Progress in drought monitoring and prediction Dennis P. Lettenmaier Department of Geography University of California, Los Angeles International Top-Level Forum on Engineering Science and Technology Development Strategy Nanjing Hydraulic Research Institute May 29, 2015 Droughts as a natural hazard

Economic losses ($B) and loss of life in recent U.S. droughts Source: NOAA Billion-dollar weather/climate disasters

Drought characteristics “creeping disaster” – most evolve slowly over time Typically cover large areas Hence losses can be large Definitions vary (meteorological, agricultural, hydrological)

Outline of this talk Characterizing droughts Drought predictability Reconstructing drought Flash Drought Concluding thoughts

1. Characterizing droughts

Severity-Area-Duration Analysis Depth-Area-Duration technique, widely used in probable maximum precipitation analysis Replace depth with measure of drought severity S : severity, ΣP : cumulative percentile of soil moisture (runoff), t : event duration from Andreadis et al., 2005

Soil Moisture S-A-D Curves (a) 3 Month Duration (b) 12 Month Duration (c) 24 Month Duration (d) 48 Month Duration 1 0.65 7 Severity Severity Area (million sq. km) Area (million sq. km) Severity Severity Area (million sq. km) Area (million sq. km) 1928-32 1932-38 1950-57 1960-67 1975-79 1987-93 1998-03

http://www. hydro. washington http://www.hydro.washington.edu/forecast/monitor/archive/htm/arch_grid.htm

2. Drought predictability

Long-term climatology Simulations now Monitoring Long-term climatology long-year spinup model climatology recent future model spin-up Real-time data forecast Forecast Products streamflow soil moisture runoff snowpack important point(s): after bias correcting and downscaling the climate model forecasts, the procedure for producing hydrologic forecasts is as follows: we spin up the hydrologic model to the start of the forecast using observed met. data (from 2 sources: NCDC cooperator stations through 3-4 months before the start of the forecasts, then LDAS 1/8 degree gridded forcings thereafter). The GSM forecasts comprise 2 sets of ensembles, one for climatology and one for the forecast. The climatology ensemble yields a distribution of the conditions we’ve seen over the period 1979-1999, while the forecast ensemble yields the distribution of the conditions we might see for the next 6 months. Although the climatology ensemble is nominally unbiased against a simulated climatology based on observed met. data (rather than bias-corrected, downscaled GSM met. forcings), we compare the forecast and GSM climatology so that any unforeseen biases (resulting, perhaps, from the downscaling method) occur in both climatology and forecast. Eventually this cautionary step may be eliminated, and we’ll compare directly to the simulated observed climatology. at the end of the spin-up period and one month before month 1 (out of 6) of the forecasts, we save the hydrologic model state. The state is then used for initializing the forecast runs. Through the first month, the model runs on observed data to the last date possible, then switches to the forecast data. Usually, we process the observed forcings up through the 15th to 25th of this initialization month, then the forecast forcing data carries the run forward for the remaining days in the month, and throughout the following 6 month forecast period. Note, the state files used for the climatology runs correspond to the spin-up associated with the particular year (out of 1979-1999) from which the climatology ensemble member is drawn. the spin-up period captures the antecedent land surface hydrologic conditions for the forecast period: in the Columbia basin, the primary field of interest is snow water equivalent. forecast products are spatial (distributed soil moisture, runoff, snowpack (swe), etc.), and spatial runoff + baseflow is routed to produce streamflow at specific points, the inflow nodes for a management model, perhaps. A

Experiment 1: Ensemble Streamflow Experiment 2: Reverse-ESP Approach: Experimental setup for partitioning the role of IHCs and FS in seasonal hydrologic forecasts Experiment 1: Ensemble Streamflow Prediction (ESP) Experiment 2: Reverse-ESP (revESP) Known IHC, Climatological forcings. Derives skill solely from knowledge of the IHCs. Known forcings (i.e. perfect climate forecasts skill) and climatological IHCs. Derives skill solely from climate forecasts. Wood and Lettenmaier, 2008

Experiment 1: Ensemble Streamflow Experiment 2: Reverse-ESP Approach: Experimental setup for partitioning the role of IHCs and FS in seasonal hydrologic forecasts Experiment 1: Ensemble Streamflow Prediction (ESP) Experiment 2: Reverse-ESP (revESP) Known IHC, Climatological forcings. Derives skill solely from knowledge of the IHCs. Known forcings (i.e. perfect climate forecasts skill) and climatological IHCs. Derives skill solely from climate forecasts. Wood and Lettenmaier, 2008

Variation of RMSE ratio for cumulative runoff forecast Climate forecast skill dominates IHCs dominate Shukla and Lettenmaier, 2011, HESS

Shukla and Lettenmaier, 2011, HESS. 1 2 3 5 4 6 Plot of the maximum lead (in months) at which RMSE Ratio [RMSE(ESP)/RMSE(revESP)] is less than 1, for cumulative runoff forecasts, initialized on the beginning of each month.

Soil moisture forecast (a) Initialization: 01 January (b) Initialization: 01 July If you prefer showing a spatial map like this: feel free to use this. I have similar figures for SWE and CR predictability as well. Climate FS dominates IHCs dominate RMSE ratio of mean monthly soil moisture forecast at lead-1, -3 and -6 months for the forecast initialized on (a) 01 January and (b) 01 July. (masked areas had RMSE resp and esp =0)

ESP (Ensemble streamflow prediction) vs NMME_VIC Fcsts Starting date ESP- P T inputs taken from randomly selected observations Run VIC with observed P and Tsurf Jan 1,1915 from UW Jan 1, 1979 IC s Fcst forward Feb 5 Feb 6--- NMME_VIC :forcings were taken from error corrected T P from CGCM Both ESP and NMME_VIC have the same initial conditions, but ESP has no climate forecast information of P and Tsurf

Fcst skill for SM Lead-1 : correlation >0.8 (WOW!!!) Lead-3: Over the western interior dry region, the fcsts are still skillful for all seasons and the North Central for January (high skill regions) Low skill regions are circled

Differences btw NMME-ESP for Lead-3 No significant differences for Lead-1 and Lead-2 Only October and January forecasts pass the Livezey Chen test Differences are in the areas that the skill is low and dynamically active areas Oct fcsts are helped by skillful P forecasts

3. Reconstructing drought

China Meteorology Administration (Wang 2005)

Motivation Frequency agricultural drought occurrence in China during past decades. Widely used, but link to direct observations (e.g., of soil moisture) is weak – hence reliance on indirect methods, such as PDSI. Need for reproducible basis for identifying drought-affected regions. Land surface model representations of soil moisture offer an alternative means for estimating severity, frequency, duration, and variability of current droughts, and linking them to the climatology of observed droughts. An ensemble approach to help reduce the uncertainty due to the bias of single model for reproducing soil moisture.

Models VIC: Variable Infiltration Capacity Model (Liang et al. 1994) CLM3.5: Community Land Model version 3.5 (Oleson et al. 2007) NOAH LSM: NCEP, OSU, Air Force, Hydrol. research lab (Mitchell et al. 1994, Chen and Mitchell 1996) CLM-VIC: a hybrid of CLM 3.5 with the VIC soil hydrology scheme (Wang et al. 2008).

13-month moving averages of normalized soil moisture indices from the ensemble average Top panel: Location of regions (“China” refers to the entire domain). Right: 13-month moving averages of normalized soil moisture indices from the ensemble average (dark lines) and range of individual models (gray shading). Inter-model variations are generally greatest in arid areas, and least in humid areas.

Up:22% Dn:37% Up:2.5% Dn:2.3% Up:6.1% Up:6.5% Dn:1.1% Annual trends in a) soil moisture percentile; b) drought severity; c) drought duration; and d) drought frequency for 1950-2006. The construction of time series of drought duration, severity and frequency are described in section 4e). The blue “+” indicates increasing trends, and the red “-” indicates decreasing trends.

Can we do longer reconstructions (to ~1870; including major Chinese famines of late 19th and early 20th centuries) using 20 C Reanalysis? Visual courtesy Aihai Wang, CAS/IAP

4. Flash drought

Two types of flash droughts 1. Heat wave flash drought— Initialized by heat waves High Tair => High positive ET anomalies=> Low SM% P plays important but indirect role 2. P deficit flash drought Initialized by P deficits P deficits => Low SM% => Negative ET anomalies=> high sensible heat=> High Tair P plays a direct role

Frequency of occurrence – heat wave flash drought 1. FOC is only 4-6% of the total record 2. Maxima over the North Central and the Ohio Valley and the PNW 3. SAC has the least events

Downward trends on the occurrence of the heat wave flash drought

P deficit flash drought-example P deficits increase ET<0 and increase Tair increases

Frequency for the P deficit flash drought Larger % than the heat wave flash drought A band of maxima are located over the southern U.S with the center at the southern Plains Less events over the maxima of the FOC for the heat wave flash drought but nor mutually exclusive

5. Concluding thoughts Drought monitoring systems based on land surface models are well developed, and serve as reasonable surrogates for hydrologic and agricultural droughts Drought prediction capabilities are limited by weak predictive capability of global atmospheric models – most of the predictive skill is in hydrologic initial conditions Drought monitoring systems have demonstrated ability to reconstruct extreme historic droughts over the period of instrumental record. For the pre-instrumental period, capability has yet to be confirmed. Flash drought identification and predictability is currently evolving. Given the rapid evolution of these events (over time periods ~1 week or so) for which there may be precipitation forecast skill, some predictability may be possible.