PosterPDF Available

Global-Local Interactions Modulate Tropical Moisture Exports to the Ohio River Basin



Regional-scale extreme rainfall and flooding are temporally and spatially associated with the occurrence of tropical moisture exports (TMEs) in the Ohio River Basin (ORB). TMEs are related to but not synonymous with atmospheric rivers, which refer to specific filiamentary organizational processes. TMEs to the ORB may be driven by strong, persistent ridging over the Eastern United States and troughing over the Central United States, creating favorable conditions for southerly flow and moisture transport from the Gulf of Mexico and Caribbean Sea. However, the strong inter-annual variation in TME activity over the ORB suggests dependence on global-scale features of the atmospheric circulation. We suggest that this synoptic dipole pattern may be viewed as the passage of one or more high-wavenumber, transient Rossby waves. We build a multi-level hierarchical Bayesian model in which the probability distribution of TME entering the ORB is a function of the phase and amplitude of the traveling waves. In turn, the joint distribution of the phase and amplitude of this wave is modulated by hemispheric-scale features of the atmospheric and oceanic circulation, and the amplitude and synchronization of quasi-stationary Rossby waves with wavenumber 1-4. Our approach bridges information about different features of the atmospheric circulation which inform the predictability of TME at multiple time scales and develops existing understanding of the atmospheric drivers of TMEs beyond existing composite and EOF studies.
Abstract ID: 129624
Final Paper Number:
Global-Local Interactions Modulate Tropical Moisture Export to the Ohio River Basin
James Doss-Gollin1,2David Farnham1,2Upmanu Lall1,2
1Columbia Water Center 2Department of Earth and Environmental Engineering, Columbia University
Conceptual Framework
Figure 1: NASA images of Eastern U.S. (L) April 2010; (R) May 2011
Floods associated with persistent extreme regional
rainfall may be determined by cross-scale atmospheric
circulation dynamics that link persistent global surface
conditions as well as regional interactions with these
evolving patterns. For example, the April 2011 flooding
in the Ohio-Mississippi river system (figs. 1 and 2) was
driven by steering of multiple storms into the region
along highly similar trajectories.
Boundary Forcings
Circulation / Planetary
Synoptic Circulation
Moisture Transport &
Flood Potential
Figure 2: (L) Conceptual diagram. (R) Tracked cyclones (lines) and
strongest 20% monthly-mean 250 hPa winds for April 2011. Region
affected circled in blue.
Research Questions & Data
1. What synoptic weather patterns lead to intense
moisture transport into the Ohio River Basin?
2. How do planetary-scale circulations modulate these
synoptic weather patterns?
3. What are the conditional probabilities of moisture
transport into the region, given synoptic and planetary
circulation indices?
4. How can forecasts of low-frequency modes of circulation
inform risk of extreme, regional flooding?
We use the Ohio River Basin (fig. 4) to anchor a case
study, focusing on the DJF season during which the
well-known 1937 flood occurred. We use 6-hourly
reanalysis data from ERA-Interim [1] (1979-2015) and
cyclone tracks generated by Donna Lee [2] using the
TRACK algorithm [3].
Study Area & Variable
Figure 3: Composite anomalies of 850 hPa geopotential height (color)
and precipitable water (contours) from 4 days preceding 113 regional
intense precipitation days in the Ohio River Basin to one day
following. Regional intense precipitation days defined as at least 10%
of GHCN rain gauges exceeding 99th percentile. From [4].
Study of regional intense precipitation events in the Ohio
River Basin (fig. 3; [4]) reveals the importance of a ridge
(somtimes stationary, sometimes transient) over the
Western Atlantic and a transient cyclone propagating
eastward. We define indices for the Eastern U.S. low and
Western Atlantic high to capture the geopotential height
gradient over the region, representing dynamical drivers
(see [4, 5] for details), and an index of SST anomalies
over Gulf of Mexico [see 6, and references therein] to
represent thermodynamic drivers:
Moisture the daily net moisture flux into Ohio River
Basin region (fig. 4, green box)
Planetary 30-day moving average of the PNA and NAO
indices from the CPC
Synoptic Mean daily SST anomalies over the Gulf of
Mexico (fig. 4, blue); and the 850 hPa the
geopotential height anomalies over the West Atlantic
(purple) and the Eastern United States (red).
W. Atl. RidgeEast U.S. Low
−90 −80 −70 −60
Figure 4: Study Area. Shaded area shows the Ohio River Basin.
Boxes show locations of indices described above.
Synoptic Patterns
Inspection of cyclone track locations and the associated
moisture flux into the region shows that moisture flux
into the region is maximized in the presence of a cyclone
to the north-west of the region. Figure 5 shows this in
terms of cyclone tracks: tracks associated with
extremely high flux (R) follow highly similar trajectories
from the south-west to north-east (SWNE). This
pathway is much less frequently observed among
randomly sampled tracks (L).
High Flux
−110 −100 −90 −80 −70 −60 −50 −110 −100 −90 −80 −70 −60 −50
Figure 5: Tracks of extratropical cyclones associated with (Left) 70
random DJF days, and (Right) the 70 days with highest DJF
moisture transport. Color indicates cyclone intensity. Moisture box
of fig. 4 is highlighted.
Global-Scale Modulation
A primary mechanism by which particular modes of the
global-scale circulation, such as the PNA, modulate
moisture flux into the mid-latitudes is by shifting the
preferred pathways of extratropical cyclones. Figure 6
shows the cyclone tracks over the study area for the
negative and positive PNA terciles. Relative to the
negative phase, the positive PNA shifts the SWNE
tracks to the east and favors a NWSE trajectory over
the Ohio River Basin.
−110 −100 −90 −80 −70 −60 −50 −110 −100 −90 −80 −70 −60 −50
Figure 6: Cyclone tracks given PNA in (L) negative; (R) positive
terciles). Moisture box of fig. 4 is highlighted.
Probabalistic Prediction
To estimate the probabilities of moisture flux conditional
on synoptic features, and of synoptic features conditional
on planetary-scale features, we build a two-level
Bayesian model. This model takes three data blocks:
1. yN×1, in this case the moisture flux
2. XN×J, “local” factors: here X1is GMX SST anomaly,
X2is W. Atl. Ridge, and X3is Eastern U.S. Low
3. ZN×k, “global” factors: here Z1is PNA and Z2is NAO
Then, the model is formulated as
yt N β0+X0
tβ, σ2for t= 1, . . . , T (1)
Xt,j N α0,j +Z0
tαj, τ2
jfor j= 1, . . . , J (2)
To estimate the parameters β0, β, σ, α0, α, τ we use
Hamiltonian Monte Carlo Markov Chain sampling using
the probabilistic programming language Stan [7].
Model Inferences
0 1
Posterior Parameter Distributions
The X variables: high, low, sst. The Z variables: pna, nao
Figure 7: 4000 parameter draws from the posterior distribution
specified by eqs. (1) and (2).
Interpreting the results of fig. 7 requires noting the
ordering of the Xand Zvariables; for example,
alpha[1,3] gives the coefficient of Z1(PNA) on X3
(GMX SSTs). In the first (local-moisture) step, we note
that the effect of the dynamical Xvariables (the low
and high height anomalies; β1, β2) is far greater than the
effect of the thermodynamic variable (the GMX SSTs;
β3). This comparison is valid because all Xand Zhave
been standardized. It is also of note that the magnitude
of the effect of increasing the W. Atlantic High (β1) is
somewhat larger than increasing the magnigude of the
Eastern U.S. Low (β2). In the second (global-local) step,
we see that a positive PNA provides feedbacks that both
support and inhibit moisture flux to the Ohio River
Basin, suppressing the high and GMX SSTs (α1,1, α1,3)
but enhancing the low (α1,2) in the positive phase. The
NAO also provides both moisture-inhibiting and
moisture-enhancing effects (α2,1:3).
Summary of Findings
1. A finite set of specific synoptic circulation patterns
accounts for most DJF moisture transport into the
Ohio River Basin
2. Results suggest that the W. Atlantic high, particularly
when it is persistent (blocking), may be more relevant
to DJF Ohio River Basin moisture transport than the
transient low or source region temperature
3. The PNA and NAO each provide positive and negative
feedbacks on moisture transport to the Ohio River
Basin. Modeling these intermediate feedbacks explicity
improves model propagation of opposing
feedbacks and variance as compared to modeling
moisture transport directly on the global-scale indices.
Next Steps & Discussion
IPosterior checking (not shown) reveals that modeling all
interactions as normal (eqs. (1) and (2)) is an overly
strong assumption; specific distributions for specific
interactions may improve prediction
IExplore additional parameters and interactions
IConsideration of GMX SST anomalies as a local feature
addresses thermodynamic impact on moisture
availability, but is not independent of large-scale
circulations; likely not a one-directional causality
IModel dependence of global features explicity using
copulas [i.e. 8]
[1] D. P. Dee et al. Q.J.R. Meteorol. Soc. (2011).
[2] J. F. Booth et al. J. Appl. Meteor. Climatol. (2015).
[3] K. I. Hodges. Mon. Wea. Rev. (1994).
[4] D. J. Farnham, J. Doss-Gollin, and U. Lall. “Space-Time Characteristics and
Statistical Predictability of Extreme Sub-Weekly Precipitation Events in the
Ohio River Basin”. AGU Fall Meeting. 2016.
[5] J. Nakamura et al. J. Hydrometeorol. (2012).
[6] S. Steinschneider and U. Lall. J. Hydrometeorol. (2016).
[7] B. Carpenter et al. J. Stat. Softw. (2016).
[8] C. Genest et al. Water Resour. Res. (2007).
Thanks to Donna Lee for sharing cyclone tracks.
Thanks to Yochanan Kushnir, Jimmy Booth, Scott
Steinschneider, Pierre Gentine, Casey Brown, Linda
Mearns, Katherine Schlef, Melissa Bukovsky, Rachel
McCrary, and Seth McGinnis for insightful
conversations. This project is funded by SERDP grant
2516 “Climate Informed Estimation of Hydrologic
Extremes for Robust Adaptation to Non-Stationary
For a digital version of this poster & more information, scan QR code (top right) or visit
ResearchGate has not been able to resolve any citations for this publication.
  • D P Dee
D. P. Dee et al. Q.J.R. Meteorol. Soc. (2011).
  • J F Booth
J. F. Booth et al. J. Appl. Meteor. Climatol. (2015).
  • K I Hodges
K. I. Hodges. Mon. Wea. Rev. (1994).
Space-Time Characteristics and Statistical Predictability of Extreme Sub-Weekly Precipitation Events in the Ohio River Basin
  • D J Farnham
  • J Doss-Gollin
  • U Lall
D. J. Farnham, J. Doss-Gollin, and U. Lall. "Space-Time Characteristics and Statistical Predictability of Extreme Sub-Weekly Precipitation Events in the Ohio River Basin". AGU Fall Meeting. 2016.
  • J Nakamura
J. Nakamura et al. J. Hydrometeorol. (2012).
  • S Steinschneider
S. Steinschneider and U. Lall. J. Hydrometeorol. (2016).
  • B Carpenter
B. Carpenter et al. J. Stat. Softw. (2016).
  • C Genest
C. Genest et al. Water Resour. Res. (2007).