Ad ancing Wa e shed-Scale Hyd odynamic Modeling:
Techniques o Compa a i e Analysis o Sain -Venan Sol e s
Edwa d Tie nan, Cheng-Wei Yu, Saz Sha io , Ben R. Hodges
No embe 2025
Objec i e
This echnical epo p esen s he me hodological ounda ion and supplemen a y analyses associa ed wi h
a s udy e alua ing he pe o mance o a newly de eloped Sain -Venan Equa ion (SVE) sol e , SWMM5+,
in compa ison wi h wo exis ing hyd aulic models: he U.S. EPA S o m Wa e Managemen Model (EPA
SWMM) and he Simula ion P og am o Ri e NeTwo ks (SPRNT). The compa a i e modeling amewo k,
in oduced in he accompanying jou nal a icle i led “Ad ancing Wa e shed-Scale Hyd odynamic Modeling
wi h he Sain -Venan Equa ions”, is applied o egional-scale i e ne wo ks o assess nume ical beha io ,
solu ion s abili y, and hyd odynamic accu acy ac oss di e en sol e a chi ec u es. This epo p o ides ad-
di ional echnical documen a ion, implemen a ion de ails, and supplemen al igu es ha suppo he indings
and discussions p esen ed in he main manusc ip .
In his s udy, SWMM5+ is alida ed h ough model- o-model compa ison a he han di ec compa ison
wi h obse a ional da a, due o he subs an ial challenges inhe en in alida ing la ge-scale i e ne wo k
models. Obse a ional low and s age da a a e o en spa ially spa se, empo ally inconsis en , and subjec
o measu emen unce ain ies, which complica e he assessmen o how model e o s eme ge and p opaga e
h oughou he sys em. Addi ionally, i e ba hyme ic da a—c ucial o sol ing he Sain -Venan Equa-
ions—a e equen ly incomple e o o unce ain accu acy ac oss la ge domains, making i di icul o isola e
algo i hmic pe o mance om bounda y condi ion unce ain ies. Compounding hese challenges is he p es-
ence o hyd aulic in as uc u e, such as dams and wei s, whose ope a ional ules a e o en p op ie a y o
undocumen ed. These unmodeled an h opogenic con ols can signi ican ly in luence low and s age obse -
a ions, he eby obscu ing he a ibu ion o model-obse a ion disc epancies (Tu ne , S eyae , Condon, &
Voisin, 2021). Consequen ly, alida ion in such se ings ends o e lec he deg ee o pa ame e calib a ion
a he han he in insic accu acy o he nume ical schemes. Model- o-model compa ison hus p o ides a
mo e anspa en means o e alua ing he pe o mance o nume ical sol e s, independen o he con ounding
e ec s o obse a ional and s uc u al unce ain ies.
In con as o adi ional model-obse a ion compa isons, model- o-model compa ison p o ides a consis en
and sys em-wide amewo k o e alua ing nume ical accu acy and s abili y unde iden ical o cing condi-
ions, bounda y speci ica ions, and pa ame e se ings. This app oach allows pe o mance o be assessed
ac oss a wide spa ial domain, o e coming he limi a ions associa ed wi h spa se obse a ional da a. Fu -
he mo e, i enables de ailed examina ion o hyd odynamic beha io s—such as wa e a enua ion, nume ical
dissipa ion, and sol e s abili y—ac oss complex, wa e shed-scale ne wo ks. By isola ing he e ec s o nu-
me ical a chi ec u e, model- o-model compa ison o e s c i ical insigh s in o he s eng hs and limi a ions o
each sol e , independen o calib a ion a i ac s o obse a ional unce ain ies.
This echnical documen a ion ou lines he me hodological componen s ha suppo such model in e -
compa ison s udies and may se e as a e e ence o simila e alua ions. The key opics co e ed include:
1. De elopmen o egional-scale i e ne wo k models using publicly a ailable da ase s.
2. C i e ia and p ocedu es o e alua ing SWMM5+ sui abili y o model- o-model compa ison.
3. Ini ializa ion p ocedu es o Sain -Venan Equa ion (SVE) simula ions, including model spin-up e-
qui emen s.
1
4. Pos -p ocessing wo k lows designed o enable obus compa ison o spa ially and empo ally dis ibu ed
esul s.
5. P esen a ion o supplemen al simula ion esul s om he SVE-based models.
6. Sol e adjus men s and con igu a ion se ings applied o ensu e ai and in e p e able compa isons
ac oss models.
1 C ea ion o i e ne wo k models
1.1 Regional-scale Ri e Basins
The egional-scale i e basins examined in he main jou nal a icle a e he La aca and San Jacin o Ri e
basins, bo h loca ed wi hin he Texas-Gul hyd ologic egion. The La aca Ri e o igina es in cen al Texas
and d ains an a ea o app oxima ely 5,900 km2be o e discha ging in o La aca Bay. The ull i e ne wo k
consis s o 3,049 km o channel eaches; howe e , o s eamline he model while p ese ing hyd odynamic
ideli y, i s -o de ibu a ies—which ha e negligible in luence on he o e all low dynamics—a e ep esen ed
as nodal in lows a he han explici ly modeled channels. This simpli ica ion educes he e ec i e modeled
channel leng h o 1,291 km. The La aca Ri e basin ea u es a ela i ely mild opog aphic g adien , wi h
ele a ions anging om 0 o 23 m abo e sea le el. Na u al channel slopes all be ween 0.0001 and 0.002,
wi h an a e age slope o app oxima ely 0.0003. The ne wo k s uc u e and opological con igu a ion o he
La aca Ri e a e illus a ed in Figu e 1(a).
In con as , he San Jacin o Ri e —which d ains an a ea o app oxima ely 4,600km2, exhibi s a mo e complex
b anching opology and g ea e ulne abili y o ex eme lood e en s, as e idenced du ing Hu icane Ha ey
in 2017. Applying he same channel il e ing me hodology as used o he La aca Ri e , he o al modeled
channel leng h is educed o 1,716 km a e excluding i s -o de ibu a ies wi h limi ed hyd aulic in luence.
Loca ed wi hin he Eas Texas Plain, he San Jacin o Ri e basin spans ele a ions om 0 o 120 m abo e
sea le el. Excluding localized slope dis up ions caused by hyd aulic s uc u es, he na u al channel slope
anges om 0.0001 o 0.009, wi h an a e age slope o 0.00045. The s uc u e and opological con igu a ion
o he i e ne wo k a e depic ed in Figu e 1(b).
2
Fig. 1. The wo selec ed i e ne wo ks wi hin he Texas-Gul egion in his s udy. (a) La aca Ri e , (b)
San Jacin o Ri e . The blue lowlines a e ex ac ed om he NHDPlus da ase (USGS, 2020). Gold s a s
(1) and (2) a e hyd og aph loca ions in Fig. 5; gold s a s (3) and (4) a e hyd og aph loca ions in Fig. 7.
1.2 F om Raw Da a o SPRNT
One-dimensional hyd ological models o he La aca and San Jacin o Ri e s we e de i ed om he USGS
Na ional Hyd og aphy Da ase (NHDplus) lowlines o Hyd ological Uni Code 12 (USGS, 2020). De ini ion
o channel geome y and oughness ollowed he app oach used in he Na ional Wa e Model (NWM), which
de ines apezoidal c oss-sec ions wi h cons an side-wall slope o 0.5 and a bo om wid h ha depends on
he S ahle ’s s eam o de o he each (Da id e al., 2011; NOAA, 2016). The bo om wid h anges be ween
7mand 40 m. The oughness om NWM (S ickle -Manning, n) anges om 0.045 o 0.06. Subc i ical
low a he downs eam bounda y is ensu ed by ixing he wa e dep h (H) o 3 m. Hyd ological lows
h oughou he i e basins we e ex ac ed om he NWM a chi e and in oduced as ime- a ying nodal
bounda y condi ions (NOAA, 2016). Following he app oach by Fal e e al. (2016) and Pai a, Collischonn,
Bonne , and de Gon¸cal es (2012), i s o de and mino s eams (i.e. eaches ha a e d y a p eponde ance
o he simula ion ime) we e emo ed. Flow con ibu ions om he excised eaches, when hey occu , a e
inco po a ed in o he hyd ological in lows a neighbo ing nodal bounda y condi ions. O he ne wo k e ine-
men s eps we e aken, including he omission o 86 “b aided” channel eaches, as well as he eplacemen
o 106 ese oi s/dams wi h simple channels. Such ea u es should be handled by sophis ica ed i e - ou ing
so wa e a emp ing o make eal wo ld p edic ions. Howe e , o ou p esen scope o compa ing di e en
models’ nume ical me hods o iden ical simula ion pa ame e s, he addi ional complexi y con ibu ed by
b aided channels o dam-ga e ope a ions is no wa an ed.
Gene ally, a o able beha io om hyd aulic models can be expec ed when channel segmen s a e o oughly
equal leng h. F om he NHDplus da ase , he a e age i e each leng h o ou es case was 3.1 km, wi h
a maximum o 68 km and a minimum o 1 m. The i e sys ems he ein we e e ined such ha each i e
channel segmen was ∼150 mby subdi iding longe segmen s and leng hening sho e ones. Table 1 con ains
de ails abou he disc e iza ion o he La aca and San Jacin o NHDplus ne wo ks in o he NWM o m.
3
Table 1. Topological in o ma ion o majo i e ne wo ks, excluding i s -o de eaches, in he Texas-Gul
wa e shed (Yu e al., 2021).
Ne wo k
Name
Channel
Leng h (km)
NHDPlus
Flowlines
Compu a ional
Nodes
In low
Bounda ies
La aca 1291 360 6973 67
San Jacin o 1732 690 9659 116
The La aca and San Jacin o Ri e ne wo k models we e o iginally buil as inpu iles o he Simula ion
P og am o Ri e NeTwo ks (SPRNT). As SPRNT iles, hese sys ems we e used by Yu, Liu, and Hodges
(2017) o s udy consis en ini ial condi ions o he Sain -Venan equa ions in i e ne wo k modeling, by
Yu e al. (2021) o e alua e a new me hod o de ec ing ins abili y-causing geome y ansi ions.
1.3 F om SPRNT o SWMM
The SPRNT modeling a chi ec u e a ies om he EPA SWMM modeling a chi ec u e in c i ical ways. To
enable a cohe en compa ison o he esul s be ween he modeling pa adigms, he models, o iginally con-
s uc ed in he SPRNT a chi ec u e, mus be con e ed in o he EPA SWMM a chi ec u e. Fo una ely, he
SWMM5+ model also used he EPA SWMM a chi ec u e, so he La aca and San Jacin o models only need
o be con e ed once.
All SVE sol e s ely on he same physical pa ame e s, e.g., ele a ion g adien s, c oss-sec ion geome y, e c.,
bu how hese pa ame e s a e handled can a y be ween modeling en i onmen s. Con e ing om one mod-
eling en i onmen o ano he is an exe cise in inding he analogous da a s uc u es o each pa ame e . Some
SPRNT objec s, such as “segmen s”, a e di ec ly analogous o SWMM “condui s”; p ope ies like leng h and
ups eam/downs eam connec ions a e simply copied om SPRNT in o he SWMM o ma . O he ne wo k
objec s equi e da a s uc u e ans o ma ion. An example o a ne wo k objec equi ing ans o ma ion
a e mul i-nodal junc ions in SPRNT being collapsed in o single node junc ions in he SWMM pa adigm
(shown wi h a ed ∆ in Fig. 2). In SPRNT, a “junc ion” consis s o mul iple “nodes” in a complex, whe e
each node can only ha e one o wo connec ions. Whe eas, in SWMM, any node wi h one o mo e con-
nec ions is a conside ed a junc ion. Me ging junc ions om SPRNT in o SWMM in ol es he eplacemen
o SPRNT junc ion complexes wi h SWMM junc ion nodes (illus a ed in Fig. 3, below). O he ne wo k
objec pa ame e s (e.g. channel geome y, in e nal bounda y condi ions, e c.) a e simply s o ed in di e en
da a s uc u es be ween SPRNT and SWMM. The ans o ma ion om SPRNT-compa ible inpu iles o
SWMM-compa ible inpu iles, conse es ne wo k opology, geome y, and o cing da a (i.e. he ne wo k
in o ma ion in Table 1 is p ese ed).
4
Fig. 2. Schema ic showing he da a s uc u e shu ling needed o con e om SPRNT- o ma o SWMM-
o ma . No e he changes in ne wo k objec ja gon be ween SPRNT and SWMM. Red ∆ indica es Junc ion
complex u he illus a ed in Fig. 3
Fig. 3. Example schema ic o “Junc ion” complex o SPRNT, SWMM5+, and EPA SWMM. S a e a iables
like low (Q) and head (h) a e epo ed as a ibu es o di e en ne wo k objec s depending on he modeling
amewo k.
The SWMM5+ model u he ans o ms he condui -node EPA SWMM inpu ile in o a ini e- olume
ne wo k o elemen s and aces, while s ill p ese ing he ne wo k p ope ies. Fo addi ional in o ma ion
5
abou he disc e iza ion app oach used by SWMM5+, please see Hodges e al. (2024).
2 E alua ion o SWMM5+
In he s udy, i e ne wo k simula ion esul s om mul iple SVE sol e s a e compa ed o assess he e ec s
o p og am a chi ec u e on lood wa e peak iming and o he hyd odynamic ac o s. The SPRNT p og am
was exp essly designed o i e ne wo k simula ion (Liu, 2014), while p e ious s udies ha e es ablished
EPA SWMM as a sui able model o some i e ou ing simula ion (Niazi e al., 2020; Swa hi, Raju, &
Va ma, 2020; Xiong & Melching, 2005). The SWMM5+ p og am, he newes o hese SVE-sol e models,
was designed as an unde - he-hood, pa allelized, mass-conse a i e e sion o EPA SWMM5 (Hodges e al.,
2024). Howe e , SWMM5+ has no been explici ly es ed on a egional-scale i e ne wo k p io o i s use
he ein. This sec ion b ie ly desc ibes some ele an a chi ec u al de ails abou SWMM5+, hen compa es
simula ion esul s o obse ed da a a wo loca ions wi hin he La aca i e basin.
2.1 Pa alleliza ion
The pa alleliza ion s a egy in SWMM5+ adop s he single p og am, mul iple da a (SPMD) app oach.
Unlike implici me hods commonly used in SVE sol e s, which ypically equi e ma ix decomposi ion, he
explici nume ical scheme in SWMM5+ enables di ec subdi ision and dis ibu ion o he compu a ional
domain ac oss mul iple p ocesso s. The ne wo k subdi ision algo i hm om Tie nan and Hodges (2022)
is adap ed o op imize load balancing ac oss p ocesso s. Ins ead o con en ional mul ip ocessing lib a ies
(e.g., OpenMP), SWMM5+ employs Coa ay Fo an o pa alleliza ion (Hodges e al., 2021). Al hough
he p ima y ocus o his s udy is no on pa allel compu a ion, we include his in o ma ion o highligh ha
SWMM5+ suppo s e icien pa alleliza ion, demons a ing i s capabili y o handle la ge-scale simula ions.
2.2 Nume ical Me hod
The nume ical amewo k o SWMM5+ employs an explici second-o de Runge-Ku a (RK2) ime-ma ching
scheme o compu e low a e (Q) and olume (V) wi hin ini e- olume elemen s. While he RK2 algo i hm
exhibi s highe empo al disc e iza ion disc epancies compa ed o highe -o de ime-ma ching me hods, i s
s uc u e is pa icula ly well-sui ed o pa allel compu ing a chi ec u es. To educe nume ical dependency be-
ween adjacen compu a ional elemen s, SWMM5+ implemen s a no-neighbo disc e iza ion app oach wi hin
he RK2 scheme (Hodges & Liu, 2019), e ec i ely minimizing communica ion bandwid h equi emen s du ing
pa allel compu a ion. The ime-space disc e iza ion s uc u e o he RK2 algo i hm is depic ed in Figu e 4.
6
Fig. 4. The uppe panel illus a es he domain o dependency o an RK2 ime ad ancemen in a ini e-
olume scheme wi h hi d-o de upwind ace econs uc ion, while he lowe panel depic s a no-neighbo
complian scheme wi h second-o de ace econs uc ion. The hicke a ows in he uppe panel highligh
he inc eased spa ial dependency in oduced by he hi d-o de scheme. Figu e adap ed om (Hodges &
Liu, 2019).
As illus a ed in Figu e 4, he ini e- olume elemen aces a e econs uc ed dynamically du ing ime in eg a-
ion. T adi ional app oaches ypically ely on a combina ion o upwind and downwind elemen alues o ace
econs uc ion. In con as , he no-neighbo me hod implemen ed in SWMM5+ ensu es ha econs uc ed
aces u ilize only adjacen elemen s, he eby minimizing spa ial dependency and educing communica ion
o e head be ween p ocesso s. This design signi ican ly dec eases da a exchange be ween compu a ional co es,
imp o ing he e iciency o pa allel compu a ion. A de ailed explana ion can be ound in (Hodges & Liu,
2019).
The nume ical s abili y o he RK2 scheme is go e ned by he CFL condi ion, which necessi a es ha
u∆
/∆x≤Cmax o p e en nume ical di e gence. In SWMM5+, howe e , o ensu e global nume ical s abili y,
he CFL cons ain is u he es ic ed o CFL <√2/2, wi h a dynamically adjus ed ime s ep. This
limi a ion is ypically imposed by he p esence o small elemen s, such as junc ions, which o en esul in
a ime s ep ha is disp opo iona ely smalle han he ime s ep in he EPA SWMM model o he same
sys em. While his smalle ime s ep can be a limi a ion, i can be e ec i ely add essed h ough he use o
pa allel compu ing, which alle ia es he compu a ional bu den (Hodges e al., 2024).
2.3 Compa ison o Obse ed Da a
Al hough he SWMM5+ simula ion con igu a ion and o cing in his s udy a e di ec ly adop ed om he
NWM se up wi hou speci ic hyd odynamic ailo ing (i.e., he model is uncalib a ed), i is s ill aluable
o include a p elimina y compa ison o obse ed da a. Figu e 5 shows he discha ge imese ies esul s o
SWMM5+ compa ed agains USGS gauge da a o wo junc ions in he La aca i e ne wo k (seen in Fig. 1).
The esul s demons a e ha SWMM5+ can p oduce easonable consis ency wi h ield obse a ions, cap u -
ing compa able peak alues and peak a i al imes. The simpli ied con igu a ion o apezoidal ba hyme y
inhe i ed om NWM may ha e lead o o e es ima es o discha ge in he simula ions. None heless, as he
p ima y objec i e o his s udy is o assess he pe o mance o he SWMM5+ model ela i e o SPRNT and
EPA SWMM, he simpli ied ba hyme y assump ion was e ained o ensu e con e gence ac oss all models
and main ain consis en compa ison s anda ds.
7
2017-01-15
2017-02-01
2017-02-15
2017-03-01
2017-03-15
2017-04-01
2017-04-15
2017-05-01
2017-05-15
Da e ime
0
20
40
60
80
100
Discha ge (
m
3/
s
)
08164450
Gauge Da a
SWMM5+ Da a
2017-01-15
2017-02-01
2017-02-15
2017-03-01
2017-03-15
2017-04-01
2017-04-15
2017-05-01
2017-05-15
Da e ime
08164503
Fig. 5. Compa ison o SWMM5+ simula ion esul s wi h ield-obse ed daily discha ge da a om wo
gauges in he La aca Ri e ne wo k. Le panel: USGS 08164450 (i.e., S a (1) in Fig. 1); igh panel: USGS
08164503 (i.e., S a (2) in Fig. 1).
3 Ini ialize SVE Simula ion – “Spin-Up Time”
The amoun o ime be o e simula ed esul s a e ee om he impac s o unce ain ini ial condi ions is
commonly e e ed o as he model “spin-up ime”. In nume ical simula ion, he spin-up ime is gene ally
una oidable, as i is a ely possible o ob ain p ecisely accu a e spa ially dis ibu ed ini ial condi ions ha
a e ully consis en wi h he co esponding spa ially dis ibu ed bounda y condi ions a = 0. As highligh ed
by Yu e al. (2017), no uni e sal c i e ion exis s o de ining spin-up ime in i e ne wo k simula ions; a he ,
i is in luenced by he quali y o ini ial condi ion con igu a ions and he deg ee o consis ency be ween ini ial
and bounda y condi ions.
He ein he La aca and San Jacin o i e basins we e “spun-up” using sligh ly a iable app oaches. Fo he
La aca Ri e , ela i ely s able bounda y condi ions we e imposed; o he San Jacin o Ri e , eal o cing
da a, including a signi ican uno e en , we e u ilized o e alua e model esponses unde dynamic hyd o-
logical condi ions. Ze o ini ial condi ions we e applied in he SWMM5+ and EPA SWMM models o assess
hei con e gence o he SPRNT model, wi h he du a ion de ined as he spin-up ime. Con e gence was
de e mined based on he pe cen age di e ence be ween simula ed hyd og aphs a he inal junc ion ups eam
o he ou le , conside ing hem con e ged when he di e ence emained wi hin ±1% o a minimum pe iod
o six hou s. The esul ing discha ge hyd og aphs and pe cen age di e ences be ween he models and he
benchma k SPRNT model a e p esen ed in Figu e 6.
8
La aca Ri e San Jacin o Ri e
Discha ge (m3/s)Di e ence (%)
Time (hou ) Time (hou )
Fig. 6. Spin-up analysis o he La aca and San Jacin o Ri e ne wo ks using SWMM5+, EPAkin, and
EPAdyn. Hyd og aphs a e e alua ed a he inal junc ion ups eam o he ou le o bo h i e ne wo ks.
The op panels depic he ime equi ed o hyd og aphs om each model o achie e con e gence wi h he
benchma k model, SPRNT. The bo om panels p esen he pe cen age di e ences be ween he h ee models
and SPRNT o e he simula ion pe iod.
The ime a sys em akes o e ase any memo y o he ini ial condi ions is la gely a physical p ocess, a he
han a nume ical one. La ge , shallowe sys ems equi e mo e spin-up ime han smalle , s eepe ones. The
esul s o Fig. 6 show ha he discha ge om he ou all o each compa ison model con e ged o he SPRNT
model wi hin ˜
350 hou s o he La aca Ri e , ˜
200 hou s o he San Jacin o Ri e . No ably, he SWMM5+
model ook he longes o con e ge o he La aca Ri e wi h consis en bounda y condi ions, despi e being
he i s o con e ge o he San Jacin o i e wi h hyd ologically ealis ic bounda y condi ions.
4 Pos p ocessing Techniques o Model- o-Model Analysis
4.1 Hyd og aph Smoo hing
Jagged, discon inuous beha io o solu ion a iables p esen s an impedimen o au oma ed p ocessing o
hyd og aph and s age esul s. To add ess his issue, ou analysis wo k low employs a non-linea median
il e (NLMF) o smoo hing noisy imese ies da a, simila o he app oach o Whi e and Hodges (2005) o
smoo hing ba hyme ic da a. The beha io o he NLMF is simila o a adi ional mo ing-a e age window,
excep he use o he median alue a he han he a i hme ic a e age educes he bias om spu ious hyd o-
g aph spikes.
NLMF smoo hing can be ep esen ed o a il e window o size 2N + 1 as
˜
ϕ[xi, k] = MEDIAN ϕ[xi, k−N ], ϕ[xi, k−N +1], ...ϕ[xi, k+N ](1)
whe e ˜
ϕ ep esen s he NLMF- il e ed da a se . Values o ϕ[xi, k< Ts a ] and ϕ[xi, k> Tend] a e igno ed.
The NLMF p ese es key hyd og aph ea u es such as he slope o he ising limb and iming o peak lows.
9
a simila endency o p oduce excessi e SD alues in downs eam channels, a ibu able o oscilla o y
nume ical esul s, as we obse ed in he La aca Ri e case. F om he au ho ’s obse a ion, he EPAdyn
simula ions exhibi mo e p onounced nume ical ins abili y in he San Jacin o Ri e compa ed o he La aca
Ri e . Despi e he applica ion o a mo ing a e age il e o mi iga e luc ua ions, he il e is insu icien
o add ess he se e e oscilla ions, pa icula ly in downs eam channels whe e he nume ical ins abili y is
mos p onounced. Fo compa ison and demons a ion pu poses, hese samples a e e ained in he sca e
plo . The EPAkin con inues o exhibi simila beha io o wha was obse ed in he La aca case, wi h a
compa able le el o SD o he SWMM5+ model bu a b oade ange o SD , pa icula ly in channels nea
he i e ou le .
7500100
Ma ke Size - Ca chmen A ea (km2)
Ma ke Colo
SWMM5+
EPA kin
EPA dyn
Ma ke Shape
Jan 20 E en
Feb 21 E en
May 16 E en
SD (h )
SD (m3/s)
Fig. 11. Sca e plo o iming se ies di e ence (SD ) e sus magni ude se ies di e ence (SD ) o he San
Jacin o Ri e , based on h ee hyd ological e en s (Janua y 20, Feb ua y 21, and May 16, 2017). Box plo s
o SD and SD a e included along he op and igh ma gins, espec i ely, o illus a e he dis ibu ion o
iming and magni ude di e ences o each model.
16
6 Adjus men s and con igu a ions made in he sol e s o ensu e
he acqui emen o compa able esul s
To p ese e he ull dynamics o he Sain -Venan go e ning equa ions, a speci ic code block in he Dynamic
Wa e module (dw low.c) o he SWMM5 sou ce code was commen ed ou , ollowing guidance om Rossman,
Dickinson, and Tie nan (2022). Figu e 13 illus a es he simula ed low a e a he ou all o he La aca Ri e
ne wo k o e he i s 40 days o simula ion using he EPAdyn sol e , unde wo con igu a ions: (a) wi h he
no mal low limi a ion emo ed o enable ull hyd odynamic complexi y, and (b) wi h he de aul no mal
low limi a ion enabled. The esul s demons a e ha disabling he no mal low limi e imp o es nume ical
s abili y a he ne wo k ou le . Acco dingly, all EPAdyn simula ions in his s udy we e conduc ed wi h he
no mal low limi a ion disabled o ensu e consis en applica ion o he comple e Sain -Venan equa ions.
Fig. 12. Sc eensho o no mal low limi ing code block in dw low.c. When he block is ac i a ed, a no mal
low condi ion is en o ced.
17
(a) De aul no mal low limi e deac i a ed.
(b) De aul no mal low limi e ac i a ed.
Fig. 13. Simula ed low a es a he ou all o he La aca Ri e Ne wo k om EPAdyn showing s abili y
di e ence om oggling he no mal low limi ing code block in he dw low.c module.
7 Conclusion
This echnical epo p o ides suppo ing documen a ion o he jou nal a icle “Ad ancing Wa e shed-Scale
Hyd odynamic Modeling wi h he Sain -Venan Equa ions”, wi h a ocus on he implemen a ion de ails,
adjus men s, and supplemen al analyses used in he model- o-model compa ison. The ma e ial p esen ed he e
includes p ac ical s eps o cons uc ing la ge-scale i e ne wo k models, se ing up nume ical expe imen s
using Sain -Venan sol e s, and pos -p ocessing me hods o e alua ing lood wa e beha io .
This epo ou lines he p ocedu es used o gene a e egional-scale i e ne wo k models o he La aca
and San Jacin o Ri e basins, including simpli ica ions applied o channel geome y, node in low handling,
and ba hyme ic assump ions. I also desc ibes modi ica ions made o he SWMM5 sou ce code o p ese e
ull hyd odynamic dynamics, as well as he a ionale o including each sol e in he compa ison.
The pos -p ocessing me hods de ailed in his epo —such as e en -based se ies dis ance me ics and
spa ial e o mapping—se e as p ac ical ools o diagnosing nume ical beha io ac oss complex i e sys-
ems. In addi ion, supplemen al esul s a e p o ided o illus a e model di e ences in hyd og aph shape,
lood wa e iming, and nume ical s abili y unde iden ical o cing.
18
O e all, his epo is in ended o enhance he anspa ency and ep oducibili y o he modeling wo k-
low desc ibed in he main a icle. I also se es as a e e ence o esea che s seeking o apply o ex end
he compa ison amewo k o addi ional i e sys ems o sol e s. By documen ing implemen a ion-speci ic
choices and sol e con igu a ions, we hope his esou ce suppo s con inued de elopmen and e alua ion o
wa e shed-scale hyd odynamic models using he Sain -Venan equa ions.
19
Re e ences
Da id, C. H., Maidmen , D. R., Niu, G.-Y., Yang, Z.-L., Habe s, F., & Eijkhou , V. (2011). Ri e
Ne wo k Rou ing on he NHDPlus Da ase . Jou nal o Hyd ome eo ology,12 (5), 913–934. doi:
10.1175/2011JHM1345.1
Eh e , U., & Zehe, E. (2011). Se ies dis ance–an in ui i e me ic o quan i y hyd og aph simila i y in e ms
o occu ence, ampli ude and iming o hyd ological e en s. Hyd ology and Ea h Sys em Sciences,
15(3), 877–896. doi: 10.5194/hess-15-877-2011
Fal e , D., Dung, N., Vo ogushyn, S., Sch ¨o e , K., Hundecha, Y., K eibich, H., . . . Me z, B. (2016).
Con inuous, la ge-scale simula ion model o lood isk assessmen s: p oo -o -concep : La ge-scale
lood isk assessmen model. Jou nal o Flood Risk Managemen ,9(1), 3–21. Re ie ed 2022-04-11,
om h ps://onlinelib a y.wiley.com/doi/10.1111/j 3.12105 doi: 10.1111/j 3.12105
Hodges, B. R., & Liu, F. (2019). Timescale in e pola ion and no-neighbou disc e iza ion
o a 1D ini e- olume Sain -Venan sol e . Jou nal o Hyd aulic Resea ch, 1–17. doi:
10.1080/00221686.2019.1671510
Hodges, B. R., Sha io , S., Tie nan, E. D., Jenkins, E., Ria˜no-B ice˜no, G., Da ila-He nandez, C., . . .
Yu, C.-W. (2024). In oducing SWMM5+. , 150 (10), 02524003. Re ie ed 2024-11-07, om
h ps://ascelib a y.o g/doi/10.1061/JOEEDU.EEENG-7680 doi: 10.1061/JOEEDU.EEENG-
7680
Hodges, B. R., Yu, C.-W., Tie nan, E. D., Ria˜no-B ice˜no, G., Sha io , S., & Jenkins, E. (2021).
SWMM5+ Alpha Release Documen a ion [Compu e so wa e manual]. Texas Da a Reposi o y. doi:
10.18738/T8/WQZ5EX
Liu, F. (2014). SPRNT Use ’s Manual.
Niazi, M., Nie ch, C., Magh ebi, M., Jackson, N., Benne , B. R., T yby, M., & Massoudieh, A. (2020).
S o m Wa e Managemen Model: Pe o mance Re iew and Gap Analysis. , 59.
NOAA. (2016). The Na ional Wa e Model [Wa e .Noaa.Go ]. wa e .noaa.go /abou /nwm.
Pai a, R. C. D., Collischonn, W., Bonne , M. P., & de Gon¸cal es, L. G. G. (2012). On he sou ces o
hyd ological p edic ion unce ain y in he Amazon. Hyd ology and Ea h Sys em Sciences,16(9),
3127–3137. Re ie ed 2022-04-11, om h ps://hess.cope nicus.o g/a icles/16/3127/2012/
doi: 10.5194/hess-16-3127-2012
Rossman, L., Dickinson, R., & Tie nan, E. (2022). p i a e communica ion. (Email co espondence
da ed Ma ch 7-11, 2022 be ween manusc ip au ho s and no able SWMM5 de elope s.) doi:
h ps://doi.o g/10.6084/m9. igsha e.29474117. 2
Swa hi, V., Raju, K. S., & Va ma, M. R. R. (2020). Addi ion o o e land uno and low ou ing me hods
o SWMM—model applica ion o Hyde abad, India. En i onmen al Moni o ing and Assessmen ,
192(10), 643. doi: 10.1007/s10661-020-08490-0
Tie nan, E. D., & Hodges, B. R. (2022). A Topological App oach o Pa i ioning Flow Ne wo ks o Pa allel
Simula ion. Jou nal o Compu ing in Ci il Enginee ing,36 (4), 13. doi: 10.1061/(ASCE)CP.1943-
5487.0001020
Tu ne , S. W. D., S eyae , J. C., Condon, L., & Voisin, N. (2021). Wa e s o age and elease poli-
cies o all la ge ese oi s o con e minous Uni ed S a es. Jou nal o Hyd ology,603, 126843. doi:
10.1016/j.jhyd ol.2021.126843
USGS. (2020). Na ional Hyd og aphy Da ase . h ps://www.usgs.go /co e-science-sys ems/ngp/na ional-
hyd og aphy.
Whi e, L., & Hodges, B. R. (2005). Fil e ing he signa u e o subme ged la ge woody deb is om ba hyme y
da a. Jou nal o Hyd ology,309 (1-4), 53–65. doi: 10.1016/j.jhyd ol.2004.11.011
Xiong, Y., & Melching, C. S. (2005). Compa ison o Kinema ic-Wa e and Nonlinea Rese oi Rou ing o
U ban Wa e shed Runo . Jou nal o Hyd ologic Enginee ing,10 (1), 39–49. doi: 10.1061/(ASCE)1084-
0699(2005)10:1(39)
Yu, C.-W., Hodges, B. R., & Liu, F. (2021). Au oma ed De ec ion o Ins abili y-Inducing Channel Geome y
T ansi ions in Sain -Venan Simula ion o La ge-Scale Ri e Ne wo ks. Wa e ,13(16), 2236. Re ie ed
2022-02-07, om h ps://www.mdpi.com/2073-4441/13/16/2236 doi: 10.3390/w13162236
Yu, C.-W., Liu, F., & Hodges, B. R. (2017). Consis en ini ial condi ions o he Sain -Venan equa ions in
i e ne wo k modeling. Hyd ology and Ea h Sys em Sciences,21(9), 4959–4972. doi: 10.5194/hess-
20
21-4959-2017
21
