Oct 12, 2010

Oct 12, 2010Hydrologic Early Warning System

for East Africa

Ashutosh Limaye, John Gitau, Eric Kabuchanga

CRAM WorkshopSeptember 26, 2011

• Data and Models• Online Maps• Visualizations• Decision Support• Training• Partnerships

Mapping Fires in Guatemala Mexico

Training and Capacity Building

Flood Forecasting in Africa

SERVIR

Strengthen the capacity of governments and other key stakeholders to integrate Earth observations into development decision-making

SERVIR Network

SERVIR @ CATHALACCity of Knowledge, Panama

Inaugurated on February 3, 2005

SERVIR-Africa @ RCMRDNairobi, Kenya

Inaugurated on November 21, 2008

SERVIR-Himalaya @ ICIMODKathmandu, Nepal

Inaugurated onOctober 5, 2010

SERVIR Applications have several dependencies:

• NASA Applied Science Program Agriculture, air quality, climate, disasters, biodiversity, public health, water resources

• GEOAgriculture, biodiversity, climate, disaster, ecosystems, and human health

• USAIDClimate change adaptation, carbon tracking and GEO focus areas

• Regional Needs Assessment

SERVIR Applications

• Spatially distributed hydrologic model CREST, developed by University of Oklahoma (based on Variable Infiltration Capacity (VIC) model)

• Uses near real-time satellite rainfall estimates from TRMM and forecasts from Kenya Meteorological Department (KMD) to produce soil moisture, evapotranspiration & streamflow

CREST model

SERVIR Hydrologic Modeling

KMD East African Domain

• Spatial extent of CREST runs match the KMD domain (~2800 x 3000 km)

• Spatial resolution: 1km

• KMD temperature and rainfall forecasts (QPF), available hourly at 14km spatial resolution, to provide boundary conditions.

• Forecasted soil moisture, evapotranspiration and streamflow will enable KMD to issue early flood warning, especially in the flood prone watersheds in western Kenya.

• KMD intends to use the modeled fields to initialize the next model run

SERVIR Hydrologic Forecasting

48-hr KMD QPFSept 20, 2011 18z

• Last week, we completed the 10-year CREST model run with the available TRMM data. We are calibrating CREST model using observations at Nzoia River in Kenya.

• We plan to use that calibration for the entire region. Needless to say, we welcome additional observational data to make the model results more robust.

• Using the 10-yr CREST model run, we have generated a streamflow history for each 1km pixel.

• We are using that historic data to assess 5th, 20th, 80th and 95th percentiles for each pixel Based on the four quantiles, we can assess whether the near real time model output falls within one of five categories:

Five Quantile Categories• Very Wet• Wet• Normal• Dry• Very Dry

Providing Historic Data Perspective in Near Real Time CREST Model Outputs

Providing Historic Data Perspective in Near Real Time CREST Model Outputs

• Additionally, we have made Land Information System (LIS) reanalysis runs with Princeton land surface forcings.

• The Princeton forcings go back to 1949. In next two months, we plan to use the 10 years of TRMM data to bias-correct the resampled dataset form 1949. It will extend our historic range to over 60 years.

• Together, the 10-years of TRMM data, and 60-years of Princeton data will provide the historic perspective to contextualize the near real time model estimates and to quantify hydrologic extremes including floods and drought.

Incorporating Seasonal Outlookfrom ICPAC or IRI

• Ensembles of seasonal forecasts need to be factored in the hydrologic predictions.

• Historic reanalysis allows us to assess the “normal”, “above” and “below” conditions.

• Expect to produce the hydrologic forecasts with the seasonal forecasts (target: March 2012).

SERVIR-East Africa Products• Near Real Time Hydrologic Datasets

– Streamflow – Soil moisture– Quantiles of Streamflow, Soil Moisture

• Short Term Forecasts using KMD QPF– Rainfall– Streamflow– Soil moisture

48-hr Streamflowbased on KMD QPFSept 20, 2011 18z

SERVIR Web Portal

CREST User Tool

•Enables time querying of maps

•Enables time querying of time series

•Users can download time series data

•Users can extract time series data for specific sites

•Supports OGC standards (WMS, WMS-T)

•Enables extraction of pixel values based on date

•Continuous enhancements and updates being carried out

SERVIR One-Stop Web Portal

Geospatial Catalog Interactive Web Maps

SERVIR Web Portal

Hydrologic Modeling for KMD, Kenya Dept. of Water Resources and Beyond

• KMD has indicated that this hydrologic modeling information will be useful for their monthly Weather Outlook bulletin.

• Kenya Department of Water Resources would like to tailor our hydrologic modeling tools to their specific interests of producing three-monthly forecasts.

• We are seeking additional government and non-governmental groups, including FEWS NET, in sharing our near real time, short term as well as seasonal forecasts.

In a Nutshell…

SERVIR-East Africa is running an operational hydrologic model using near real time NASA satellite data sets and Kenya Meteorological Department forecasts.

In next few months, we plan to include seasonal forecasts in our hydrologic modeling. We anticipate those products to become available on our website (www.servirglobal.net) by the beginning of next year.

We are committed to making our products useful to governmental and non-governmental agencies for their decision making.

Thank you

Ashutosh LimayeSERVIR Science Applications Lead

[email protected]