Trees, large wood and streams: Using archive survey data to inform changing interactions in a human‐impacted landscape

Monitoring is the key to understanding fluvial systems and a crucial foundation for assessing the outcomes of river restoration. The New Forest, southern England, was designated a National Park in 2005 in recognition of its highly valued landscape, which has been both positively and negatively impacted by over 1000 years of human management. Here we analyse archive field maps and tabulations extracted from walkover surveys that record the distribution and character of wood jams along ~59 km of New Forest streams in 1991. By integrating the 1991 archive survey data with other historical information, we analyse associations among stream channel characteristics, wood jams, riparian land cover, and stream and land management at that time. We reveal associations among these factors that reflect the imprint of centuries of grazing, forestry and stream management practices. Along ~10 km of one stream (the Highland Water), we analyse data collected during additional walkover surveys in 1997 and 2021, to track changes since 1991. We illustrate how a reduction in stream and wood management over recent decades and the restoration of some stream reaches in 2005 has resulted in overall increases in stream sinuosity and the number of wood jams. The walkover surveys on which our results are based provide an approach to characterizing the impacts of river corridor management in the New Forest landscape that are both cost and time effective and could be applied by non‐specialist volunteers (‘citizen scientists’) following appropriate training. Survey data of the spatial coverage achieved in 1991 are rare but need to be encouraged in the New Forest and farther afield to provide a robust framework into which more local, specialist surveys can be integrated, and from which the broad impacts of river corridor management approaches can be monitored.


| INTRODUCTION
Scientific research investigating the role of wood in river and stream ecosystems commenced in the 1970s. Much early work was conducted in the Pacific Northwest of the United States, focusing on forested catchments under native, mainly coniferous, tree species.
Since then, research has investigated the hydrological, geomorphological and ecological roles of large wood within a wide variety of forest types across the world in both semi-natural and managed contexts (Gurnell & Bertoldi, 2022;Swanson et al., 2021). Following the proposal of the river continuum concept (Sedell et al., 1989;Vannote et al., 1980), numerous field investigations have refined concepts of how trees and wood influence river channel and floodplain ecosystems. These include the floodplain large wood cycle hypothesis (Collins et al., 2012), the natural wood regime in rivers , and proposals regarding how these natural dynamics are disrupted by human actions (Collins et al., 2012;Scott & Wohl, 2018).
The impacts of humans on river systems have become more intense, varied and widespread during the Anthropocene (Brown et al., 2017;Downs & Piégay, 2019). Wohl (2014Wohl ( , 2019 discusses the legacy effects of human alterations of river corridors, including manipulation of large wood. She stresses the importance of recognizing the existence, nature, timing and spatial distribution of human alterations, as well as natural processes, on river-floodplain morphodynamics, to inform river restoration design. Recently, Gurnell and Hill (2021) presented a multi-scale investigation of the legacy effects of human activities on stream environments in the New Forest, southern England. Their study revealed both positive and negative impacts of humans on this sensitive, predominantly wooded landscape, which has been subject to a variety of land and river management practices for over 1000 years. In particular, an historical analysis of a $10 km reach of the Highland Water revealed the negative effects of historical woodland and channel management practices, and the recovery of a stream section restored in 2005.
In this article, we further investigate tree-wood-management interactions in the New Forest, focusing on the stream network (i.e., channels typically < 10 m wide, though predominantly < 5 m) of the wider Lymington River catchment in which the Highland Water reach is located (Figure 1). We revisit a walkover survey of $66 km of streams conducted in 1991, which was previously analysed by Gregory et al. (1993) to identify spatial patterns in the load of large wood (kg.m À2 channel area) retained in the streams. The 1991 survey has not been repeated in full in subsequent years, but as an example of the potential information that could be revealed from analysis of repeat surveys, we have extracted the same set of variables collected in 1991 from surveys of $10 km of the Highland Water undertaken in 1997 and 2021 (Gurnell & Hill, 2021). The straightforward nature of the surveys allows us to consider the potential of additional surveys being repeated by river scientists as well as by volunteers ('citizen scientists') following appropriate training. In the New Forest and other locations, such a 'crowd-sourcing' approach (e.g., Gurnell et al., 2019;Shuker et al., 2017) would allow much larger areas to be surveyed than would otherwise be possible in a cost-and time-effective manner, with the repeat surveys potentially identifying responses to changing river corridor management practices.
Analysis of remotely-sensed data is increasingly reducing the need for traditional ground surveys (Tomsett & Leyland, 2019), but this is not always a solution for all environments and survey types.
Airborne remote sensing approaches have been devised to monitor changes in wood retention in rivers, particularly the analysis of LiDAR (light detection and ranging) data (e.g., Abalharth et al., 2015;Atha & Dietrich, 2016). However, such approaches are not currently viable on the small, lowland streams investigated here, which are < 10 m wide and are usually overhung by a dense tree canopy. Furthermore, even if wood accumulations could be identified by such means, additional ground survey would be needed to distinguish between different large wood jam types and/or fallen trees (Figure 2), many of which are associated with distinctive physical habitat assemblages and dynamics (Gregory et al., 1985).
In this article we analyse the aforementioned field-survey data sets to address the following specific aims: i. To use the areally extensive 1991 surveys and related archive data to examine inter-relationships and interactions among the characteristics of the predominantly single-thread surveyed stream channels, the number and types of wood accumulations and fallen trees, and the nature and management of riparian land cover in the New Forest at that time.
ii. To compare data extracted from repeat surveys of the Highland Water in 1997 and 2021 to illustrate spatial and temporal trends in stream channel characteristics, wood jams and fallen trees, and F I G U R E 1 (a) Location of the Lymington River catchment within the New Forest in southern England. (b) Locations of the surveyed streams within the Lymington River catchment (127 reaches) that were surveyed in 1991 and within the Highland Water (22 reaches) that were resurveyed in 1997 and 2021. Unsurveyed streams are those that either were not included in the original 1991 survey (i.e., ephemeral headwater streams, streams draining catchments lacking woodland, streams where access was not possible), or were parts of the 6.7 km stream length for which original maps and tables from 1991 are no longer available so identify the impacts of changing riparian land cover and stream management practices.
iii. To discuss how our findings support and extend previous research regarding trees, large wood and streams, and how walkover surveys could provide a broad scale framework for monitoring streams and their responses to changing river corridor management practices.

| THE STUDY AREA
The study area within the New Forest, southern England, is the second most densely populated National Park in England after the South Downs (63 people per km 2 ). Gurnell and Hill (2021) reviewed the landscape character and management of the New Forest that is relevant to its stream systems. We summarise their findings in this section.
The New Forest is underlain by Eocene deposits of the Bagshot gravels and sands, the Barton sands and clays, and in the south by the Headon marls. Locally, these are overlain by Pleistocene river gravels. The New Forest landscape is gently undulating with abundant peat accumulations forming valley and hillslope mires. Streams generally have gravel and finer bed materials with some local exposures of clay.
The New Forest was subject to land clearance for agriculture and settlement long before it was established as a hunting forest by King William in 1079 CE. Since then, in addition to hunting, the land has been used continuously to the present by 'commoners' (i.e., persons with rights of commons; Short, 2008) for grazing their animals but also in past centuries for the collection of wood and peat for fuel.
Woodland is a characteristic feature of the landscape, and includes both natural and planted stands. Planting trees to produce timber commenced in the 17th century and the plantations were usually fenced to exclude grazing animals. In 1877, the New Forest Act formally permitted 'the Crown' to enclose (fence off from grazing) a maximum fixed area of land for timber production (Statutory Inclosures), formalizing a complex mosaic of land cover and management that persists to the present with open (unenclosed) grazed areas referred to as 'Open Forest'.
The Inclosures were planted initially with native deciduous hardwood species, particularly beech and oak, but from the 1930s, commercial, predominantly non-native, conifer species were also planted.
Over the same timeframe, a mix of heathland, grassland ('lawns') and 'ancient and ornamental woodland' containing predominantly native, deciduous, hardwood species have persisted within the Open Forest, including water-related habitats such as wet woodland and carr woodland adjacent to valley mires (Grant & Edwards, 2008). The land cover mosaic in the Open Forest is maintained by grazing by wildlife (e.g., deer) and the animals of 'commoners', particularly by cattle and horses.
Over the last two centuries, deliberate human interventions in the streams of the New Forest have expanded dramatically. Initially, these interventions were aimed at 'improving' land drainage. In Inclosures, drainage schemes were especially widespread. Streams were frequently realigned (straightened), re-sectioned (deepened) and networks of feeder drains were often cut prior to planting single species stands of trees up to the stream channel margins. In the Open Forest, many valley mires and lawns were also subject to drainage schemes to improve grazing.
F I G U R E 2 Examples of the four wood jam types that were mapped. (a) Active (note the higher upstream (left) than downstream (right) water level, (b) Complete, (c) Partial, (d) High. All photographs by A.M. Gurnell [Color figure can be viewed at wileyonlinelibrary.com] Over the last $30 years, however, drainage schemes have given way to stream and wetland restoration programmes. The latter have intensified since the designation of the New Forest as a National Park in 2005. Our spatially extensive 1991 survey, and the more focused 1997 and 2021 Highland Water surveys, thus collectively span a period during which there was a move away from fallen tree and wood jam management and a move towards restoring streams back to a more natural form. In 1991 there had been no stream restorations and the main direct human impacts on streams were related to drainage 'improvements' and indirect impacts from forestry operations.
Following a severe wind storm in 1987 and a lesser storm in 1989 that caused widespread tree fall, there had been clearance of many of the largest trees that had fallen across or into New Forest streams.
This added to a history of removing or reducing the size of some of the largest wood jams to improve flow conveyance by the stream network. The 1991 survey, and to some extent the 1997 survey, record the character of New Forest streams at the end of this accumulated management history. In 2005, two of the surveyed reaches of the Highland Water were subject to a major restoration involving: (i) clearance of a non-native conifer plantation on the floodplainstream margins that had been planted c. 1960; (ii) leaving any native deciduous trees that were present on the floodplain-stream margins; (iii) returning the straight, incised stream channel that had been cut c.
1960 to accompany planting of conifers to its previous sinuous course; and (iv) adding gravel to the channel to raise its bed to the level that existed prior to straightening and incision. A further four reaches that had been straightened in the early 19th century to accompany planting of native deciduous trees continued to recover

| Wood jam types
Our analysis of the changing interactions between trees, large wood and streams is based on the separation of wood accumulations into active, complete, partial and high types. These wood jam types are widely observed in small, lowland streams and rivers (Gregory et al., 1985) and have different hydraulic impacts ( Figure 2 with stream flows when the water is sufficiently deep to fill the area under the jam. To qualify as a high jam, the contact water depth must be less than or equal to the bankfull channel depth (see Figure 2d).
Fallen trees are those that span streams but remain suspended above the bank full level, supported by other trees or by their branches.
3.2 | Extracting a data set for analysis from the 1991 archive materials The 1991 survey maps locate the end points of reaches ($500 m length) used in the original analysis (Gregory et al., 1993) and indicate the locations of active, complete, partial and high wood jams and fallen trees. In places, the maps are annotated to indicate stream channel width, land cover types, and contemporary management actions. Despite the 30 year gap since they carried out the field surveys, two of the co-authors of this article (RJD and ST) were able to provide clarifications that helped to assemble and interpret the data analysed in this article.
In the present analysis, the end points of the original $500 m stream reaches were retained. The numbers and types of wood jams and the numbers of fallen trees in each of the reaches were extracted from the maps, as was information on potential controlling factors on wood input/retention within each reach including channel characteristics, proportions of channel margins that were within an Inclosure, and proportions of channel margins that were under five land covermanagement types (Table 1). Although our core data came from the original survey maps, wherever possible cross-checks were made using other data sources. The reach centre-line length and straightline length (i.e., straight line length between the upstream and downstream ends of the reach), which were used to estimate sinuosity,

| Data analysis
The channel centre line lengths extracted from 1:2500 Ordnance Survey maps and, where necessary, corrected for known differences due to channelization/restoration of channel planform at each survey date, were used to adjust the observed wood jam and fallen tree numbers to a standard 500 m reach (centre-line) length (Table 1) prior to further analysis.
All data analyses were conducted using XLSTAT version 2020.5.1.1079 (Addinsoft, 2021). Inter-relationships among the assembled variables from the 1991 survey were explored through scatter plots, cumulative line plots, correlation analyses, and principal components analysis (PCA) of sets of variables selected from those listed in Table 1. across PC biplots as well as highlighting spatial differences in plotting positions of sections (groups of reaches) affected by different land cover-management. Bar graphs were then constructed to represent the changing temporal PC scores of each Highland Water reach from upstream to downstream on each of PC1, PC2 and PC3. These graphs were examined for evidence of both temporal (1991,1997,2021) and spatial (upstream to downstream) trends in reach PC scores. They were also examined for further evidence of contrasting temporal and spatial patterns in the PC scores within sections of the Highland Water that had been the subject of different land cover-management.
The statistical significance of changes in the PC scores across the Highland Water reaches between monitoring years was tested using paired t tests. Finally, the wood jam data alone were considered. The total number of wood jams and the number recorded as active, T A B L E 1 Variables quantified for each of 127 stream reaches included in the 1991 survey

Variable name
Units and explanation

Channel dimensions
Reach (centre-line) length (m) The length (in metres) of the channel centre-line measured from 1:2500 scale Ordnance Survey maps (approximate field survey date -1960) with corrections for any major changes (mainly channel realignments) to match the 1991 survey date.

Reach (start to end point straight line) length (m)
The straight-line length (in metres) between the upstream-downstream 500 m reach end points.
Distance from source Cumulative channel centre-line length (in metres) from the stream source.

Sinuosity
Channel length/reach length

Wood jams and fallen trees
Active jams Number of active jams recorded in each reach expressed as active jams per 500 m reach (centre-line) length.
Complete jams Number of complete jams recorded in each reach expressed as complete jams per 500 m reach (centre-line) length.
Partial jams Number of partial jams recorded in each reach expressed as partial jams per 500 m reach (centre-line) length.
High jams Number of high jams recorded in each reach expressed as high jams per 500 m reach (centre-line) length.

Felled
Proportion of bank tops where woodland was recently felled.
Heath-scrub-lawn-mire Proportion of bank tops under heath-scrub-lawn-mire with limited tree cover.
Fields-gardens Proportion of bank tops occupied by fields or gardens.
complete, partial or high per 500 m reach (centre-line) length in each of the 22 reaches in 1991, 1997 and 2021 were investigated using line plots and cumulative relative frequency distributions.

| Data description
From the 1991 survey, data were extracted for 127 reaches of the original 138 surveyed (Figure 1b), covering a total of $59 km stream length. Original survey data were not available for 11 of the reaches.
In some places on the survey maps, measurements of channel width had been noted, indicating a median channel width of 2.8 m and a range from < 1 m to $10 m. Descriptive statistics for the continuous variables where information was available for all reaches ( was along recently felled woodland, 18% was bordered by heathscrub-lawn-mire and 1% was bordered by field-garden land cover. Data were also extracted from the 1997 and 2021 surveys of the Highland Water that contained 22 of the original 1991 surveyed reaches ( Figure 1b). These data were combined with the 1991 data set to enable temporal and spatial trends to be identified. These 22 reaches represent streams whose distance from source at their downstream end ranges from $1.6 to 12.3 km, and sinuosity ranges from 1.1 to 1.9. The reaches can be grouped from upstream to downstream into seven sections (section A through to section G Two channel characteristics were considered for inclusion in an analysis of the potential controls on wood retention in lowland streams: distance from source and sinuosity. Distance from source provides a surrogate for catchment area and thus likely river flows, and also channel size, the latter being a key influence on a channel's T A B L E 3 Spearman's rank correlation coefficients estimated between pairs of variables indicative of (a) channel dimensions; (b) wood jams and fallen trees; (c) land cover and management (all variables selected from and expressed in the units listed in Table 2  ability to retain wood. Direct measurements of channel size were not included in the analysis as measurements were not available for all of the reaches. Planform sinuosity also influences a channel's ability to snag and retain large wood but sinuosity shows only a weak association with distance from source (Table 3) and so provides additional information as a second channel dimension variable for integrative analysis.
The 1991 archive materials provide data sets regarding the types and frequencies of wood jams and fallen trees retained by the surveyed streams, and also the extent of different types of marginal land cover management. These two data sets were separately sorted before being summarized in cumulative line plots (Figure 3) to give an impression of both variability and trends across the surveyed reaches.
In the case of wood jams and fallen trees (Figure 3a), the data were sequentially sorted in order of likely hydraulic influence from least to most: the number of fallen trees, high jams, partial jams, complete jams, and active jams as a percentage of the total number within each reach. Figure 3 (Table 3).
Given the strong inter-correlations among the measured variables, PCA was used to reduce the dimensionality of the data set and to identify the broad gradients and inter-relationships within it. PCA was applied to all the variables listed in Table 3, and to the estimated age classes for the deciduous and coniferous woodland (Table 1). Variables reach (centre-line) length and reach (start to end point straight line) length (listed in Tables 1 and 2) were excluded because their ratio is used to calculate sinuosity, which is included. The eigenvalues, variance explained and variable loadings for the first four PCs (all those whose eigenvalues exceed 1) are presented in Table 4. The variable loadings and reach scores on PC1 to PC3 are presented in Figure 4.
The variables with the strongest loadings on PC1 (Table 4, bold numbers) indicate that this component describes a gradient of increasing coniferous woodland cover and age and decreasing deciduous woodland cover. PC2 describes a gradient of decreasing fallen trees and increasing heath-scrub-lawn-mire cover. PC3 describes a gradient of decreasing distance from source and increasing numbers of active wood jams. These three PCs explain over half (52.5%) of the variance in the data set. PC4 has a high loading on fields-gardens but the eigenvalue for this PC is only slightly larger than 1 and this land cover type is only present along a very small length (1%) of the surveyed channels ( Figure 3b, Table 2), so this component was not considered further.
The variable vectors on the PC1-PC2 biplot (Figure 4a) separate heath-scrub-lawn (upper half of the plot) from the two woodland types, wood jams and fallen trees (lower half of the plot). The coniferous woodland vector is closely associated with (plots close to) the Inclosure vector. The deciduous woodland vector is located close to the sinuosity vector; partial jams, active and high jams plot closer to deciduous than coniferous woodland; and complete jams and fallen tree vectors are positioned roughly equidistant between the two woodland types. The proximity of the distance from source vector to the deciduous woodland vector suggests that the latter, particularly older age classes, may be observed more with increasing distance from the stream source.
The PC1-PC3 biplot ( Figure 4b) illustrates a strong inverse association between distance from source and active, complete and high wood jams, with the latter increasing in number towards a stream's    However, in 1997 virtually all of the jams were partial or high whereas by 2021four active and approximately two complete jams were present per 500 m. Straightened section E shows notable increases in high (from $5 to $10 per 500 m) and complete (from $2 to $5 per 500 m) jams between 1997 and 2021 with a smaller increase (from $2 to $3 per 500 m) in active jams. Although section E has not been restored, its reaches retain more jams than reaches in section C, and only slightly less than would be expected from the downstream trend in total jams through sections B, F and G, which are not in Inclosures.

| DISCUSSION
We have shown how a carefully assembled data set, extracted from $30 year old survey maps and tabulations, has supported an informative analysis of associations among wood jams and fallen trees and two groups of controlling factors (distance from source/sinuosity and land cover type and management) along $59 km of streams in the Lymington River catchment. By combining this data set from 1991 with more recent survey data (1997,2021), our analysis has also revealed changes through time along $10 km of the Highland Water in both relatively naturally functioning and more managed sections, but also notable differences in the sections subject to different types and levels of management. These findings can be used to support and extend previous investigations into the natural and anthropogenic factors controlling the spatial and temporal patterns of large wood in streams, and our approach has wider implications for the monitoring of river restoration schemes involving large wood, both in the New Forest streams and farther afield.

| Supporting previously-recognized distributions of large wood in streams
Our analysis has identified patterns within the 1991 data set that support previous understanding of wood in stream channels. First, with increasing distance from source (i.e., as channels increase in size) the number of wood accumulations per unit channel length tends to decrease, and second, in these small channels the wood is not transported very far once it enters the channel, since relatively few wood jams were observed in channels not bordered by woodland. These tendencies are captured by PC2 (Figure 4a), which identifies a gradient of decreasing numbers of fallen trees and wood jams with increasing heath-scrub-lawn-mire (i.e., non-woodland) cover and by PC3 ( Figure 4b) which describes a gradient of increasing numbers of jams, particularly active jams, with decreasing distance from source. These characteristics of in-channel wood distributions have long been recognized. Indeed, some of the earliest research on wood in channels revealed inter-relationships among wood supply, wood piece sizes and jams, channel dimensions and the flow regime (e.g., Bilby & Ward, 1989, 1991Meehan et al., 1977). These inter-relationships highlighted their importance for stream ecosystems within the River Continuum Concept (Vannote et al., 1980), and these topics have been subsequently revisited and refined (e.g., Wohl & Jaeger, 2009).

| Revealing associations among trees, wood, land cover and stream management
Other associations have emerged from our analysis of the 1991 data set that relate wood jams to local land and stream management F I G U R E 6 Scores on PC1, PC2 and PC3 of the 22 reaches of the Highland Water included in the 1991, 1997, and 2021 surveys. The reaches are separated from upstream to downstream into sections A to G, which have been subject to different management practices practices. These associations enhance our understanding of the controls on stream-wood interactions in this long-managed landscape and are most clearly illustrated by the direction and length of the variable vectors in Figure 4.
Land and stream management in the New Forest has differed historically between Inclosures and Open Forest (Chatters, 1995;Tubbs, 1986). The Inclosure vector in Figure 4 the sinuosity vector (i.e., is associated with more sinuous streams) ( Figure 4a). This follows observations by Gurnell and Hill (2021) that streams are less likely to be modified in the Open Forest, but, small, unmanaged, heath-scrub-lawn-mire streams tend to be less sinuous than those flowing through Open Forest woodland. Within the analysed data set, streams bordered by heath-scrub-lawn-mire were frequently headwater streams, explaining the inverse relationship between this cover type and distance from source ( Figure 4b).
Vectors representing different types of wood jams and fallen trees are spread across the PC1-PC2 biplot (Figure 4a) along the gradient described by PC1 from coniferous to deciduous woodland. This indicates that their relative frequency changes along this gradient, with fallen trees and complete jams showing a similar association with both types of woodland (i.e., they are found in most woodland settings) whereas the proportions of active jams, high jams and partial jams tend to increase as the proportion of deciduous woodland increases. However, the PC1-PC3 biplot (Figure 4b) indicates that the frequency of active jams, complete jams, high jams and fallen trees decreases with increasing distance from source (i.e., increasing stream size) and that the largest streams appear to be associated with deciduous woodland in the Open Forest.
The fact that these associations emerge clearly within the space defined by the first three PCs demonstrates the pervasive influence of land cover and stream management on stream ecosystems in 1991.
This reflects a time when stream management in the New Forest was concerned mainly with 'improving' land drainage and when many Inclosures were primarily being managed for timber production.

| Decadal-scale temporal changes
By including a sequence of three surveys (1991,1997,2021)  Wood jams are now recognized as crucial components of woodland stream ecosystems within the New Forest, as advocated by Gregory and Davis (1992). Since 1991, wood management has markedly declined and land cover and stream restoration have become an increasing focus for stream corridor management across the New Forest, particularly within riverine 'Sites of Special Scientific Interest' (Mainstone & Wheeldon, 2016). Similarly, local Forest Design plans have sought to restore native deciduous woodland extent, function and processes to support habitats and species (Forestry Commission, 2017). The particularly large increases in the PC1 scores of reaches within section C ( Figure 6)

| The potential of walkover surveys for monitoring
In the broader context of stream and river management, our analysis points to the potential utility of coupling historical research on management practices with relatively rapid but appropriately designed, repeated walkover surveys. Particularly in the context of the United Nations "Decade on Ecosystem Restoration" (2021-2030), citizen science approaches have been recommended as one way of increasing local community engagement with, and longer term investment in, restoration efforts (e.g., Gann et al., 2019). In river and wetland contexts, many rapid assessment approaches for assessing channel and ecosystem health and/or monitoring restoration efforts have been developed (e.g., Gurnell et al., 2019;Kotze et al., 2019).
Application of some of these approaches have included volunteer citizen scientists and have been shown to generate robust data sets (e.g., Gurnell & Downs, 2021), often in a cost-and time-effective manner. In the New Forest for instance, additional field surveys of similar detail to our 1991, 1997 and 2021 surveys could easily be collected by citizen scientists following a brief (1-day maximum) training course, capturing decadal changes or responses to particular events (e.g., wind storms, floods). Following the training, > 5 km of stream could comfortably be surveyed in 1 day by pairs of volunteers, so that a large proportion of the Lymington River stream network (e.g., equivalent to the spatial coverage of the 1991 survey) could be covered at relatively little cost with 10 to 12 days of effort.
Our approach and analysis could be extended from a primary focus on wood jams to incorporate other in-channel features of importance to stream ecosystems such as pools, riffles, bars and bank forms. This would require more training of volunteers, and would inevitably add to the time taken in surveying each stream reach, so might adversely affect the spatial coverage achieved. Nevertheless, a nested survey approach could be designed, whereby broad spatial coverage using straightforward surveys by non-specialists is then supplemented by detailed surveys of a smaller number of reaches by specialist river scientists.
As Wohl (2019, p. 5181) emphasizes, 'the existence of forgotten legacies challenges river scientists to recognize the continuing effects of human activities that have long since ceased and also poses challenges for the application of scientific understanding to resource management'. Against this backdrop, walkover surveys clearly not only have the potential to provide a broad understanding of stream environments as a framework for more specialist local monitoring of human activities, including management interventions such as restoration, but application within citizen science frameworks is likely to be a key part of ensuring their wider application. Even for locations that have no existing archival materials, the surveys of today will become the archives of tomorrow, and undoubtedly will have analytical value in the decades to come.

| CONCLUSIONS
It is tempting to conclude that landscapes that appear natural are functioning naturally, but such assumptions are nearly always flawed in areas that have long-supported human populations and/or been subject to a long history of management. The New Forest presents one such case study. It is a unique landscape that has multiple conservation designations, is now under National Park management, and is enjoyed by its many visitors as an apparently 'wild' and 'naturallyfunctioning' environment. Nevertheless, its character and existence are a product of over 1000 years of human activities, including active management of woodlands and streams.
We have revealed some of the consequences of those activities for the New Forest stream channels through analysis of field survey data sets collected over the last 30 years. While many traditional management activities continue (e.g., extensive grazing) and contribute to maintaining highly valued habitats, other activitiesparticularly those applied to streams, their riparian margins and floodplainscontinue to be reversed through an increasing programme of restoration and a move towards management for naturally-functioning systems. Our extensive 1991 survey data and associated analysis provides a benchmark against which the broader effects of these restoration strategies and activities can be judged. Indeed, our later survey data sets (1997,2021) for the Highland Water show a trajectory of channel recovery following reduced wood and stream management, coupled with restoration of two reaches.
In a global context, our decades long data set regarding the changing interactions between trees, large wood and streams is very unusual, but the need remains for ongoing monitoring of stream recovery in response to past and ongoing management activities and the contribution of additional wood jams from future storm events. In the New Forest, and in many other human-impacted landscapes, the collection of well-designed but straightforward, rapid surveys of the type analysed here can provide a template for timely and repeated monitoring over decades, perhaps increasingly as part of citizen science projects. Particularly in view of the United Nations "Decade on Ecological Restoration", monitoring is crucial for permitting analyses that develop understanding of how streams function and change, not least those that experience legacy effects of human activities across different time and space scales.