Download presentation
Presentation is loading. Please wait.
Published byCharity Clark Modified over 9 years ago
1
Approaches to Seasonal Drought Prediction Bradfield Lyon CONAGUA Workshop 24-26 Nov, 2014 Mexico City, Mexico
2
Drought Prediction What do we want to predict?
3
- Precipitation (timescale? monthly, seasonal, annual...?) - Soil Moisture (how deep a layer?) - Stream flow / Inflow - Groundwater Level - Impacts It depends on specific decisions: The best “Drought Index” is the one that is most closely associated with the specific outcome/impact of interest. A generalized drought prediction system needs to forecast several indicators, which ultimately need to be related to specific variables of interest (inflow, soil moisture, crop yield, etc.). Drought Prediction What do we want to predict?
4
Sources of Predictive Skill Sea Surface Temperatures a) Tropical Pacific (El Niño, La Niña) b) Tropical Atlantic Seager et al. 2009 May to October
5
Sources of Predictive Skill Sea Surface Temperatures Climate Model * Skill in Seasonal Rainfall Predictions Correlation (Fcst, Obs) 1982-2010 * North American Multi-Model Ensemble (NMME, 6 climate models) Jan-MarApr-Jun Jul-SepOct-Dec
6
Sources of Predictive Skill The “Initial Condition” The July NADM is a good “first guess” of the October NADM…
7
Sources of Predictive Skill The “Initial Condition” There is often month-to-month persistence in drought indicators that can provide predictive information. Consider the Standardized Precipitation Index (SPI). The SPI compares accumulated, precipitation to historical values, expressing differences as a normal distribution. SPI6(Jun) SPI6(Jul) Jan Feb Mar Apr May Jun 5 of the 6 months are in common large persistence JUL Feb Mar Apr May Jun To make a forecast of SPI6 one month ahead, the picture looks like this:
8
Number of months with lagged correlation > 0.6 for the 12-month SPI Sources of Predictive Skill The “Initial Condition” Lyon et al. 2012, JAMC
9
Sources of Predictive Skill The “Initial Condition” Yaqui Water System Inflow data courtesy of José Luis Minjares Accumulated inflow in March a potential predictor annual inflow… INFLOW ( x10^6 m^3)
10
Sources of Predictive Skill The “Initial Condition” Use accumulated inflow in March to predict yearly inflow Inflows to the Yaqui System
11
Which Indicator is Best? Yaqui Water System Inflow Departure from Average: Comparison with SPI-12 Water years 1965-2007 Yaqui Water System Inflow data courtesy of José Luis Minjares r = 0.7 The one most relevant for a specific use
12
* * * Model Soil Moisture vs. Various SPI Indicators Example from the Eastern US (1950-200) Layer 3 Layer 2 Layer 1 “VIC” Land Surface Model www.hydro.washington.eduVIC soil moisture data courtesy of Justin Sheffield, Princeton University Correlation “VIC” Soil Moisture and SPI Which Indicator is Best?
13
Inflows to the Yaqui System Ideally, predictions of specific outcomes are desired: reservoir inflow, crop yield, rangeland biomass, etc. However, more general drought indicators can be linked to specific outcomes. This provides a calibration of the index to something more relevant to the user… Tailored Forecasts
14
Drought & Agricultural Impacts in Sri Lanka (1960 – 2000) Lyon et al., 2009, JAMC 40 yrs. of agricultural impacts data available at the district level. Which meteorological drought indicator is most closely associated with drought impacts to agriculture? Tailored Forecasts
15
1. Examine drought indicators and impact occurrences2. Consider seasonality of drought & impacts 3. Quantify relationships between drought indictors and impacts. Key for development of early warning systems Lyon et al., JAMC, 2009 Tailored Forecasts
16
GCM Fcst PRCP, Wind Statistical Model Historical Inflows IRI Seasonal Fcst Pr(Below-Normal Rainfall) Tailored Forecasts
17
As input to a reservoir management tool… Tailored Forecasts
18
2-Mo. Lead Fcst for end of June 2010 2-Mo. Lead Fcst for end of June 2011 Low Risk High Towards a Water Sector Impact Forecast for Mexico Index of Water Sector Vulnerability Drought Index, Water Impact Relationship: Identify Thresholds Issued April 2010 Issued April 2011 Drought Index Forecast Probabilistic Water Supply Impact Forecast [ V + Pr(< threshold) ] [ 1+ Pr(< threshold) ] R = 0 ≤ V ≤ 1, 0 ≤ R ≤ 1 With Carolina Neri, UNAM
19
Drought Index Forecast Prob. SPI6 < -1 Issued in April Low Risk High For Jun 2011 = Probabilistic Water Impact Risk Forecast Issued in April For Jun 2010 Obs Jun 2010 Observed SPI6 in June DryWetDryWet + Water Vulnerability For Jun 2011 Obs Jun 2011 For Jun 2010
20
Available Today Forecasts of 3, 6, 9 and 12-month SPI Dec Prob. SPI12 < threshold Dec SPI Best Estimate Dec SPI12 10% probability Dec SPI12 Best Estimate Interactive: User selects Index, Thresholds, Probabilities of interest…
21
Summary Droughts are not simply unpredictable, random events. There is identifiable skill in seasonal forecasts of several meteorological drought indicators (and other variables). Skill is typically greatest in fall and winter, least in summer. Ultimately, we are interested in the likelihood of drought impacts, not just forecasts of drought indicators. Thus, there is a need to calibrate drought indicators to impacts in some fashion. Generation of drought risk forecasts will first require a vulnerability assessment of a system to drought.
22
Acknowledgements This work has been supported in part by the Modeling, Analysis, Predictions and Projections (MAPP) program at NOAA, which is gratefully acknowledged. References Lyon, B., M. A. Bell, M. K. Tippett, A. Kumar, M. P. Hoerling, X. Quan, H. Wang, 2012: Baseline probabilities for the seasonal prediction of meteorological drought. J. Appl. Meteor. Climatol., 51, 1222-1237. Lyon, B., L. Zubair, V. Ralapanawe, and Z. Yahiya, 2009: Finescale Evaluation of Drought in a Tropical Setting: Case Study in Sri Lanka. J. Appl. Meteor. Climatol., 48, 77–88. US-Mexico SPI Forecast and Monitoring Products from IRI http://iridl.ldeo.columbia.edu/maproom/Global/Drought/N_America/index.html ----------------------------------------------------------------------------------------------------------------------------------------------
23
Longer Time Scale Variations (mm/mo.) Annual Average Rainfall
24
Longer Time Scale Variations A Simple Separation of Time Scales The majority of the variation in rainfall is from one year to the next…
25
Longer Time Scale Variations
26
Seasonality of Precipitation
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.