Remarkably, all the ditches we sampled exhibited at least some characteristics of wetlands, especially in their herbaceous plant communities, almost all hydrophytic. All the ditch sites stored at least some organic matter in their soil, and all were somewhat wet. It appears that wherever topography and soils permit water to accumulate well enough to prompt people to dig ditches, wetland structure tends to follow, even after and despite drainage. Neither lower soil organic matter than forested ditches, nor substantial development and agricultural influences, prevented roadside and agricultural ditches from growing robust, dense, mostly hydrophytic, perennial, and native herbaceous plant communities. Plants were not just the same few weedy species in all ditches. Community composition varied with ditch type, and included a variety of native wetland species along with some invasive weeds. Similarly, the variation we observed among soils suggests that there is no one drainage ditch soil in the North Carolina Coastal Plain, or even soils specific to agricultural, forested, and freeway ditch types. For most variables we measured, ditches had wide ranges of values even within site types, suggesting richness in both ecosystems already present, and room to change, among Coastal Plain ditches.
We followed the USACE example in defining wetlands based on plants, soil, and hydrology (USACE 2010). The general agreement between USACE hydrologic indicators, soil water content, soil organic matter (Fig. 3), and known wetland and water landscape surroundings for (Fig. 6 and SI 3) for characterizing individual sites and site types as wetter and drier suggests that our methods captured real trends of wetland structure within these ditches. Overall, the ditches, of all types, tended wet. The organic matter contents suggest that some of these ditches, especially in forests, have substantially wetland-like soils. To compare to a natural swamp in the region, none of the ditches had as much soil carbon as the 95% in pocosins and 75% in gum swamps measured at Croatan (Bridgham & Richardson 1993), but a few came close. It is notable that ditches, forested and even agricultural, can have high carbon in their soils at all, as drainage ditches are not typically recognized as potential carbon sinks.
Our forested sites met our goal of representing relatively low human landscape influence for ditches as compared to agricultural and freeway sites, and indeed featured prime local conditions for wetlands as well. The matrix around many forested sites proved not just low in agriculture and development, but also more swamp, i.e., classified as wetland and also forested, than simply classified as forest, according to the NLCD. Lesser local disturbance clearly allowed far greater growth of trees, a critical vegetative layer to the type of wetland that once dominated this landscape, than in evidently trimmed agricultural and freeway ditches. Unmown herbaceous layers grew taller, if less dense, in forests as well, with more OBL plants. Forested ditches also proved more wetland-like in terms of soil carbon and water relative to agricultural and freeway sites. These differences make sense given land use history and preferences, with the forests lingering largely as easterly swampy remnants too wet for agriculture, and freeways constructed mostly inland within the Coastal Plain.
It is unclear from this short-term study what proportion of the accumulation of water and carbon in these ditches occurred since ditch construction, due to ongoing processes, and what persists as an artifact, perhaps degraded by ditching, of the broader wetland landscape. Some of the forested ditches may date to past clearing for agriculture, thereby constituting more of a scar of past land use than contemporary silviculture infrastructure. Longer term study might better sort through which human actions, past and present, influence the ditches to function how within the swampy forests of coastal North Carolina. Such study could inform possibilities not just in current forestry work there, but also in restorations and responses to saltwater intrusion. For now, though, we can say that these forested ditches do exhibit multiple strong signals of wetland structure. We did not have to visit these most natural of ditch contexts in the North Carolina Coastal Plain to find appreciable wetland structure, though; roadside and agricultural ditches also exhibited wetland characteristics, and distinct structure of their own.
Forested, agricultural, and freeway ditch types each had their own herbaceous vegetation communities. The differentiation of plant communities primarily by ditch type, which is to say, by human land use, construction, and management, suggests that humans exert substantial control over the plant communities of ditches already. Engineers and land managers could choose to exert this control more deliberately, beginning by experimenting with management to determine achievable goals. The high variability across sites in many of our measurements suggests correspondingly wide management potential. That extremes of restriction or permission of growth within site types tended to drive sites away from their site type clusters within the NMDS, towards other site type clusters, suggests a promising starting point for experimentation. It appears that mowing, applying herbicide, letting trees grow, or choosing to avoid or space out these management actions shifts ditch taxonomic composition. That percent cover increases most strongly along with cultivation within 500m of the ditch, of all landscape variables, suggests a possible landscape influence of fertilizer application as well. Likely these ditch wetland plants capture some excess fertilizer and store it as biomass, especially in agricultural settings, providing an underappreciated regulating ecosystem service. We see opportunities to not just appreciate what drainage ditches already do ecologically, but to potentially shape that ecology through management, in all types of ditches studied.
The significance of the vector of mean maximum herbaceous vegetation height in the NMDS (Fig. 6) suggests that mowing already substantially conditions roadside herbaceous plant communities, along with some agricultural ones. The correlations between vectors for herbaceous vegetation height and for OBL taxa proportion and overall taxa count (Fig. 6c, f, and i) suggest that letting plants grow more could lead to establishment of greater richness and more wetland taxa, though demonstrating causation would require experimentation. The North Carolina Department of Transportation (NCDOT) manages freeway roadsides. It already includes both environmental scientists, and regional divisions for localized management. Perhaps this agency could incorporate experimental variation in ditch management, designed with wetland-style considerations in mind, into duties they already execute, such as mowing and trimming, seeding and planting, chemical application, and dredging and re-grading. NCDOT could vary timing, frequency, and intensity of these practices to try to foster wetland or other conservation characteristics, such as soil carbon accumulation, or more or less growth of species of interest. Then again, freeway ditches’ status as transportation infrastructure, and the high proportion of development surrounding them, mandate caution for human welfare. Experimental approaches would have to allow water to flow without creating a flooding risk or adding a maintenance burden.
While balancing all these considerations is difficult, that freeway roadside ditches were relatively dry compared to other drainage ditches suggests the possibility of leeway to innovate with water management in freeway ditches. Freeways routinely come with drainage features all along their length, rather than focused construction of drainage in existing wetter patches, as is more common in agriculture and silviculture. So, it makes sense that freeway ditches generally would tend drier on average, even before factoring in that these freeway ditches tended further inland than others in the study. Systems like highway ditches built for one or few ecosystem services, just flood control and water transport, are often overbuilt for those purposes, leaving potential for additional uses (Palta et al. 2017). The range of herbaceous community variability, across the NMS (Fig. 6), in highway sites, also suggests broad potential outcomes. Experimenting with managing for additional ecosystem services may not inevitably increase maintenance workloads. Designed experiments incorporating natural elements into infrastructure have even reduced maintenance burden before (Felson & Pickett 2005). For future studies, private ditches along smaller roads, which are more numerous that freeways, may be just as interesting (after McPhillips et al. 2016, Schilling et al. 2018, and Tatariw et al. 2021). Roads, for better or worse, cross all types of terrestrial and freshwater ecosystems. The North Carolina Coastal Plain has upwards of 2,711 km of freeway and 93,502 km of roads (USGS 2014b). So, by quantity alone, roadside ditches present great potential for widespread influence.
Agricultural sites had the most vigorous herbaceous plant assemblages, including the densest herbaceous cover (Figs. 5 and 6). Despite the stress of surrounding development and agriculture (SI 3 and Fig. 6b, e, and h), these sites often had high taxa richness (Fig. 5a), albeit featuring maximum upland taxa along with maximum obligate wetland taxa across their assemblage range (Figs. 4 and 6a, especially c, and e). These agricultural ditches’ palette of plant taxa was broad, albeit nonnative as well as native (Fig. 5, SI 2), without deliberate intervention to make it so. All these characteristics suggest that agricultural ditches too have abundant potential, likely opportune for shaping by management. The very sites we sampled were agricultural research stations, already a hub for experimentation. The existing frequent need of maintenance like mowing and dredging in agricultural ditches (Dollinger et al. 2015), along with the need to mitigate high nutrient exports from agriculture in many cases, could make them great places to experiment with goal-directed management (De Meester et al. 2005, Koetsier & McCauley 2015).
Any efforts towards managing ditches as an aquatic resource, or first, to finish accounting for what they already do, would benefit from thorough mapping of ditches. However, few surveyed ditches show up in the NHD + v2, NHD + HR, or the NWI, which means that traditional mapping means fail to account for these ditches as potential aquatic resources. Ditches may be narrow, but their net length is long, particularly here in the North Carolina Coastal Plain. So, this under-accounting likely adds up to a substantial area and connective web of wetland structure across the landscape. Perhaps worse, the newer, eponymously higher resolution version of the NHD + included fewer of this study’s ditch sites than the older NHD + v2 did, a step backwards. The lack of inclusion is also worse for sites not surrounded by acknowledged wetlands, not heavily covered by tree canopy, and thus visible on simple aerial imagery. So, exclusion of ditches from geospatial datasets appears to be a growing choice. Admittedly, we selected possible sampling sites partially through permission to access, rather than strictly randomly. However, we did not consult the NWI or NHD about these sites before site selection. So, the datasets’ lack of inclusion of sites in this study likely reflect a broader deficiency in ditch mapping. Even a casual glance at either database superimposed on aerial imagery for this region makes clear that many ditches, and other small water features, are left out, or miscategorized. This lack of information and misinformation on extent and distribution of ditches apparently leaves a good deal of what we have shown to be wetland-like habitat uncounted, suggesting that artificial ecosystems are neglected in natural resource inventories.
However, the technology and even data to improve ditch mapping already exists. Where simple aerial imagery will not suffice, our LiDAR processing suggests the feasibility of finding ditches to add to spatial datasets, given enough computing power to deal with the large amount of data that using LiDAR at scale entails. Remote sensing scientists have already made significant progress on this task elsewhere (Bailly et al. 2008, Cazorzi et al. 2013, Rapinel et al. 2015, Mahor et al. 2016, Hasselquist et al. 2018, Lidberg et al. 2023). The decline in coverage of our ditch sites between the NHD + v2 and NHD + HR suggests that unfortunately, mapping of ditches may not automatically improve over time with general improvements in data, technology, and water mapping. Choices must change first. Perhaps papers like this one, documenting the existence of environmental features of interest in ditches, will inspire inclusion of these and other ditches in future mapping improvements.