satellite) 0. 25° resolution Daily rain gauges (24 hour) totals Hydroestimator (1 MSG channel) and CRR (3 MSG channels) added to this suite of products

RMSE and Correlation Coefficient (v 2013)

CRR validation thus far… • Encouraging • Better than Hydroestimator in Oct-Dec 2014 • Will continue monitoring…

RDT example 9 Nov 2012 (v 2012) 1130 UTC

RDT example 28 Dec 2013 at 1200 UTC Zimbabwe (v 2012)

RDT example 20 September 2014 – Afternoon thunderstorms and hail in Jh. B (v 2013) Operational! v 2013

RDT validation against lightning data over South Africa (v 2012) – No lighting data input – Bufr to hdf 5 – Growing and mature phases used for validation – Validated against SALDN – CG lightning only. 10 min before and after each RDT time step – Object-oriented methodology (R, Spatival. Vx)

9 Oct 2012 v 2012 Graphs supplied by Morne Gijben R statistical software used with Spatial. Vx package for O-O validation developed at NCAR

RDT validation against lightning data over South Africa (v 2013) – No lighting data input – No OT detection input – Bufr 3 output since Dec 2013 – Growing, mature and decaying phases used for validation – Validated against SALDN – CG lightning only. 10 min before and after each RDT time step – Between 1100 and 1800 UTC (13: 00 and 20: 00 SA local time) – Object-oriented methodology (R, Spatival. Vx)

23 Dec 2014 v 2013

Validation for Dec 2014 and Jan 2015 (v 2013)

Conclusions and Future work • • Encouraging results especially for data sparse regions! Submission to national journal accepted for publication Add lightning (and OT? ) as input data (this year) Validate against radar (where available) over SA New HPC, latest upgrades of UM (this year) – 4 km, 1. 5 km Funded research project 2015 -2017 NWCSAF V 2015?