Real Time Nowcasting In The Western Us OR

Real Time Nowcasting In The Western Us OR Why you can’t use nodes C 0 -2 George Thomas Andy Wood Dennis Lettenmaier Department of Civil and Environmental Engineering LAND SURFACE HYDROLOGY RESEARCH GROUP Group Seminar July 5, 2006

Outline q Objective q background q surface water monitor q westwide forecast system q challenges q index station method for real-time forcing generation q implementation details q progress / results to date q future plans

Objective q Monitor the hydrologic state of the US land surface q Initially, western US; soon, mexico + remainder of US q 1/8 degree q Daily updates in real-time (lag of 12 -36 hrs) q Soil moisture, snow, runoff q Fully automated q Consistent with retrospective: back to at least 1950 q Can be used for hazard assessment q Drought q Flood risk q Can be used to initialize hydrologic forecasts q Short lead, 1 -15 day q Long lead, 1 -12 month

Background q This nowcast draws upon procedures and data from two existing systems assembled by Dr. Wood q UW Experimental Surface Water Monitor (1/2 degree) Ø Methods for real-time forcing generation & model updating q UW West-wide Seasonal Hydrologic Forecasting System Ø VIC model data at 1/8 degree Ø Eventually, forecast methods The nowcast will eventually be an integral part of the Westwide forecasting system

SW Monitor Background q An outgrowth of the west-wide forecasting system that adds a national scale perspective on land surface moisture q directly relevant to retrospective drought reconstruction work going on in our group q Andreadis et al. (2005) paper on drought q ½ degree VIC input parameters q enabled by recent NCDC extension of digital data archives back to 1915 q will be used as platform for drought and hydrologic analyses in real-time q nowcasts are used now by US Drought Monitor & US Drought Outlook authors (at CPC and elsewhere) q many products possible, such as following one:

Drought Severity and Spatial Extent

Monitor Webpage daily updates 1 -2 day lag soil moisture & SWE percentiles ½ degree resolution archive from 1915 -current uses ~2130 index stns

Background: UW SW Monitor trends: 1 week 2 week 1 month Archive!

Background: UW SW Monitor Archive from 1915 -current conditions are a product of the same simulation (same methods, ~same stations) as historical conditions allows comparison of current conditions with historical ones can navigate by month or year People: Andy, Ali, Kaiyuan, Dennis

Background: UW SW Monitor

Background: West-wide Forecasting System Snowpack Initial Condition Soil Moisture Initial Condition

Background: West-wide Forecasting System NEW: West-wide overview of flow forecasts (mouse-over/clickable for more details)

Background: West-wide Forecasting System Now clicking the stream flow forecast map also accesses current basinaveraged conditions As previously, flow location maps give access to monthly hydrograph plots, and also to data.

Background West-wide Forecasting System An earlier G. Thomas contribution: Automating plots of west-wide SWE data Daily Updating West-at-a-glance SWE from NRCS, EC, CADWR Analyses: q Current Anomalies q Percentiles: Ø Current Ø 1 -week change Ø 2 -week change

Background: Central Challenge # stations q Model simulations are calibrated and validated using a uniform or consistent set of forcing data q Nowcasts and forecasts use models calibrated and validated retrospectively q Problem: the station data used to create forcings are not as widely available in real-time as they are for the retrospective calibration/validation period time 3 months before present q Solution: the “index-station method” present

VIC model spinup methods: index stations estimating spin-up period inputs dense station network for model calibration sparse station network in real-time

Outline q Objective q background q surface water monitor q westwide forecast system q challenges q index station method for real-time forcing generation q implementation details q progress / results to date q future plans

Index station method: example for precipitation q uses time-varying precipitation signal ONLY FROM stations that report reliably in real-time and for over 45 years (many go back longer) q precipitation percentiles calculated from raw precip for time period no shorter than 21 days. q percentiles interpolated to 1/8 degree grid q at 1/8 degree, percentiles used to extract corresponding observed value from 1/8 degree restrospective distribution (based on dense observing network, standard VIC forcing methods) q period 1/8 degree precip amount disaggregated using the fractional daily precipitation for that period (interpolated to 1/8 degree grid). q temperature is treated differently – daily interpolated anomalies for Tmin & Tmax are used

Index station method: example for precipitation Index stn pcp (mm) pcp percentile gridded to 1/8 degree monthly 1/8 degree pcp (mm) 1/8 degree dense station monthly pcp DISTRIBUTION (N years for each 1/8 degree grid cell) (MM) daily disagg. to daily using interpolated daily fractions from index stations

Index station method: example for precipitation q In real-time, with daily updates, this method actively updates the forcings for a period from 3 weeks to 7 weeks. case 1: current day is less than day 21 of months treated as 1 period for percentile calculation case 2: current day is greater than day 20 of months treated as 2 periods for percentile calculation the first month becomes fixed in forcing data

Index station method q test of method for streamflow

Outline q Objective q background q surface water monitor q westwide forecast system q challenges q index station method for real-time forcing generation q implementation details q progress / results to date q future plans

Nowcast Information Flow 1930 s 1955+ NOAA ACIS / Other Prcp Tmax Tmin Coop Stations Index Station Method Gridded Forcing Creation VIC Retrospective Simulation Daily, 1915 to Near Current Hydrologic values, anom’s, %-iles w. r. t. retrospective PDF climatology (PDF) of hydrologic values w. r. t. defined period VIC Real-time Hydrologic Simulation State (~1 month long) Hydrologic State (-1 Day) vals, anoms %-iles w. r. t. PDF

Implementation Details Computing Environment: Flood Cluster 46 cores 9 AMD Opteron 2 x dual core 2 Intel Xeon 2 X single core 1 AMD Opteron 2 x single core Rocks 4. 0. 0 / Cent. OS 4. 0 Linux Nowcasting is implemented on nodes c 0 -2 and c 0 -6 Useful phrases: WTF? ! (what the flood? !) RTFM!! (read the flood manual)

Implementation Details SW Monitor coding scheme download obs P, tx, tn update Station Index files climatology Forcings 2 mon Params Soil, etc. force VIC output

Implementation Details SW Monitor coding scheme – NCAST implementation Node 0 -2 Node 0 -6 download obs P, tx, tn update Loop over basins ca, colo, gbas, riog Station Index files climatology Station Index files Loop over basins pnw, mexn, mexs force Forcings 2 mon Params Soil, etc. VIC Forcings 2 mon VIC output

Implementation Details Observation data: Real-time and retrospective stations. 3 Primary sources

Implementation Details Observation data: Real-time and retrospective stations. 3 Primary sources 1. ACIS for CONUS 2123 stations 1915 -present

Implementation Details Observation data: Real-time and retrospective stations. 3 Primary sources 1. ACIS for CONUS 2123 stations 1915 -present 2. Environment Canada 10 stations 1915 -present

Implementation Details Observation data: Real-time and retrospective stations. 3 Primary sources 1. ACIS for CONUS 2123 stations 1915 -present 2. Environment Canada 10 stations 1915 -present

Implementation Details Observation data: Real-time and retrospective stations. 3 Primary sources 1. ACIS for CONUS 2123 stations 1915 -present 2. Environment Canada 10 stations 1915 -present 3. Mexico (retrospective) 739 stations 1925 -2003

Implementation Details Observation data: Real-time and retrospective stations. 3 Primary sources 1. ACIS for CONUS 2123 stations 1915 -present 2. Environment Canada 10 stations 1915 -present 3. Mexico (retrospective) 739 stations 1925 -2003 Real-Time: EDAS (Eta DAS) daily re-analysis

Outline q Objective q background q surface water monitor q westwide forecast system q challenges q index station method for real-time forcing generation q implementation details q progress / results to date q future plans

Results daily forecast of SM percentiles

Results daily forecast of SM percentiles – animations of recent forecasts

Results daily forecast of SM percentiles - comparison with SW Monitor

Results daily forecast of SM percentiles - comparison with CPC Drought Monitor

Results daily forecast of SM percentiles - 2 week change - comparison with SW Monitor

Results April 1 SWE Archive (1997 – 2006)

Outline q Objective q background q surface water monitor q westwide forecast system q challenges q index station method for real-time forcing generation q implementation details q progress / results to date q future plans

Future Work Ongoing and Future Work q data products q expansion (Arkansas, etc. ) q routing q constraints using SWE

END Thank you!