Influence of intra-event rainfall variation on surface–subsurface flow generation and soil loss under different surface covers by long-term field observations
Abstract. Rainfall is the main driver of runoff generation and soil erosion. The impacts of natural rainfall on water erosion have been extensively studied at an inter-event scale; however, very few studies have explored the intra-event influences and associated responses to different surface cover types. In this study, long-term in situ field observations of surface runoff, subsurface flow, and soil loss characteristics in three surface cover plots (bare land, litter and grass cover) under natural rainfall events were conducted from 2002 to 2012 in the red soil hilly region of southern China. According to the period of most concentrated rainfall, 262 rainfall events were classified into four types of intra-event variation: advanced, intermediate, delayed, and uniform patterns. For bare land, the advanced pattern with the shortest duration and the highest intensity was main rainfall type for surface runoff and soil loss; the contribution rates were 57.24 % and 75.17 % for surface runoff and soil loss, respectively. Sediment yields were more sensitive to intra-event rainfall variation than surface runoff. The highest subsurface flow was found in the delayed pattern with the longest duration and high depth, followed by the uniform, intermediate, and advanced patterns. For all rainfall patterns, compared to the bare land, surface cover significantly reduced surface runoff and soil erosion by 88.01 to 91.69 % and by 97.80 to 97.95 %, respectively, while subsurface flow was increased from 3.55 to 5.92 times. The reduction benefits of litter cover were comparable to those of grass cover. However, the increasing benefit of subsurface flow for litter cover for each rainfall pattern ranged from 1.38 to 2.67 times those of grass cover. Moreover, surface cover weakened the influences of intra-event rainfall variation on surface-subsurface flow and soil loss. The results demonstrated that intra-event rainfall variation had important effects on surface-subsurface flow and soil loss, and provided a basis for optimizing surface cover measures to effectively respond to extreme water erosion and drought caused by global climate change.
Jian Duan et al.
Status: open (until 14 Jul 2023)
Jian Duan et al.
Jian Duan et al.
Viewed (geographical distribution)
This paper analyses runoff and soil loss data collected from several experimental plots, with data covering about a decade of observations under natural rainfall.
I would first commend the authors for having adopted field study under natural rainfall, rather than the generally inappropriate reliance on "rainfall" simulation (which typically uses unrealistic, fixed, high intensities - often with little or no attempt at offering a reasoned basis for this, and typically neglecting intensity profile altogether). I think that observations made under natural rainfall are arguably far more informative and valuable than "rainfall" simulation studies that neglect the kinds of issues that are discussed in the current manuscript.
The use of four classes of intra-event rainfall pattern (advanced, intermediate, delayed, and uniform) is not unusual, and does provide some basis for identifying and categorising the intra-event rainfall intensity variation. However, this classification provides no indication of what the actual intensities during an event were. Indeed, it is a limitation of the current manuscript that remarkably little is actually said about the rainfall intensities, beyond reporting of the mean event intensity (I presume) for the advanced, delayed, etc. (e.g. lines 234-236). I think that it would have been helpful and appropriate for the paper to report additional detail concerning intensity - perhaps using measures like I5, I15, I30, I60, and so on. In the absence of this, we cannot really understand whether, for instance, the 'advanced' event type had different peak intensities than the other classes, and for how long those peak intensities might have lasted. We cannot see whether there was more intensity variation in advanced than in delayed, for instance, and it is well-known that short-term intensity peaks can exert critical influences on soil erosion (e.g. see for instance Dunkerley, D. L. (2019). Rainfall intensity bursts and the erosion of soils: an analysis highlighting the need for high temporal resolution rainfall data for research under current and future climates. Earth Surf. Dynam., 7(2), 345-360. doi:10.5194/esurf-7-345-2019).
I would like to have seen in the paper more information on rainfall continuity or intermittency also, as breaks in rainfall can be critical in allowing overland flow to slow and for soil to be re-deposited, for ephemeral surface ponding to dissipate, for soil infiltrability to partially recover, and so on. Much is known about all of this, and could have been considered in the manuscript. Many rainfall events reported in the manuscript have durations of more than 24 hours, and up to 48 hours. Was rainfall actually continuous through these long durations, or were there breaks (cessations) in rainfall? The authors might also say something about when the rain occurred, especially for the 'advanced' type. These might for instance have been late afternoon convective events, whilst the 'delayed' type might have been long, overnight falls. The timing, diurnal or nocturnal, would influence evaporation rates, especially during any temporary breaks in rainfall. This is too often ignored when rainfall data are reported simply in terms of event amounts and intensities.
I was not entirely convinced that simply classifying rainfall events as 'advanced', 'delayed', etc. is sufficient. I think that the authors recognised this too. Clearly, the depth and intensity of the events differed widely, and this is where measures such as I30 or a related index might have been useful. In my own experimental studies, designed primarily to understand how intensity profile affects infiltration and runoff, all events, regardless of intensity profile, had the same depth, duration, and average intensity. An example is Dunkerley, D. (2012). Effects of rainfall intensity fluctuations on infiltration and runoff: rainfall simulation on dryland soils, Fowlers Gap, Australia. Hydrological Processes, 26(15), 2211-2224. doi:10.1002/hyp.8317. In that work, all events lasted 90 minutes and delivered 15 mm of rainfall at an average rainfall rate of 10 mm h-1. This ability to hold depth and duration constant experimentally isolates, at least to some extent, the intensity profile itself as the factor than could account for differences in plot infiltration and runoff. But this is not possible to do when using natural rainfall. Therefore, event durations were different among the intensity profiles analysed by the authors of the present manuscript. Were the 'advanced' events primarily different in their effects on the soil plots from the 'delayed' events in terms of their intensity profile, or their duration, or their peak intensity, or the duration of intensity exceeding, say, 10 mm h-1, or some other factor or factors? I think that the present manuscript leaves this unresolved. Perhaps issues of this kind could be addressed in future work by the authors. I well understand that not all questions can be addressed in a single manuscript of manageable length. It is clear from Figure 3 for instance that the 'advanced' events are much more variable in their intensity profiles than any of the other classes. The spread of event characteristics declines in the sequence advanced > delayed > intermediate > uniform. This suggests, for instance, that it might be advantageous to consider some sub-categories within at least the 'advanced' class (and perhaps also in the 'delayed'). A suitable measure might be something like 'time to peak intensity', say. It is evident from Figure 3 that some 'advanced' events are more advanced than others (i.e., that their rainfall arrives much earlier). This might be worth considering in future work.
The authors also neglect, as is often the case in such studies, the time sequence of rainfall events. This sequence influences the critical antecedent wetness (or dryness) of the plot soils prior to the start of an event. Events that begin soon after a prior event, with soils that are already partially wetted-up, are likely to have infiltration and runoff characteristics that are different to those seen when an event begins after weeks without rain, say. The literature is full of findings on such effects, and they really should not be ignored. In my own experimental work, referred to above, each experiment was made on a fresh runoff plot, and none had received any rain for some weeks, such that all plots could be regarded as having the same initial condition - 'dry'. So I would speculate that an 'advanced' event soon after a prior event would show very different runoff behavior than one on dry soil, say. And it makes limited sense to compare what happens in an 'advanced' event on wet soil with what happens on a 'delayed' event on dry soil - because it is not only the class of event that is different, but the antecedent soil wetness. The two effects are then confounded in any statistical analysis. Clearly this cannot be controlled for in work under natural rainfall, but nevertheless, antecedent conditions and their influence have to be borne in mind, and their possible effect analysed as well as can be managed. One could, perhaps, analyse as a test case only rainfall events that occurred after at least 5 days without rain, or some such criterion.
The manuscript is generally clear and easy to read. There are minor errors that could be corrected easily. The authors variously report that they analysed 226 rainfall events (line 225) or 262 events (lines 321, 504). I am not sure which number is correct. I doubt that the plot walls were actually 100 cm (1 m) tall - as stated in line 148. I cannot see how one can obtain 1 min rainfall data (line 105) from a standard tipping-bucket rain gauge. During many clock minutes, a bucket would simply be filling progressively, and might finally tip during a minute that actually delivered very little rain. This just requires some care in data processing. I generally don't see much validity in expressing tipping-bucket gauge data at any finer accumulation duration than 5 - 15 minutes, depending on the rainfall intensity.
In line 65, perhaps find a better expression than 'inter-fluctuations'. In line 166, avoid using '5-cm'. The international SI metric system requires that only a space can come between a quantity (in this case, 5) and the units of measurement (in this case, cm). The correct form is therefore "5 cm". In lines 171-172, I did not understand what the 5 measurements were for. Was this to measure the depth of water in a collecting vessel? In line 218 I think that the word 'and' is superfluous in the expression "the and runoff plot".
Among the 'key words', I would suggest adding something like 'intensity profile' or 'storm type' or 'storm pattern' (the terminology is currently not settled).
As the foregoing will have suggested, there is a considerable body of literature that could have been cited (e.g. on the time-sequence of rain events and the effect this has on antecedent wetness and hence on plot behavior). The authors might find the following paper to be of interest, for example, as it addresses the significance of intensity profile, though I do not suggest that this needs to be cited, as I am among the authors.
Liang, Y., Gao, G., Liu, J., Dunkerley, D., & Fu, B. (2023). Runoff and soil loss responses of restoration vegetation under natural rainfall patterns in the Loess Plateau of China: The role of rainfall intensity fluctuation. CATENA, 225, 107013. doi:https://doi.org/10.1016/j.catena.2023.107013