Evidence of increase in woody vegetation in a river corridor, Northwest England, 1984–2007

Abstract

Based on observations on rivers in England over a 30-year period, it was hypothesised that the extent of woody vegetation in riparian areas has increased in many locations. Methods to map and measure changes in vegetation cover from aerial photographs were developed and tested on a 5 km reach of the River Dane in Northwest England. Riparian vegetation was divided into seven classes, and these were mapped from aerial photographs of four dates, 1984, 1996, 2001 and 2007. The results indicate an increase from 9% woody vegetation cover in the river corridor in 1984 to 32% cover in 2007, with a 40-fold increase in areas of mature woody vegetation. The results have implications for river processes, because of effects on bank erosion and sediment supply, and for land management policies, with grazing control hypothesised to be a major cause of the changes.

Keywords:

• river corridor
• riparian vegetation
• river channel
• aerial photographs
• vegetation cover
• vegetation change
• grazing impact
• decadal timescale

1. Introduction

Changes in riparian vegetation have important implications for river channel processes, for flood management and for river habitats. Increase in bank vegetation decreases rates of bank erosion and sediment supply (e.g. Beeson & Doyle, 1995; Corenblit, Steiger, Gurnell, Tabacchi, & Roques, 2009; Micheli & Kirchner, 2002) and in the longer term may influence channel pattern and behaviour (e.g. Crosato & Saleh, 2011). Increase in riparian vegetation, particularly woodland, can reduce flood capacity but also increases roughness thus slowing the passage of flood waves (Darby, 1999; James & Makoa, 2006). River corridors support a range of habitat types and are often key zones of biodiversity (Décamps et al., 2004).

Quantification of change in riparian cover is also important in terms of assessment and prediction of effects of agricultural and land management policies and practices, for example, the development and effectiveness of buffer zones for nutrient control (Ducros & Joyce, 2003; Howard-Williams & Pickmere, 2010; Mainstone, Dils, & Withers, 2008; Stockan, Langan, & Young, 2012) and concerns over fine sediment inputs to rivers affecting fish populations (Kemp, Sear, Collins, Naden, & Jones, 2011). Statutory agencies, therefore, need evidence to develop vegetation management policies along river corridors (Collins et al., 2010; Defra, 2012). In terms of biological conservation, there is great concern at the loss of total area of semi-natural habitats, and the extent to which it has become fragmented with a consequent loss of connectivity (Hooftman & Bullock, 2012). River corridors are prime locations for maintenance of such connected habitats, and riverbanks can be refuges (Smart et al., 2005). The presence of riparian vegetation delivers many benefits, and much recent research has also been stimulated by a need for greater understanding of dynamics in order to implement protection and restoration (Bennett, Wu, Alonso, & Wang, 2008; Richter & Richter, 2000; Sandercock & Hooke, 2010; Schirmer et al., 2014; Tabacchi, Steiger, Corenblit, Monaghan, & Planty-Tabacchi, 2009).

Monitoring and observation of several watercourses (Hooke, 2008) suggested that large increases in woody vegetation cover had taken place in recent decades. Some data are available on riparian vegetation change over this period but have produced differing national trends for England and Wales (Norton et al., 2012; Seager, Baker, Parsons, Raven, & Vaughan, 2012). Therefore, the aim here was to develop methods of mapping this type of vegetation and test the hypothesis that woody vegetation has increased along a case-study river. It would allow quantification of changes and timing of any increase, and insights into possible contributory causal factors and relevance of these changes for wider national trends.

Methods for mapping and analysing channel changes have been widely developed and applied (Garófano-Gómez et al., 2013; Gurnell, 1997; Hooke, 1984; Hooke & Yorke, 2010). Likewise, vegetation changes have been studied both in riverine and other environments, many studies using satellite imagery (Bertoldi, Drake, & Gurnell, 2011) but many, especially on decadal timescales, have used aerial photographs, as here (Dixon & Johnson, 1999; Garófano-Gómez et al., 2013; Gonzalez, Gonzalez-Sanchis, Cabezas, Comin, & Muller, 2010; Kondolf, Piegay, & Landon, 2007; Lunt, Winsemius, McDonald, Morgan, & Dehaan, 2010; Yang & Tien, 2010). Increases in woody riparian area have been found but many of the cases are associated with change in river regime (Dixon & Johnson, 1999; Johnson, 1994; Perona, Molnar, Savina, & Burlando, 2009), especially due to dam control (Dixon, Johnson, Scott, Bowen, & Rabbe, 2012; Gonzalez et al., 2010; Merritt & Cooper, 2000). Other causes of change such as land-use management impacts have been investigated or demonstrated (Sawtschuk, Delisle, Mesmin, & Bernez, 2014), as well as climatic and natural hydrologic variations (e.g. Stromberg, Lite, & Dixon, 2010). Types and zonation of riparian vegetation are strongly controlled by hydrological regime (Osterkamp, Hupp, & Stoffel, 2012; Sandercock & Hooke, 2010; Tabacchi et al., 2009), and occurrence of floods can affect recruitment and survival (Pasquale, Perona, Francis, & Burlando, 2014; Perona et al., 2009). Succession, particularly on river bars, is well known (Corenblit et al., 2009). Much recent research has focused on the mechanisms and dynamics of growth and survival and interaction with processes (Bertoldi et al., 2009; Crouzy, Edmaier, Pasquale, & Perona, 2013; Gilvear, Francis, Willby, & Gurnell, 2007), including research on Salix species (Pasquale et al., 2014).

2. Study area

The area studied is a 5 km reach of an active, meandering river in Northwest (NW) England, which has been monitored for 30 years (e.g. Hooke, 2008). It is a Site of Special Scientific Interest (SSSI) in Great Britain, designated for its fluvial geomorphology (Gregory et al., 1997). Lateral channel movement is rapid, with active bank erosion reaching 1 m a⁻¹ on many bends; new areas of depositional point bars are created each year, over time developing into mature floodplain at 1.5 m above river bed level. Most of the woody vegetation on the point bars is willow (Salix spp.). Isolated trees along the river banks are mostly alder (Alnus spp.). Some older hedges and isolated oak and ash trees are also present on the valley floor. Old mixed woodland is present on the steep valley walls.

3. Methods

3.1. Data sources

Vegetation monitoring through remote sensing can be achieved using satellite imagery (e.g. Lillesand, Kiefer, & Chipman, 2007), but can have limitations in resolution and detection of subtle vegetation change. Large-scale aerial photography was used here as it is often better for delimiting certain categories of vegetation cover (Akasheh, Neale, & Jayanthi, 2008; Bradter, Thom, Altringham, Kunin, & Benton, 2011; Xie, Sha, & Yu, 2008). Moreover, for long-term comparisons various types of remote sensing, particularly high-resolution images, are not available for early dates. A geographical information system (GIS) was used for mapping, quantification and analysis of changes (Sawtschuk et al., 2014; Yang & Tien, 2010), as applied here.

Vegetation was mapped from a time sequence of aerial photographs (1984, 1996, 2001 and 2007) commissioned from Cambridge University Aerial Photography Unit. All photography was taken in April/early May, at low flow and at scales between 1:4460 and 1:7000. Bank lines of the river were mapped from the 1984, 1996 and 2001 photographs by a photogrammetrist using a stereo-digital photogrammetric plotter (Kern DSR 14 analytical plotter). The photographs were geo-referenced using a minimum of eight ground control points per photograph which were surveyed to 2 cm accuracy using a differential global positioning system receiver. Bank lines were calculated to be accurate to <1 m. The 2007 photographs were processed slightly differently; they were scanned using an aerial photograph scanner at a resolution of 21.2 µm (a pixel resolution of 0.095 m), and then each photograph was geo-referenced using a minimum of 25 ground control points from both field survey and Ordnance Survey Land-Line data. Root mean square errors (m) were in most cases <1 m and were consistent with the 1984–2001 photogrammetric plotting. The 2007 photographs were mosaiced and the bank lines digitised using Esri ArcGIS (Hooke & Yorke, 2010).

3.2. Vegetation mapping

Methods to classify the vegetation automatically using ERDAS unsupervised and supervised classification techniques were tested with various spectral signatures (ERDAS, 2005; Lillesand et al., 2007). Neither method was successful because both techniques failed to distinguish between woody vegetation types and could only detect gross differences in land-use type. This study needed to detect more subtle changes within semi-natural vegetation so the vegetation was mapped in detail using categories that were identifiable on the air photographs. Seven vegetation categories were identified for non-pasture areas for assessing vegetation change within the river corridor (Table 1); two further categories were included for other areas (pasture/grassland and bare soil). The categories were tested by mapping from the most recent aerial photography (2007) and then ground-truthing with respect to the vegetation in the field (2011). When interpreting the aerial photography, stereoscopic viewing of the original photographs assisted in determining vegetation height of the woodland classes. Field checking shows that almost all the areas of new growth near the river were Salix spp.

Table 1. Vegetation categories used to map vegetation along the River Dane, Cheshire, between 1984 and 2007.

	Category	Definition
1	Pasture/grassland	Fields
2	Shrub/herbs	Rough texture, low vegetation, not pasture
3	Juvenile sparse woodland	<4 m high, >2 m between tree crowns
4	Juvenile Dense Woodland	<4 m high, <2 m between tree crowns
5	Mature, sparse woodland	>4 m high, >2 m between tree crowns
6	Mature, dense woodland	>4 m high, <2 m between tree crowns
7	Old woodland	Mixed trees on valley wall
8	Bank trees and linear	Lines of trees along river or old channel; hedges
9	Bare soil/other	No vegetation cover visible

For each sampling year, each vegetation patch along the river was digitised, heads-up at a viewing scale of 1:1000, as a polygon in ArcGIS and allocated one of the nine vegetation classes (Table 1); thus four date layers of the spatial distribution of the delimited vegetation classes were produced (Main Map). To ensure that the digitising and classification was performed accurately, it was checked by two independent researchers. Some difficulties in categorisation were encountered, especially in separating juvenile and mature trees; here the division was made at approximately 3–4 m tree height. A further problem was where trees were either scattered or quite sparse; here decisions had to be made over the extent of sub-division or joining of these patches, and as the trees grew some areas joined together. Isolated individual trees in the floodplain were not included, but lines of trees along the river bank were mapped. The maps produced for each date were then compared with each other. Blank areas are pasture/low grass/herbs on the floodplain and bare areas on the point bars.

To analyse change between categories, an intersect overlay operation was applied to the pairs of layers of different vegetation categories between different sampling dates one-by-one. Analysis focussed on patterns of recruitment from younger to older categories, but all combinations of category change were checked to detect any misclassifications, for example, change from mature to juvenile, and changes in the opposite direction, possibly due to floods removing vegetation or human clearance.

4. Results

Comparison of the maps for each date shows a large increase in total area of riparian woody vegetation (Main Map), from 4000 m² in 1984 to 150,000 m² in 2007, with the area of mature woodland increasing from 2868 m² in 1984 to 11,7961 m² in 2007 (Main Map figures). Taking the whole area of the valley floor for the measured reach, including all the pasture, then in 1984, 5% of the area was woody vegetation but by 2007 this had increased to 18%; within the river corridor the increase was from 9% to 32%.

The maps of the areas in each category for each date indicate that the areas of woody vegetation increase are immediately adjacent to the river (Main Map), mainly on the new areas created by channel movement (point bars and young floodplain). However, some of these areas were not covered with woody vegetation in 1984, and thus the increase was more than simple colonisation and succession on new bar areas associated with river movement.

Some of the increase in woody cover could be an increase in the size of the canopy due to differing states of the trees and an earlier spring time at the time of photography. Comparison of the air photos from each date indicates that, although all images were taken from mid-April to early May, trees were barely in leaf in 1984, leaves were well out in 1996, less so in 2001 and well advanced in 2007. The proportion of the change that can be attributed to such variations can be assessed by comparing the areas of old woodland on the valley slopes at different dates. Comparison in a sample area indicated that these areas increased by 7% in 1984–1996, decreased by 4% in 1996–2001 and increased by 16% of that area from 2001 to 2007. The overall increase from 1984 to 2007 was 19%. Thus, of the overall changes recorded, this proportion could be due to simple canopy growth and variation in seasonal conditions and not to new vegetation. This percentage net increase is much smaller than the increases measured in the riparian vegetation.

Overall, major increases in extent of woody cover have occurred, with transition from juvenile to mature stages over the period (Main Map figures). River floods have not been of sufficient magnitude to remove floodplain trees other than by channel migration, nor has human clearance taken place. Significant channel movement occurs almost every year (Hooke & Yorke, 2010) but the increase in woody area is still much greater than the creation of new areas of floodplain. The transition from juvenile to mature vegetation during the whole period and of changes in individual patches indicate increased recruitment and growth of young trees in the period prior to 1984 through to 1996. It is worth noting that the riparian zone was also very open in air photographs of 1947.

The data derived from the River Dane analysis can be compared with those produced nationally by the Centre for Ecology and Hydrology (CEH) Countryside Survey (2007) from stratified-random sampling of 1 km grid squares. Both data sets have been calculated as a proportion of total area in river valleys/riparian corridors that are woody at the dates of survey. The results of the local survey are very similar to those of the national survey, both in proportions and timing. However, the Environment Agency found little difference in extent or distribution of channel shading from riverside trees between the 1995–1996 and 2007–2008 surveys, either nationally or in the NW region (Environment Agency, 2010) from their River Habitat Surveys of 500 m sample reaches of channel (Seager et al., 2012).

Various possible causes have been hypothesised for the changes in the Dane valley, but the most likely is the exclusion of cattle by increased use of fencing adjacent to the river channels. Comparison with a heavily grazed and unfenced portion of floodplain on the neighbouring River Bollin, with very similar conditions, indicates much less growth (see air photo in Hooke, 2004). The changes, if widespread, have implications for occurrence of habitats and also for river channel processes, and therefore, for the European Water Framework Directive (WFD). They have implications for the assessment of effectiveness of the creation of buffer zones, promoted by the Environmental Stewardship Scheme, cross-compliance measures and Catchment Sensitive Farming. Climate change may also be having an effect.

5. Conclusions

A large increase in the area of mature woody vegetation, mainly of Salix spp, in this sample of river corridor in a lowland, agricultural area has been found for the period 1984–2007, an increase much greater than would be expected from simple succession on river bars. The air photographs of four dates provide snapshots of extent but tracking of individual vegetation patches indicates an increase in extent and maturity. Whilst increases in woody riparian vegetation have been found elsewhere, many of those cases have been related to changes in river regime and flood frequency, associated, for example, with dam construction. On this study river, no trend is present in data for annual peak flows or other discharge parameters for the period of record from 1949 to 2007, nor in the study period, 1984–2007, nor is there any trend of net narrowing of the river, increasing floodplain area (Hooke & Yorke, 2010). It is hypothesised that the woody vegetation increase here is mainly due to fencing control of grazing near the river banks, preventing browsing by cattle on young willow saplings. Further research is needed to test this hypothesis and to assess how widespread is this increase, which could have major process and ecological impacts, on rivers in Britain and elsewhere. The methods applied here distinguish effectively the types and maturity of riparian woody vegetation.

Software

The Main Map was produced using Esri ArcGIS 10.

Acknowledgement

The authors thank Rob Marrs (University of Liverpool) and Simon Smart (CEH, Lancaster) for advice.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This project was undertaken with a small grant from the University of Liverpool Living with Environmental Change Theme (LWEC) scheme.