| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Slope Conservation Section, Geohazards Division, Disaster Prevention Research Institute, Kyoto University, Uji, Kyoto 611-0011 JAPAN
sidle{at}slope.dpri.kyoto-u.ac.jp
Received for publication September 19, 2006.
Dear Editor:
I was encouraged to see a contribution in JEQ on the very important issue of sustainable land use related to soil erosion in the Loess Plateau of China. As noted by Wang et al. (2006), this vast region of deep, wind-deposited silty soil is among the most erodible areas in the world. This erosion is of critical concern to sustainable production of the land base, water quality, air pollution, and downstream flooding. Herein I focus on some problems in capturing the dominant erosion processes and land use in the Loess Plateau using the APEX (Agricultural Policy–Environmental eXtender) model employed by Wang et al. (2006). My comments cut to the heart of a broader issue—the importance of incorporating process function in distributed catchment models, particularly related to effects of land management activities.
As background, Wang et al. (2006) state that water-induced erosion was simulated by MUST (Williams and Izaurralde, 2005), a modified version of the universal soil loss equation (USLE); slope length/steepness equations derived from the revised universal soil loss equation (RUSLE) were used in this application of MUST because better results were achieved compared with using the USLE slope length/steepness equations. Nevertheless all of these models (USLE, RUSLE, and MUST) focus on inter-rill and rill surface erosion and are not appropriate for assessing deep gully erosion and related processes which dominate in the Loess Plateau. Such concerns are especially true at the moderate-size catchment scale where complex hillslope–channel linkages occur. Furthermore, other important erosion processes are not considered in MUST, nor are the effects of certain management practices on these processes. Important omissions in background information for this study include: (1) the spatial distribution of rain gages; (2) the resolution of the digital elevation model (DEM) used in the simulations; and (3) the location, method, and accuracy of sedimentation measurements. The first two points greatly affect the accuracy of distributed model inputs and flow paths, while the third point is essential for judging the accuracy of annual sediment outputs. My comments that follow raise more fundamental issues in this article related to: (1) erosion processes; (2) sediment transport and storage; (3) effects of land management; and (4) catchment-scale modeling.
Erosion Processes in the Loess Plateau
Gully erosion is the dominant erosion process in the Loess Plateau (e.g., Leger, 1990; Shi and Shao, 2000; Wu and Cheng, 2005), yet the uniqueness of this process is not discussed in this paper. Gullies initiate when a specific upslope area supplies the critical flow shear stress to incise channels—steeper slopes require smaller contributing areas for incision (e.g., Dietrich et al., 1992; Poesen et al., 2003; Wu and Cheng, 2005). However, once gullies are incised, they may actively headcut in erodible soils (Shi and Shao, 2000; Moeyersons, 2003; Hessel and van Asch, 2003). Gullies exhibit fundamentally different behavior than rilled and inter-rilled slope segments. When runoff occurs, gullies behave as major channels. In loess soils, gullies incise, banks erode, and bank undercutting occurs during storm runoff. These processes lead to immediate or later bank collapse and mass wasting due to the oversteepening of the gully sidewalls and around headcuts (Piest et al., 1975; Hessel and van Asch, 2003; Poesen et al., 2003) (Fig. 1c). Following gully bank oversteepening, the surficial mass wasting process of dry ravel can occur (Sidle and Ochiai, 2006). As gullies erode more deeply, mass wasting processes dominate over surface erosion (Wieczorek et al., 1987; Xu, 1999; Sidle et al., 2004).
|
Landslides are also a significant erosion processes in the Loess Plateau that need to be considered in terms of land management practices (Billard et al., 1993; Cao and Coote, 1993; Derbyshire, 2001; Messing et al., 2003). Earthquakes and seasonal rainfall in this otherwise dry region trigger numerous landslides and landslide-debris flow combinations in loess hillslopes (Derbyshire, 2001). These processes cannot be estimated by the erosion model employed by Wang et al. (2006). While landslides are episodic, their contribution to long-term sediment budgets must be considered related to sustainable land management alternatives. Short-term tests (e.g., <10 yr) of erosion models using catchment outlet sedimentation rates cannot hope to capture episodic landslide sediment contributions.
Wind erosion is another significant process in the Loess Plateau, partly because of the poor vegetation cover in many areas and partly because of anthropogenic disturbances. Li et al. (2005) estimated that wind erosion comprised >18% of the total erosion from a loess area in Shaanxi Province. Xu et al. (2006) note the linkage between wind erosion and subsequent debris flows and hyperconcentrated flows in channels. Wind erosion needs to be considered in the Loess Plateau related to sustainable land use and sedimentation. Although the APEX model can estimate wind erosion (Williams and Izaurralde, 2005), and it is implicitly addressed in the introduction of Wang et al.'s (2006) article, it was apparently not estimated in the erosion simulations because wind speed data were not available.
Sediment Transport and Storage
A very important role of gullies is their potential to couple sediment delivery from hillslopes to channels (Kasai et al., 2005). Based on my experiences in the Loess Plateau, such connectivity is facilitated by high rates of erosion not only from active gullies, but also from road networks servicing terraced areas, slope failures along terrace risers, wind erosion, and episodic mass wasting. None of these processes can be effectively simulated by a USLE-based erosion model. The seemingly ‘good’ annual sediment predictions obtained by Wang et al. (2006) were likely the result of using the Bagnold-based sediment routing model, which can rather efficiently route the abundant sediment already stored in the channel system (i.e., an energy-limited fluvial system). The Bagnold equation is based on the concept of stream power, thus for energy-limited fluvial systems it can be used to route stored sediment downstream. However, such routing does not address the debris flows and hyperconcentrated flows that occur in channels of the Loess Plateau (Xu, 1999; Xu et al., 2006), nor does it allow for an adequate description of sediment process linkages between hillslopes and channels (Poesen et al., 2003; Sidle and Ochiai, 2006). Such episodic erosion events may overwhelm the system, dwarfing values that are measured during ‘quiescent’ periods with little mass wasting. Sediment storage may occur on landscapes or in channels in the intervals between large runoff events, priming the system for future episodic evacuation. Annual sedimentation rates measured at outlets of moderate-sized catchments do not provide good estimates of within-catchment erosion, particularly related to spatially distributed land uses.
Effects of Land Management Activities
Wang et al. (2006) focused their erosion simulations on manipulations of vegetation cover. Based on my experiences in the Loess Plateau, many other management practices affect soil erosion at the catchment scale. While the widespread terracing in the region is generally believed to an effective soil conservation measure, such practices have been shown to cause problems related to slope stability. The wide terraces recently constructed in steep hillslopes of the Loess Plateau require larger and more extensive service roads and higher terrace risers. Significant mass wasting as well as surface erosion occurs on terrace risers (Billard et al., 1993; Sidle et al., 2006). Roads in terraced terrain and paths are persistent sources of sediment (including gullies), as well as effective conduits for transporting eroded soil to gullies and channels (e.g., Ziegler et al., 2000; Sidle et al., 2006). Also, back-sloped terraces concentrate water during storms, thus sometimes causing landslides (Johnson et al., 1982; Billard et al., 1993; Sidle and Ochiai, 2006). In steep, unterraced terrain where forests were cleared and replaced with grasses or shallow-rooted crops, landslide potential increases substantially due to the loss of root strength (Sidle et al., 2006). These important, but episodic, sources of management-induced erosion are not considered in the APEX model.
Reductions in soil loss related to grass cover, although less than for trees, are likely overestimated by Wang et al. (2006) because shallow-rooted grasses provide negligible reinforcement against the various mass wasting processes in gullies or on the landscape (e.g., Rice, 1977; Sidle and Ochiai, 2006). Wang et al. (2006) recognize that the benefits of sediment basins are largely artificial—while they trap sediment downstream, they do not ameliorate on-site erosion. An important omission, however, is the fate of these basins once they fill or experience an episodic event such as a debris flow. While unlikely, the failure of such sediment dams and downstream consequences must be considered.
Lessons to be Learned in Applying Models in Managed Catchments
Whenever hydrogeomorphic processes are modeled at the catchment scale related to distributed effects of land use, several questions should be asked: (1) Does the model capture the important processes at the source? (2) Are the pathway dynamics of the materials being modeled adequately represented? (3) Are the land management practices sufficiently represented? and (4) How can we test the model given the study objectives? Given the dominant erosion processes in this region, it appears that the APEX model used by Wang et al. (2006) is not suitable to estimate erosion and sediment transport in this area. To imply that the model accurately depicted erosion processes and to make land management recommendations based on this application is misleading. Moreover, it is irresponsible to advocate applying this model as a management tool for the entire Loess Plateau region—this represents a classic example and potential dangerous outcome of modeling that gives the ‘right answer’ for the ‘wrong reasons’ (cf. Sidle, 2006).
Modeling the dominant erosion processes in the Loess Plateau related to past and future land management is a difficult task. While some models (e.g., WEPP and EGEM; Flanagan and Nearing, 1995; Woodward, 1999) attempt to capture the process of concentrated flow detachment of sediment in gullies by comparing flow shear stress to the critical shear stress of the gully bed material, they have not been extensively tested and only simulate the fluvial erosion processes in gullies (not mass wasting). Such process-based models could be combined with models that incorporate soil mechanics principles to estimate mass wasting along gully banks and at headcuts, along with slope-area thresholds for gully initiation. The effects of piping could also be included as a subprocess in estimates of gully and bank retreat (Poesen et al., 2003). Much more emphasis needs to be placed on understanding the effects of anthropogenic practices such as roads, paths, ditches, and terraces, on erosion and sediment delivery—not simply vegetation cover. These linear terrain features augment the contributing hydrologic area, thus promoting gully formation (e.g., Moeyersons, 2003; Poesen et al., 2003). To predict the effects of alternative land use on soil erosion, it is also necessary to model or estimate the processes of mass wasting on hillslopes as well as within channels and gullies. Effects of vegetation conversion and forest clearing on long-term landslide probability have been modeled (e.g., Dhakal and Sidle, 2003). Additionally, sediment accumulation in channels and periodic evacuation by debris flows or hyperconcentrated flows needs to be estimated by empirical models that are based on process function or magnitude-frequency relationships (e.g., Liu and Lei, 2003; Aleotti, 2004). Finally, the widespread process of wind erosion should be estimated; while this process can be assessed by APEX, such estimates may benefit from recent advances in regional-scale predictive systems (e.g., Sun et al., 2006; Webb et al., 2006).
Incorporating this complex suite of erosion processes in a regional predictive model is difficult. Such process-based, integrated models are currently not available and supporting field research is undoubtedly needed to develop such models. However great the temptation, it is not wise to apply an existing model that does not emulate the dominant erosion processes in the Loess Plateau if the objective is to demonstrate how the model can be used to evaluate alternative soil erosion management strategies—such was the case in the paper by Wang et al. (2006). Furthermore, erosion models calibrated against catchment ‘outputs’ largely ignore internal processes (or at least relegate these to parameter fitting) related to entrainment, transport, and storage, and thus are of limited use or even misleading in evaluating the effects of temporally and spatially distributed land uses (Sidle, 2006). Before developing detailed physically-based models of erosion processes in the Loess Plateau, dynamic sediment budgets should be considered as a plausible approach to evaluate sediment sources and sinks related to land use practices and natural processes.
Editor's note: The authors of the original article were given the chance to reply, but did not. Therefore, there is no separate reply from them.
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Crop Science | |||
| Journal of Natural Resources and Life Sciences Education |
Vadose Zone Journal | ||||
| Soil Science Society of America Journal | Journal of Plant Registrations | The Plant Genome | |||
