Appendix 5: RCA

Appendices ››
Parent Previous Next

Application of River Condition Assessment to River Habitat Survey data


1. Introduction

All 32 River Condition Assessment (RCA) indicators, the Preliminary Condition Score and the Final Condition Class can be estimated from a standard River Habitat Survey (RHS). Indicators are converted with one set of explicit rules and scored with the published MoRPh thresholds, unchanged.

The RCA (Gurnell et al., 2020; RCA guide, January 2024) is the river condition component of the statutory Biodiversity Metric used for Biodiversity Net Gain (BNG) in England. It is built on the Modular River Survey (MoRPh). Many restoration and development sites already hold RHS data, and RHS has an established accreditation scheme, so an RHS route to the RCA lets those data be used rather than resurveyed.

This document describes the RCA module of the RHS Toolbox v221 and above, which applies the RCA to RHS data. It covers:


2. The RCA method in brief

The RCA scores a river subreach against what its River Type could achieve if it functioned naturally. It has four stages.

    1. Indicative River Type. Eight indicators (A1 to A8) assign the reach to one of 13 near-natural types (A to M), or to Large rivers or Canals and navigable rivers. A1 to A5 come from a desk study (braiding, sinuosity, anabranching, confinement, valley gradient); A6 to A8 describe the bed material from the field survey. A decision tree combines them.
    2. Condition indicators. 32 indicators summarise the field survey: 19 positive (0 to +4) for the diversity and extent of natural habitats, and 13 negative (0 to −4) for human pressures. They cover the bank top (B1 to B5), bank face (C1 to C10), channel margin (D1 to D5) and channel bed (E1 to E12).
    3. Preliminary Condition Score = average of the positive scores + average of the negative scores.
    4. Final Condition Class. The Preliminary Condition Score is translated into Good (3), Fairly Good (2.5), Moderate (2), Fairly Poor (1.5) or Poor (1) using River Type thresholds (guide, Table 3). The class can be lowered by one if the channel's Shape (width ÷ (water depth + lower bank height)) shows it is overdeep and other evidence agrees.

Each indicator is computed in the same way in MoRPh. Field abundances are converted to numbers: Absent = 0, Trace = 2, Present = 19, Extensive = 67, roughly the mid-point percentage cover of each class. These are then weighted, summed over both banks and the 5 modules, and banded into a score with thresholds calibrated on 40 subreaches.

This structure is what makes an RHS conversion possible. Each indicator is a sum of approximate percentage covers, counts or presences, so it can be estimated from any survey that records the same features at a comparable level of detail.


3. Compatibility of RHS data with the RCA

RHS records almost every feature the RCA uses, but it samples them differently. RHS records presence at points plus a whole-site summary; MoRPh records abundance over five continuous strips. Only a few RCA inputs are missing outright. The main difficulty is converting one sampling design into the other.

3.1 Survey design

Aspect

MoRPh5 subreach

RHS site

Length

10 channel widths: 50, 100, 150 or 200 m by width (250 m for canals and rivers wider than 30 m)

Fixed 500 m

Sampling units

5 contiguous modules, each about 2 widths long, each surveyed in full

10 spot-checks every 50 m (physical attributes on a 1 m transect, vegetation over 10 m), plus one sweep-up of the whole site

Coverage

Continuous: every metre of the subreach is in a module

Spot-checks sample about 2-20% of the length; the sweep-up covers 100% at low resolution

Abundance scale

A / T / P / E per module (0, <1%, 1–33%, >33% of the module)

Spot-checks: mostly presence, or N / P / E for channel vegetation; sweep-up: N / P / E for the whole site (P = present, <33%; E = 33% or more)

Banks

Most features recorded per bank, per module

Spot-checks per bank; most sweep-up features for the whole site, some per bank (bank profiles, land use, berms)

Counts

Counts per module (pools, riffles, weirs, outfalls...)

Counts for the whole 500 m (riffles, pools, weirs, culverts, bridges, outfalls, deflectors)

Corridor

Bank top to 10 m from the edge

Land use within 5 m at spot-checks; within 50 m in the sweep-up

Vegetation structure

Presence of each of 5 structural types

One class per spot-check (Bare, Uniform, Simple, Complex), defined on the same 5 types

Three consequences follow:


3.2 RCA inputs missing from RHS

Missing input

Indicators affected

Importance

Leaning and J-shaped trees

B2, C2

Moderate. Added to the Toolbox as a user question.

Bank top vs bank face location of tree features

B2, C2

Minor. RHS tree features are site-wide, so the same records feed both indicators.

Bench, bank toe

D3, D4

Minor. 2 of 9 margin features and berms cover benches in RHS; D4 can still reach its maximum score.

Tributary junctions

D3, D4

Minor. The presence of a significant tributary is recorded in RHS Map data optional field as presence/absence. We created an optional user field populated with that information, if available and changeable by users .

Channel margin vs channel bed location of aquatic vegetation

D1, D2, E1

Moderate. The same RHS vegetation records feed both groups.

Steps

E4, E5

Minor except in step-pool streams (Type C), where steps matter most.

Trees growing from the river bed

E2

Minor.

Organic bed material

E6

Minor. Not missing in practice: organic bed material is recorded in RHS channel substrate as earth (EA), which E6 now counts as organic.

Patchy vs continuous silt layer

E7

Moderate. RHS has a site-level silt drape (newer surveys only) and discrete silt deposits.

Bed reinforcement material

E9

Moderate. Optional user field, otherwise approximated from bank materials.

Bridge piers in the bed

E10

Moderate where present. Optional user field.

NNIPS on the channel bed and bars (balsam, knotweed, hogweed)

E11

Moderate. RHS records only floating pennywort and “other” in the channel and the other species are generally on the banks and occasionally in the channel. Bank NNIPS record cover it to a large extent.

Trace (T) abundance class

All extent indicators

Minor. RHS Present covers MoRPh Trace and Present.

Upper and lower bank material separately

C5

Minor. One dominant material per bank per spot-check.

Natural profile types Vo and Vu separated from V

C4

Minor. RHS has 5 of MoRPh's 7 natural profiles.

3.3 RHS information the RCA does not use

RHS also records material the RCA does not ask for, and some of it could strengthen the assessment. The Toolbox uses some of it already.

4. Solution: converting RHS records to MoRPh values

Every indicator is converted in the same two steps. First the MoRPh indicator value is estimated from RHS records using five rules. Then the published MoRPh thresholds turn it into a score, unchanged. The RHS version therefore differs from MoRPh only in how the value is estimated, and each rule is a stated assumption that can be tested by changing a constant.

4.1 The five conversion rules

Rule 1 – Pseudo-modules (spot-check presence data). The 10 spot-checks are grouped in consecutive pairs (1–2, 3–4, 5–6, 7–8, 9–10) to form 5 pseudo-modules, each 100 m long, mirroring the 5 MoRPh modules. A feature recorded at one spot-check of a pair is scored as MoRPh Present (19); at both, as Extensive (67). Bank-specific records (bank material, bank modification, bank features, land use, vegetation structure) are treated bank by bank, as in MoRPh.

Why: one spot-check of a pair hit means the feature covers part of that 100 m stretch on that bank. Two hits mean it is likely continuous over a large part of it. The pairing keeps the MoRPh structure (5 units × 2 banks), so the MoRPh thresholds apply directly.

Rule 2 – Spot-check abundance (channel vegetation N/P/E). Each spot-check scores P = 19 and E = 67. For channel-margin indicators (D1), each of the 10 spot-checks stands for one of MoRPh's 10 margin-modules (5 modules × 2 margins). For channel-bed indicators (E12), the sum is halved to match 5 bed-modules.

Rule 3 – Sweep-up data. The sweep-up is used in two ways.

Rule 4 – Counts. RHS counts cover 500 m; MoRPh counts cover 50–250 m. Counts of positive features (riffles, pools) are rescaled to the expected number in a MoRPh5 subreach of the same width (count × MoRPh5 length ÷ surveyed length; the surveyed length is 500 m unless the surveyor enters another one in `Survey length`). They are then spread evenly over the 5 modules and scored with MoRPh's rule (19 for a count of 1 in a module, 67 for more than 1). Counts of negative features (weirs, culverts, outfalls, deflectors) are not rescaled by default. This reflects the RCA rule that at least one subreach must capture the most degraded part of the river; either option can be switched. The Rescale tick box on the BNG form (ticked by default) turns all rescaling off when unticked, so every count is used as recorded over the surveyed length.

Rule 5 – Unchanged thresholds. The MoRPh scoring bands are applied as published. Where a range is open (“>4”), integer values are used as written.

Constants used in the calculation

Default

Meaning

GAP_P, GAP_E

19, 134

Sweep-up record missed by all spot-checks

SWEEP_P, SWEEP_E

19, 335

Sweep-up record with no spot-check equivalent

SCALE_POSITIVE_COUNTS

True

Rescale riffle and pool counts to MoRPh5 length

SCALE_NEGATIVE_COUNTS

False

Rescale weir, culvert and outfall counts

BRIDGES_AS_DEFLECTORS

False

Count RHS bridges as bank deflectors in D5

4.2 Data added to the Toolbox

Inputs RHS does not collect are requested on the BNG form. The code works without the optional ones, with a stated default.

5. Assigning the indicative River Type

The River Type is derived in three steps. The user enters the desk-study indicators A1 to A5. The Toolbox derives the bed-material indicators A6 to A8 from the RHS survey, using the predicted natural substrate for A8 where one exists. A decision tree then proposes a type, which the user can override. Large rivers and Canals and navigable rivers are selected by the user, as in the guide.

5.1 Desk-study indicators (A1 to A5)

Indicator

Input on the BNG form

Calculation

Used for

A1 Braiding index

Mean number of flowing threads across 10 cross-sections

Entered directly; 1 for single thread

Multi-thread (>1.1) vs single thread

A2 Sinuosity index

River length and valley length (m)

River length ÷ valley length, button cmdCalcA2

Straight-sinuous (<1.5) vs meandering (≥1.5)

A3 Anabranching index

Mean number of channels separated by islands

Entered directly; 1 for single thread

Anabranching (≥1.5) on sand or finer beds

A4 Confinement

Confined, partly confined or unconfined

Entered directly

Sense-check only (Types A–D are usually confined)

A5 Valley gradient

Upstream and downstream valley elevations (m)

(upstream − downstream) ÷ valley length, button cmdCalcA5

Steep plane-bed streams (≥0.01)


5.2 Bed-material indicators from RHS (A6 to A8)

A6 Bedrock reach. MoRPh sets A6 when bedrock is Extensive in at least 3 modules, or Extensive in 2 and Present in 3. With mid-point covers of about 65% (E) and 15% (P), those two cases mean bedrock covers about 39% and 35% of the bed. In RHS, each spot-check represents about 10% of the site, so A6 is set when bedrock (BE) is the substrate at 4 or more of the 10 spot-checks (40%).

A7 Coarsest bed material. A7 is the coarsest natural material recorded at any spot-check, or listed as an additional substrate in the sweep-up. The order is bedrock, boulder, cobble, gravel-pebble (RHS GP, G and P), sand, silt, clay. Artificial substrate (AR) is excluded. As the guide's note 3 warns, boulders may be artificial (rip-rap, washed-out reinforcement). Where the valley gradient is below 0.01 and boulders are coarsest, the user should check them and, if artificial, replace A7 by A8.

A8 Average natural bed material. A8 should describe the natural bed, but an RHS site may be silted, dredged or reinforced. The RCA guidance is not clear about whether the typology should represent the natural or observed river type. As is, it mainly records the observed type but allows users to change it when they consider it 'inappropriate for a valid geomorphological reason'. As condition assessment should ideally be assessed against 'natural' condition and we have predictions of natural substrate, we have here an opportunity to select a river type reflecting naturalness. A8 is therefore taken, in order of preference, from:

    1. The predicted natural substrate. The Channel Substrate Index for semi-natural conditions is predicted from map data and stored in the Toolbox. It is classed as: below −1.8 SI; −1.8 to −1.69 CL; to −1 SA; to 0.21 GP; to 0.74 CO; 0.74 and above BO. The CSI scale places silt below clay, so this order is intended.
    2. The observed substrate. Without a prediction, the guide's weighted phi average is applied to the spot-check substrates, each spot-check counting as one unit of cover: phi = (10 CL + 6.5 SI + 1.5 SA − 3.5 GP − 7 CO − 9 BO) ÷ (CL + SI + SA + GP + CO + BO). Bedrock and artificial substrate are excluded. The result is classed as: above 9 CL; above 4 SI; above −1 SA; above −6 GP; above −8 CO; otherwise BO.

The comment field records which source was used.

5.3 Decision tree

cmdPrelimType first checks that A1, A2, A3, A5, A6, A7 and A8 are filled in. If A6 is set, the reach is Type A (bedrock). Otherwise the tree reads A7, then A8, then the planform rules.

A7 (coarsest)

A8 (average)

Rule

Type

BE or BO

BO

–

B Cascade

BE or BO

CO

–

C Step-pool

BE or BO

GP

A5 ≥ 0.01, else gravel planform

D Plane bed, else E / F / G

BE or BO

SA

A5 ≥ 0.01, else sand planform

D Plane bed, else J / H / I

CO or GP

GP or CO

Gravel planform

E / F / G

CO, GP or SA

SA

Sand planform

J / H / I

GP, SA, SI or CL

SI or CL

Fine planform

M / K / L

Planform rules:

Combinations not in the table, such as boulders with a silt average, give no type. The user then selects one, and the flag records any type that differs from the derived one. The tree is the transcription of the guide's Figure 3 from the original Toolbox code, unchanged apart from the missing-data check.

6. The condition indicators

Each indicator below lists what it represents, the MoRPh data and scoring, how it is derived from RHS with a worked example, and how an RHS-based score is likely to compare with a MoRPh-based one. (+) marks a positive indicator and (−) a negative one. P and E are Present and Extensive; “pair” means a pseudo-module of two spot-checks (Rule 1).

Group B: Bank top

B1 Bank top vegetation structure (+)

B2 Bank top tree feature richness (+)

B3 Bank top water-related features (+)

B4 Bank top NNIPS cover (−)

B5 Bank top managed ground cover (−)

Group C: Bank face

C1 Bank face riparian vegetation structure (+)

C2 Bank face tree feature richness (+)

C3 Bank face natural bank profile extent (+)

C4 Bank face natural bank profile richness (+)

C5 Bank face natural bank material richness (+)

C6 Bank face bare sediment extent (+)

C7 Bank face artificial bank profile extent (−)

C8 Bank face reinforcement extent (−)

C9 Bank face reinforcement material severity (−)

C10 Bank face NNIPS cover (−)

Group D: Channel margin

D1 Channel margin aquatic vegetation extent (+)

D2 Channel margin aquatic morphotype richness (+)

D3 Channel margin physical feature extent (+)

D4 Channel margin physical feature richness (+)

D5 Channel margin artificial features (−)

Group E: Channel bed

E1 Channel aquatic morphotype richness (+)

E2 Channel bed tree feature richness (+)

E3 Channel bed hydraulic feature richness (+)

E4 Channel bed natural physical features extent (+)

E5 Channel bed natural physical feature richness (+)

E6 Channel bed material richness (+)

E7 Channel bed siltation (−)

E8 Channel bed reinforcement extent (−)

E9 Channel bed reinforcement severity (−)

E10 Channel bed artificial features severity (−)

E11 Channel bed NNIPS extent (−)

E12 Channel bed filamentous algae extent (−)

7. Preliminary score, final class and overdeepening

These steps follow the guide exactly. The only RHS-specific part is the estimate of width/depth ratio from RHS channel dimensions.

7.1 Preliminary Condition Score

The score is the average of the 19 positive indicator scores plus the average of the 13 negative scores, rounded to 2 decimals. It is calculated only when all 32 indicators have a score; otherwise the missing indicators are listed.

7.2 Final Condition Class

The guide's Table 3 thresholds are used to derive the final condition class. A score must be strictly above a threshold to reach its class. The thresholds divide the range between the worst and the likely best Preliminary Condition Score of each type: Poor takes the bottom 30%, Fairly Poor 25%, Moderate 20%, Fairly Good 15% and Good the top 10%.

River Type

Likely best

Good

Fairly Good

Moderate

Fairly Poor

Likely worst

Canals and navigable rivers

1.8

>1.4

>0.9

>0.3

>−0.5

−1.5

Large rivers

2.2

>1.8

>1.3

>0.5

>−0.4

−1.5

A, K, L, M

2.4

>1.9

>1.2

>0.2

>−1.0

−2.5

B, C, D, E

2.7

>2.2

>1.4

>0.2

>−0.9

−2.5

F, J

2.8

>2.3

>1.5

>0.4

>−0.9

−2.5

H

2.9

>2.4

>1.6

>0.5

>−0.9

−2.5

G

3.0

>2.5

>1.6

>0.5

>−0.9

−2.5

I

3.1

>2.5

>1.7

>0.6

>−0.8

−2.5

Scores at or below the Fairly Poor threshold are Poor. Source: RCA guide, January 2024, Table 3.

7.3 Overdeepening

MoRPh defines width/depth ratio as average width ÷ average (water depth + lower bank height). The RHS equivalent is bankfull width ÷ (water depth + mean banktop height of the two banks). Where RHS records that banktop is not bankfull and gives a trash-line height, that height is used instead. RHS measures these at one cross-section, where MoRPh averages five.

The Toolbox reports width/depth ratio in the comment field and flags values below 4 as “likely overdeep”. Shape alone never lowers the class; it is reported for the user's judgement, using the guide's thresholds (below 2: highly likely; below 4: likely, especially if wider than 10 m).

Overdeepening from the RHS record. width/depth ratio is taken from the value already computed in the Indices tab at the quality-assessment step; the RHS-based width/depth ratio estimate above is used only when that ratio is empty. The final class itself uses the surveyor's “overdeepened channel” entry, which is checked against the width/depth ratio and the Channel Resectioning Index at the QA step:

The message and the class before any downgrade are written to the comment field.

8. Overall fit and ways to improve it

RHS data support the RCA well for vegetation, bank modification, reinforcement and invasive species. Support is weaker for localised features (bars, bank materials, margin features) and for counts, which depend on site length. On balance RHS-based Preliminary Condition Scores are probably slightly lower than MoRPh-based ones would be: several positive indicators are underestimated and some negative ones overestimated. The method therefore errs on the cautious side for BNG baselines. This needs confirming with paired surveys.

8.1 Test on 201 RHS surveys

The rules were run on the 201 surveys in the Toolbox database, mostly restoration and urban sites. Every indicator except E11 produced a range of scores. The average positive score was 1.47 and the average negative score −0.91. Preliminary Condition Scores ranged from −1.52 to 2.16, with a median of 0.60.

The desk-study indicators were filled with RHS proxies:

197 sites received a River Type: F 97, H 62, K 18, G 7, D 6, C 3, I 3 and L 1. At 20 sites the coarsest substrate recorded (A7) was finer than the predicted dominant substrate (A8), or no substrate was recorded at all. The decision tree has no branch for these sites, so their type was derived with A7 raised to the level of A8 (see 8.3). The remaining four sites have no substrate data and no prediction, so they have no type.

Before the overdeepening step, over half the typed sites (104 of 197) are Moderate. That is plausible for a database dominated by modified and restoration sites. However, recorded overdeepening (page 1 of the RHS form) is Extensive on 136 of the 201 sites. The one-class reduction therefore applies to 131 typed sites, which moves the most common class to Fairly Poor and leaves 64 sites Poor. Before the rule is relied on, it should be checked whether an RHS record of extensive overdeepening is a strong enough signal to lower the class this often. The width-depth ratio estimation in the results allows that comparison.

8.2 Comparison with the RHS quality and impact indices

The RCA results were compared with the RHS indices stored in the ‘Indices’ form. In every RHS class 1 is best and 5 is worst:

Agreement is a Spearman rank correlation, signed so that a positive value means the two measures agree. The RCA values are medians, from the best to the worst RHS class.

RHS index vs RCA output

Sites

Agreement

Median RCA value by RHS class (best → worst)

RHQ vs Preliminary score

187

0.59

III 1.50 → IV 0.87 → V 0.39 (I and II: one site each)

RHQ vs final class

187

0.56 (0.51 before overdeepening)

III Moderate → IV Fairly Poor → V Fairly Poor

HMS score vs mean negative score

190

0.59 (HMS class: 0.45)

HMS 0–16: −0.12; 17–199: −0.15; 200–499: −0.38; 500–1399: −0.54; 1400+: −0.92

Baseline HQA class vs mean positive score

187

0.77 (HQA score: 0.87)

1.95, 1.82, 1.58, 1.37, 1.11

HIR vs Preliminary score

125

0.10 (not significant)

class 2: 0.70, 3: 0.99, 4: 0.46, 5: 0.58

HIR vs final class

125

0.08 (not significant)

Fairly Poor in every HIR class


Positive indicators and HQA agree closely, because both reward the variety of natural features on the banks and in the channel. The indicators closest to HQA are D4, E4 and E3 (about 0.6). The vegetation indicators D1, D2 and E1 are unrelated to HQA, which scores channel vegetation differently.

Negative indicators and the HMS score agree moderately (0.59). C7 (0.59), B5 (0.55), E10 (0.46) and C8 (0.42) follow the HMS score. Most of the gap comes from two differences. B4, C10, E7 and E12 (invasive plants, silt and algae) are not part of HMS, and B5 counts land use, which HMS does not score. Using the HMS class instead of the score lowers the agreement to 0.45, because 154 of the 190 sites fall in class 5 (HMS 1400+).

RHQ agrees with the RCA in ranking sites, but the RCA sits about one class higher until the overdeepening step is applied. The matrices below assume that RHQ I to V correspond to RCA Good to Poor; cells where the two match are in bold. They cover the 187 sites that have both an RHQ class and a River Type.

RHQ class vs RCA class before the overdeepening step



RCA class




RHQ

class


Good

Fairly Good

Moderate

Fairly Poor

Poor

Total

I Excellent

0

1

0

0

0

1

II Good

1

0

0

0

0

1

III Moderate

0

9

11

0

0

20

IV Poor

0

6

33

9

0

48

V Very poor

0

1

55

56

5

117


Total

1

17

99

65

5

187

RHQ class vs RCA final class



RCA class




RHQ

class


Good

Fairly Good

Moderate

Fairly Poor

Poor

Total

I Excellent

0

1

0

0

0

1

II Good

1

0

0

0

0

1

III Moderate

0

9

11

0

0

20

IV Poor

0

2

15

24

7

48

V Very poor

0

0

15

48

54

117


Total

1

12

41

72

61

187

Before the overdeepening step, the classes match at 25 of the 187 sites (13%), and 125 (67%) are within one class. The RCA class is better than the RHQ class at 161 sites. After the step, they match at 89 sites (48%), 170 (91%) are within one class, and the RCA class is better at 90 sites and worse at 8. No RHQ III site falls below Moderate, and 102 of the 117 RHQ V sites end up Fairly Poor or Poor. Much of the final agreement therefore comes from one RHS entry, extensive overdeepening, rather than from the 32 indicators.

The RCA is more optimistic than RHQ. Before the overdeepening step, the RCA class is higher than the RHQ class at 161 of the 187 sites (86%) and lower at only 1. Even after the downgrade, it is still higher at 90 sites (48%) and lower at 8 (4%). The strict typing method shows the same bias, with the RCA class higher at 154 of 178 sites before the downgrade and at 91 after it. RHS data on the same sites therefore rate river condition better under the RCA than under RHQ, and only the overdeepening rule brings the two scales closer together.

HIR shows no relation to either RCA output. HIR measures how far substrate, flow types, channel vegetation and geomorphic activity have departed from their predicted semi-natural values, which is a measure of altered processes. The RCA instead counts features and modifications, and apart from the overdeepening step it does not test whether the channel still behaves as its type should. HIR is also very high (class 5) at 92 of the 125 sites, which limits the contrast.

The dataset is dominated by modified sites, with few in the best RHS classes, so these correlations describe the lower half of the condition range only.

8.3 Typing with and without the semi-natural substrate prediction

Two ways of typing the 201 test sites were compared. Both use the same desk-study proxies (A1–A5) and the same 32 indicators.



Toolbox method

Strict RCA method

Sites with a River Type

197

187

Sites without a type

4 (no substrate data and no prediction)

14 (no substrate recorded)

River Types

F 97, H 62, K 18, G 7, D 6, C 3, I 3, L 1

F 97, H 42, K 27, D 7, G 4, I 4, L 3, C 2, B 1

Agreement with RHQ, class before overdeepening

0.51

0.50

Agreement with RHQ, final class

0.56

0.54

Final class equal to the RHQ class

89 of 187 (48%)

77 of 178 (43%)

Final class within one class of RHQ

170 (91%)

160 (90%)


The two methods agree on the River Type at only 117 of the 187 sites typed by both (63%). Cohen's kappa is 0.43, which is only moderate agreement once the matches expected by chance are allowed for. Counting the 4 sites that neither method can type, overall agreement is 121 of 201 (60%; kappa 0.42).

The types that agree best are D (Plane bed, 6 of 6) and C (Step-pool, 2 of 3). Agreement is weaker for F (Straight, sinuous gravel/pebble, 71 of 97) and K (Straight, sinuous silt/clay, 12 of 18). It is poor for H (Straight, sinuous sand, 22 of 62) and for the meandering types G, I and L.

The main swaps are:

Under the strict method, 10 sites typed by the toolbox method get no type: 8 H and 2 F sites where no substrate was recorded. One F site becomes B (Cascade).

All of these changes come from A8:

River Type from the toolbox method vs the strict RCA method (201 sites; matching types in bold)



Strict RCA method type






Toolbox

method

type


B

C

D

F

G

H

I

K

L

No type

Total

B

0

0

0

0

0

0

0

0

0

0

0

C

0

2

1

0

0

0

0

0

0

0

3

D

0

0

6

0

0

0

0

0

0

0

6

F

1

0

0

71

0

15

0

8

0

2

97

G

0

0

0

0

3

0

2

0

2

0

7

H

0

0

0

25

0

22

0

7

0

8

62

I

0

0

0

0

1

0

1

0

1

0

3

K

0

0

0

1

0

5

0

12

0

0

18

L

0

0

0

0

0

0

1

0

0

0

1

No type

0

0

0

0

0

0

0

0

0

4

4


Total

1

2

7

97

4

42

4

27

3

14

201


Overall the two A8 values agree at only 110 of the 187 sites (59%). A8 also changes the Preliminary Condition Score at 28 sites, through E7, whose silt penalty applies only when A8 is gravel or coarser. Even so, the final class is the same at 175 of the 187 sites (94%). The choice of method therefore matters more for the River Type reported than for the class. Sections 8.1 and 8.2 use the toolbox method, so its two RHQ matrices are the ones in 8.2.

Agreement with RHQ. After the overdeepening downgrade, the final RCA class equals the RHQ class at 48% of sites with the toolbox method and 43% with the strict method. On the 178 sites that both methods type and that have an RHQ class:

Before the downgrade, both methods match RHQ at only 13% of sites. Their kappa is −0.04, no better than chance (weighted 0.10 and 0.09), because the RCA class is higher than the RHQ class at about 86% of sites.

Using the semi-natural prediction therefore improves agreement with RHQ slightly, by 8 sites. In both methods, though, most of the agreement comes from the overdeepening downgrade rather than from the typing.

A more thorough analysis with a comparison to sites where both RHS and Morph were done will be published on the RHS website (www.riverhabitatsurvey.org)

Strict RCA method: RHQ class vs RCA class before the overdeepening step



RCA class




RHQ

class


Good

Fairly Good

Moderate

Fairly Poor

Poor

Total

I Excellent

0

1

0

0

0

1

II Good

0

1

0

0

0

1

III Moderate

0

12

8

0

0

20

IV Poor

0

7

29

11

0

47

V Very poor

0

1

55

50

3

109


Total

0

22

92

61

3

178

Strict RCA method: RHQ class vs RCA final class



RCA class




RHQ

class


Good

Fairly Good

Moderate

Fairly Poor

Poor

Total

I Excellent

0

1

0

0

0

1

II Good

0

1

0

0

0

1

III Moderate

0

12

8

0

0

20

IV Poor

0

2

15

21

9

47

V Very poor

0

0

16

46

47

109


Total

0

16

39

67

56

178

8.4 Fit of each indicator

Fit

Indicators

Likely difference from MoRPh

Good: same features, comparable scale

B4, C8, C9, C10, D2, E1, E3, E12

Small; D2 and E1 slightly higher (longer site)

Moderate: features present but extent or bank must be inferred

B1, C1, B2, C2, B5, C3, C4, C6, C7, D1, D4, E2, E5, E6, E7, E8, E9

B2, C2 slightly higher; C4, C6, D4, E6, E8 slightly lower; B1, C1, D1, E2, E5, E7 uncertain both ways; C3, C7 close to MoRPh; E9 depends on user input

Weak: key data missing or sampled very differently

B3, C5, D3, E4, D5, E10, E11

Lower for B3, C5, D3, E4 (spot-checks miss local features) and E11 (species not recorded); higher for D5 and E10 (counts over 500 m)


The weak group explains most of the expected difference. The positive features at risk are small and patchy, so a 1 m spot-check every 50 m often misses them. The negative indicators at risk count structures over 2.5 to 10 times the MoRPh length.

8.5 Sensitivity

The conversion constants change scores by known amounts. For example, rescaling the counts of negative features (weirs, culverts, outfalls, deflectors) raises the Preliminary Condition Score by 0.05 on average (at most 0.23). It changes the Type F class of 8 of the 201 sites, and cuts E10 = −4 from 65 sites to 5. The same test should be run for SWEEP_P, SWEEP_E and the B1/C1 richness rule before the method is used for BNG decisions.

8.6 Ways to improve the fit

    1. Paired calibration. Survey MoRPh5 and RHS on the same reaches, about 30 or more across River Types. Use the pairs to tune the constants, or to set RHS-specific score thresholds, as Gurnell et al. (2020) did with their 40 calibration subreaches.
    2. Score the worst part of the site. Instead of the whole 500 m, compute the indicators on the most degraded run of spot-checks matching the MoRPh5 length (1 pair for streams under 5 m wide, 2 pairs under 10 m, and so on). This matches the guide's requirement that a subreach capture the most degraded part, and removes most of the length bias in counts.
    3. Add a few RCA fields to the RHS form. Bench and toe, steps, trees growing from the bed, bed reinforcement material, bridge piers in the bed, tributary junctions, functional outfalls, NNIPS on bars and islands, and a Trace class for key features. Most could be optional add-ons, like the existing RHAT and urban modules.
    4. Record confidence. Flag each indicator estimated from a default or proxy (E9 without a user answer, E7 without a silt-drape record, counts with no width) and report the share of indicators at full confidence beside the class.
    5. Cross-check with RHS indices. Compare the RCA class with the Habitat Modification Score, the Habitat Quality Assessment and the RHS overdeepening and realignment entries; large disagreements show where the conversion fails.
    6. Fix and verify the typology inputs. Add a rule for sites where the predicted dominant substrate (A8) is coarser than the coarsest one recorded (A7), or where no bed was seen. Check NGR1 and NGR11 in the database so that sinuosity can be derived for every site. Then compile and test the VBA in Access against the Python results.

References

Environment Agency (2003). River Habitat Survey in Britain and Ireland: Field Survey Guidance Manual, 2003 version.

Gurnell, A.M., England, J., Scott, S.J., Shuker, L.J. et al. (2024). A Guide to Assessing River Condition: Part of the Rivers and Streams Component of the Biodiversity Metric. Updated January 2024.

Gurnell, A.M., Scott, S.J., England, J., Gurnell, D.J., Jeffries, R., Shuker, L. and Wharton, G. (2020). Assessing river condition: a multiscale approach designed for operational application in the context of biodiversity net gain. River Research and Applications, 36, 1559–1578. doi:10.1002/rra.3673

Gurnell, A.M. and Downs, P.W. (2021). The legacy of river channel modification in wadeable, lowland rivers: exploring overdeep rivers in England. Earth Surface Processes and Landforms, 46(15), 3016–3025. doi:10.1002/esp.5268

Modular River Survey (2022). MoRPh Survey Manual, version 14, October 2022.

Shuker, L.J., Gurnell, A.M., Wharton, G. et al. (2017). MoRPh: a citizen science tool for monitoring and appraising physical habitat changes in rivers. Water and Environment Journal, 31(3), 418–424.

Defra and Natural England (2026). The Statutory Biodiversity Metric: User Guide, June 2026.