Nomogram for maintaining forests to avoid watershed function problems
Meine van Noordwijk, ICRAF SE ASia
Current forest classification schemes, such as exist in Indonesia for example, try to identify areas with high risk for damage to watershed functions, largely on the basis of slope. While there is clear logic to taking slope into account in such a scheme, other factors such as rainfall regime and soil properties can have a substantial modifying effect on what should be considered as critical value for slope. No simple schemes for identifying such modifying effects exist that have found practical application. Yet refinement may be relevant as pressure on forest lands in SE Asia require a stronger justification for the forest set-asides for watershed protection.
Hornberger et al. (1998) provided a set of basic equations that allow for an interpretation of a landscape into four domains: a domain where no surface runoff or soil movement is to be expected, a domain where saturation overland flow can be expected without erosion, a domain where erosion is likely and a domain where landslide risk is substantial. The domains are defined by two parameters: slope and a topographic index that indicates the area up-hill per unit contour length (this is equal to slope length for situations with parallel streamlines towards the river).
Figure 1. Domains where saturation overland flow, erosion and landslides can be expected on the basis of slope and a topographic index’ that provides the area uphill per unit contour length (slightly modified from Hornberger et al.) (default parameters: T = 10-4 m2 s-1, qtot = 50 mm day-1 or 5.79 10-7 m s-1, alpha = 8 10-6 m2 s-1, friction angle = 35o, Rhos and Rhow = 1 Mg m-3)
Compared to the current scheme with a single critical value for slope, this nomogram offers two types of progress: 1) it explicitly relates the domain to potentially measurable characteristics of vegetation and soil, and 2) it introduces the ‘topographic index’ as a parameter that can substantially modify our interpretation of the types of problems that can be expected. For low values of the topographic index, as will obtained close to the division lines between subcatchments even steep slopes may be relatively safe, while for higher values of the index lower down the slope the same slope angles may carry greater risk.
It may be relevant to look at the equations and see where the leverage points for land cover change will be. The first equation provides the threshold for saturation overland flow:
A/c > T tan(Slope) / qtot (1)
where T is the transmissivity of the surface soil layer [m2 s-1] and qtot is the total runoff (rainfall – canopy interception – recharge to of the soil profile to field capacity to compensate for preceding evapotranspiration) for the maximum rainfall event considered.
The second equation provides the threshold for erosion:
A/c > T tan(Slope) / qtot + alpha/ (qtot tan(Slope)2) (2)
where alpha is the resistance of the soil to erosion [m2 s-1].
The risk for landslides defines a critical slope for conditions where the soil is saturated with water:
Tan(slope) > 0.5 tan(friction angle) (3)
where the friction angle depends on the cohesiveness of the soil profile, and a threshold for non-saturated conditions:
A/c > T (Rhos/ Rhow) tan(Slope) (1 - tan(Slope)) / (qtot tan(friction angle)) (4)
where Rhos and Rhow are the bulk densities of soil and water, respectively.
Effects of land cover change (‘deforestation’) can affect these domains as follows: qtot which depends primarily on rainfall can be increased if there is no intercepting canopy (maximum effect of the order or 5 mm day-1) and if there has been less preceding evapotranspiration creating storage capacity in the soil (effect depends on soil properties and vegetation change, probably less than 20 mm day-1; for non-vegetated land covers qtot may approach the gross rainfall; so combined effects can be a 50% increase in qtot for the maximum rainfall event considered. This effect can be immediate after forest conversion.
T and alpha may be reduced over time as the soil looses its structure; a tentative reduction by 40% may be expected for vegetated land covers, but for vegetation-less land use types the effects might be stronger
The combined effect of such changes is a shift in the boundaries of the domains, with more surface runoff and erosion (Fig. 2). A loss of deep-rooted trees may decrease the friction angle, and thus shift the landslide domain to the right.
Figure 2. Modifications in the boundaries of the domains of Figure 3 as a consequence of forest conversion (assuming qtot to increase by 50%, T and alpha to reduce by 0.6
Practical implementation of this approach
If this approach seems attractive, we have to consider practical implementation. The challenges are:
A. Derive the topographic index from the information in a digital elevation model (DEM); as current software allows (A1) the automatic derivation of a DEM from a set of aerial photographs and the slope at any point in a landscape (A2) can be derived from such a DEM, the main challenge is to set up a procedure for (A3) distinction of subwatersheds and streams, and (A4) derive the area uphill per unit contour length for the area between streams and sub-watershed boundaries, possibly using a classification of the area by distance to the streams
A challenge here is to establish the spatial resolution for a DEM that is needed for meaningful results.
B. Establish reasonable value for qtot on the basis of rainfall with modifiers for land cover types (interception, water use creating storage capacity). Existing plot-level water balance estimates and models can be used to derive the land-cover modifiers. Analysis of long-term rainfall records for excedance probabilities is straightforward.
C. Establish reasonable estimates for the Transmissivity value T and the way this may be modified by land cover. The T value is probably scale dependent, as it reflects the ‘bottlenecks’ rather than average soil properties. Procedures for deriving T from more detailed process-level water transport models (including HYDUS 2D and their parametrization via pedotransfer functions) are feasible.
D. Derive estimate of the resistance to erosion , alpha, on the basis of measurements of runoff and sediment transport in mini-plots. Analysis of existing data sets can probably provide sufficient basis for a ‘pedotransfer’ approach, with a protocol for field verification to support future practical use.
E. Deriving estimates for the friction angle and the way this depends on soil type and rooting pattern of the vegetation. This appears to be the most challenging parameter as yet, and further contacts need to be developed in this domain.
References
Hornberger, G.M., J.P. Raffensperger, P.L. Wiberg, K.N. Eshleman. 1998. Elements of Physical Hydrology. John Hopkins University Press, Baltimore (Maryland, USA)
2