Process-based approaches for predicting soil erosion at the field and small watershed scales.

Dino Torri, 2016, Process-based approaches for predicting soil erosion at the field and small watershed scales., THIRD CONFERENCE OF COST ACTION ES1306_ CONNECTING EUROPEAN CONNECTIVITY RESEARCH HYDROLOGICAL AND EROSION PROCESSES IN MEDITERRANEAN LANDSCAPES_ IMPACTS OF LAND MANAGEMENT ON CONNECTIVITY., Universita' Palermo, 28th February - 5th March, 2016,

Process-based approaches for predicting soil erosion at the field and small watershed scales Dino Torri IRPI, CNR, Perugia, Italy, Processes occur at their own scale and often are the results of other processes occurring at a more basic scale. This is a well-known mechanism that is at the base of the whole scientific system_ looking for more and more basic processes aiming at the basic processes/constituents of Universe and Life. Soil erosion, runoff generation, and slope stability research share this same aspect, even if these studies are often driven by practical implications and immediate utilization. Hence, discussing something that is process-based requires that we define what are the processes we want to discuss and at which scale of approximation and generalization, their representation in the synthetic landscape represented by our approximation of topography, land use and management, structures and infrastructures, landscape and land use spatially distributed history. Following the scale of representation of the processes, we will need to explicitly describe the math approximating the chosen set of physics, chemistry and biology laws which are relevant to the studied processes. Given our present knowledge of the processes involved in soil erosion at the field-to-small-catchment scale (FSC), we can expect that we will work at a scale that is usually coarser than the one proper to the processes. Furthermore, we will be limited by the input data (typology and quality) available with respect to the data requested by our equations. Hence our models, despite how well founded on a good physics, are always at risk of producing coarse approximations. FSC scale implies a quite large number of processes (Torri and Borselli, 2011). Since the first drop hits the soil, water begins to penetrate into it decreasing soil resistance to compaction and detachment. Capillary suction can cause telluric air to be trapped and compressed into the soil causing aggregates to partly disintegrate (slaking). With decreasing soil resistance, drop impact becomes effective in compressing soil and detaching soil particles and aggregates. These processes (slaking and splash) are responsible for the formation of a usually less porous layer, which causes a change in soil pore size distribution and infiltration rate. Raindrops also smoothen local roughness. These processes accelerate the formation of ponds and their expansion until partial runoff starts. The presence of water at the very surface of the soil dissipates part of the raindrop impact energy reducing raindrop detachment. The dissipated drop energy increases the capacity of the usually thin layer of diffuse overland flow to transport detached soil grains. A series of proto-rills starts to appear, to be quickly erased by episodes of intense and very localized deposition. If a sufficiently large area gets hydraulically connected to some of the proto-rills then a rill channel is usually formed. Similarly, excess-runoff gullies will be excavated, e.g. when rills coalesce. The area draining to a potential gully head ought to be 40-50 times larger than in the rill case to initiate gully erosion. When there is a subsurface layer more erodible than the capping layer then a hypodermic flow develops and, with it, pipe erosion. The occurrence of micropiping at rill and gully headcuts helps increasing linear erosion intensity and the head-cut retreat rate. Along the rill and gully walls various type of mass movements occur, from soil toppling to relatively large mass movements favored by the undercutting and the formation of tension cracks, some of these processes occur during rainfall, others later. Debris flows occur when water rapidly accumulates in the ground, during heavy rainfall or rapid snowmelt. Mass movements along slopes often occur after the causative rainfall. They can trigger the formation of other rills and gullies in subsequent events. Other erosion processes, which occur mainly or partly outside rainfall events, include soil creeping, due to soil shrinking-swelling cycles, and solifluction, due to freeze-thaw activity. All these processes are characterized by very different erosion rates, with process-caused soil loss and process occurrence being almost inversely proportional. When the most devastating soil erosion processes occur, they can damage infrastructures and buildings and strongly affect the agricultural activities in cropland. In the latter case, extra tillage operations and land leveling are requested, which accelerate and spread around the degradation of soil characteristics while the local morphology is modified. On the basis of the recalled processes, we can list the items concurring to produce a process-based approach to soil erosion modeling at the field/slope and small watershed scale_ 1- the processes, which dominate soil erosion at a given site, are actually and properly described in the software; 2- the input data, required by the processes and their descriptions, are adequate and measurable or they can be estimated through proper and reproducible procedures; 3- stochasticity and spatial variability are taken into account; 4- the landscape is described at the correct scale for the involved processes; and, if the model evaluates erosion over long periods, then 5- it must be able to work under very different climatic conditions (i.e. the model must deal with climate direct and indirect effects); 6- it must interact with the various ecosystem components. Finally, if we want the model to include all the erosion processes, then the model cannot be an event-model that does not simulate the processes occurring between rainfall events. On the contrary_ 7- the model must be a continuous model; 8- tillage and land leveling shall be included because of their effect on local topography and on soil characteristics. Item 1 seems obvious_ a model cannot be process-based if processes are missing. Actually, the way in which a process is represented can make the difference. Slaking of soil aggregate is a composite process, which can be represented in different ways_ a) through a quantity proportional to the soil saturation degree - in this case, we need four input data_ water content at saturation, initial moisture, volumetric fraction of disconnected pores and an empirical coefficient depending on soil, soil use, period of the year (or prevalent weather condition of the previous weeks), vegetation status - hence something very local and of short temporal duration; b) describing water penetration into aggregates by capillary action and get a dynamic reconstruction of the slaking, even introducing a slaking rate - temporal dynamics of the water retention curves substitute the soil empirical coefficients and their dynamics as in a); c) adding a constant to interrill detachment and call it "slaking component", maybe measuring it through soil aggregate stability. The three approaches are not equivalent; the second one can be used for understanding and depicting a wide range of remediation. The first one is probably the best for keeping the model simple and improving chances to forecast reliable erosion rates. The third is too coarse to be useful Another important aspect, still belonging to Item 1, is relative to the mathematics of the software. It often involves approximate numerical calculus, which must converge_ convergence is a responsibility of the modelers. Unfortunately, most model descriptions leave convergence into a foggy background. Similar discussions will be developed for the other items. E.g., soil resistance to erosion can be represented by undrained soil strength at saturation (tensile and/or shear; Item 2). The presence of living plants increases it because the direct effect of roots and the indirect effects of root exudates and the soil biota that plants favor (Item 6): certainly worms (Charles Darwin in his book "The formation of vegetable mould though the action of worms", 1881) and fungi (Rillig et al., 2015). Several plant-related parameters can help evaluating this increased resistance. Vegetation-growing models can be used for evaluating these effects and for changing them on the base of the history of land use and management changes (residence time of soil organic matter bindings). Items 5 through 8 are of the outstanding importance because of climate changes and ecosystem modifications. The local weather is not a constant. On the contrary, it is changing dramatically and ecosystems are following the changes, partly resisting, partly adapting. In the Mediterranean area, we are facing warmer winters and hotter summers, with more rainstorms and less rainy days, and more frequent and longer droughts (Zollo et al. 2015). Shortly, we must expect a different weather situation in 10 years from now, which will be different from the situation in 20 years and so on. Hence any prediction of soil erosion, that want to be serious and scientifically informed, must use predicted weather series under a changing climate while vegetation and soil properties must be evolved accordingly. Under the present situation, where the climate change is one of the major threats to ecosystems (Hansen et al., 2015), the ecosystem resilience is basic to maintain its functions. This makes the soil component essential because of its many fundamental services, from storing water, which can attenuate drought effects, to hosting an important part of the biota. Hence, managing soil resistance to erosion and degradation becomes paramount to contrast the expected increase of climate aggressiveness while soil erosion (process-based) models should be able to replace models based on empirical/statistical process-lumping relationships (based on observations of a past situation).

Data from