Lumped model research inventory

This is a broad, source-linked inventory, not an assertion that every lumped model ever developed has been found, nor that every entry is implemented here. “Lumped” is a spatial configuration: some model families also have semi-distributed/distributed forms. Closely named variants can have different equations, solvers, states, and parameter meanings.

Available now

Version 0.2.0 has 14 Python models: LuMod GR4J, GR2M, GR1A, modified HYMOD2, modified HBV-light and MILC; hydromodel GR3J, GR5J, GR6J, classic HYMOD, XAJ and XAJ_MZ; SMARTpy SMART; and native monthly ABCD. Additionally all 47 MARRMoT structures below have an optional persistent Octave adapter. list_models() lists Python models; list_models(include_optional=True) lists all 61 selectable implementations. Overlapping family names are intentionally not counted as distinct original families.

The six LuMod adapters use LuMod. The other engine revisions, licenses and compatibility changes are documented in THIRD_PARTY.md. Numerical consistency tests do not establish field skill or source-equivalence with every variant of a named model.

Other established families / integration candidates

Family

Official implementation/reference

Current BasinForge status

GR3J, GR5J, GR6J

INRAE GR models, hydromodel

Runnable hydromodel variants; airGR equivalence unverified

CemaNeige + GR models

INRAE CemaNeige, RavenPy

Research only; not equivalent to standalone GR4J

Classic five-parameter HYMOD

hydromodel, SPOTPY example

Runnable HYMOD_CLASSIC via hydromodel; MARRMOT_29 separate standardized structure

HBV-96, HBV-EC, TUW/HBV variants

hydromad HBV, RavenPy

Research only; LuMod HBV is a particular modified variant

SAC-SMA / Sacramento

RavenPy Sacramento, eWater

MARRMOT_33 standardized structure runnable; original engines research only

Xinanjiang / XAJ, XAJ-MZ, XAJ-SLW

hydromodel

XAJ/XAJ_MZ runnable; SLW routing variant not integrated

Dahuofang / DHF

hydromodel

Research only

SIMHYD, IHACRES

hydromad, eWater

MARRMOT_18/05 standardized structures runnable; original engines research only

HMETS, MOHYSE, Canadian Shield, HYPR

RavenPy emulators

Research only; an emulator is not automatically identical to every original model

FLEX/SUPERFLEX structures

SuperflexPy

User-supplied model interface; no bundled adapter

TOPMODEL, TANK, NAM, SMAR, MODHYDROLOG, MCRM, HYCYMODEL, GSM-SOCONT, PRMS, CLASSIC, LASCAM, TCM, VIC-inspired lumped structures

MARRMoT

Standardized MARRMoT structures runnable via Octave; original engines not integrated

SMART

SMARTpy

Runnable daily-to-hourly Python adapter; uniform forcing assumption

ABCD

Monthly equations

Runnable monthly native water balance

AWBM

hydromad AWBM

Research only; soil accounting must be paired with appropriate routing

WAPABA, IHACRES, SAC-SMA original variants

CSIRO PyGME

Research only; compiled C and hydrodiy dependencies require separate validation

TUW/HBV

TUWmodel

Research only; R/compiled engine distinct from existing HBV implementations

PDM

UKCEH official implementation

Research only; official source redistribution/license must be confirmed

CFE and TOPMODEL BMI

NOAA NextGen model inventory, CFE

Research only; compiled engines and additional soil/catchment configuration

HEC-HMS lumped configurations

USACE features

External system, not a bundled open-source Python model; event/continuous options differ

The hydromodel package already offers multi-model calibration; BasinForge is not the first such package. RavenPy, hydromad, SuperflexPy, and SPOTPY also provide substantial modelling/calibration tools. The intended contribution here is explicit units/frequencies, lightweight prepared-array adapters, equal-weight shared-basin fitting, chronological validation, and reusable completed basin jobs.

Complete 47-structure MARRMoT inventory

Names, parameter counts, and store counts below are transcribed from the official model-file directory. MARRMoT explicitly notes that its standardized structures resemble but are not identical to the original models. Some names represent experimental/unnamed structures rather than separately established model families.

All entries below are integrated in BasinForge 0.2.0 under keys MARRMOT_01 through MARRMOT_47, using the pinned v2 source and external Octave + optim. All 47 have controlled-forcing upstream parity and calibration-interface tests. This does not validate every model on real basins or guarantee convergence for arbitrary parameters. Source-derived parameter ranges/default midpoints need basin-specific judgment. The runtime is not bundled in the wheel.

Daily P/PET are mm/step; mean temperature °C is required for IDs 06, 12, 30, 31, 32, 34, 35, 37, 41, 43, 44 and 45. Parameters p01... follow each source class’s exact order; fixed initial stores s01... default to zero. Generic names avoid incorrectly assigning the same meanings/units to parameters across structures. See source-class comments before supplying bounds.

ID

Structure name in source

Parameters

Stores

01

collie1

1

1

02

wetland

4

1

03

collie2

4

1

04

newzealand1

6

1

05

ihacres

7

1

06

alpine1

4

2

07

gr4j

4

2

08

us1

5

2

09

susannah1

6

2

10

susannah2

6

2

11

collie3

6

2

12

alpine2

6

2

13

hillslope

7

2

14

topmodel

7

2

15

plateau

8

2

16

newzealand2

8

2

17

penman

4

3

18

simhyd

7

3

19

australia

8

3

20

gsfb

8

3

21

flexb

9

3

22

vic

10

3

23

lascam

24

3

24

mopex1

5

4

25

tcm

6

4

26

flexi

10

4

27

tank

12

4

28

xinanjiang

12

4

29

hymod

5

5

30

mopex2

7

5

31

mopex3

8

5

32

mopex4

10

5

33

sacramento

11

5

34

flexis

12

5

35

mopex5

12

5

36

modhydrolog

15

5

37

hbv

15

5

38

tank2

16

5

39

mcrm

16

5

40

smar

8

6

41

nam

10

6

42

hycymodel

12

6

43

gsmsocont

12

6

44

echo

16

6

45

prms

18

7

46

classic

12

8

47

IHM19

16

4

Integration acceptance criteria

For each additional model: verify upstream license and redistribution rights; identify exact variant/timestep/state convention; validate forcing and discharge units; reproduce authoritative reference outputs; check dry/wet/snow/extreme boundary cases and conservation assumptions; verify optimizer bounds and state initialization; benchmark real record lengths; only then advertise it as supported.

Event runoff methods (e.g. SCS-CN + unit hydrograph) and distributed frameworks (e.g. SWAT, mHM) should not be silently presented as interchangeable continuous daily lumped models.

Survey performed during this build using primary project documentation and official source inventories. It is not a systematic literature review with an exhaustive search protocol.