Practical thresholds to distinguish erosive and rill rainfall events
- Autori: Todisco F.; Vergni L.; Vinci A.; Pampalone V.
- Anno di pubblicazione: 2019
- Tipologia: Articolo in rivista
- Parole Chiave: Interrill; Rainfall erosivity; Rainfall hyetograph; Rainfall pattern; Rainfall thresholds; RUSLE; Soil erosion; Soil loss; USLE
- OA Link: http://hdl.handle.net/10447/392678
In this paper, 1017 rainfall events from 2008 to 2017 are used to identify the rainfall threshold that produces upland erosion at the Masse (central Italy) and Sparacia (southern Italy) experimental stations. The rainfall events are classified into three classes: non-erosive, interrill-only and rill. The threshold values for separating as correctly as possible the erosive rains (case I) and the rill rains (case II) are derived solely from the hyetograph. Each threshold value is obtained by imposing that the long-term erosivity of the events above the threshold is equal to the long-term erosivity of all erosive events (case I) or only rill events (case II). The performances of selective criteria based on 31 threshold variables are compared, and those most effective in separating erosive and rill events are identified. The identification of the best criterion depends on the aim of the analysis. It could be required to provide the greatest accuracy for separating erosive and rill events or the lowest error in the prediction of long-term erosivity. In general, the results clearly show that the best variables are those that quantify the characteristics of rainfall patterns, such as rain showers (periods of continuous rain) and the deviations in the rain records over a truncation level. These results are especially significant for the operational estimation of rainfall erosivity and for identifying the trigger of the erosion process and rill development by using only a hyetograph. This is obtained by relatively simple field measurements and is also widely available on a global scale. The most effective variables are potentially usable in water erosion prediction models as proxies of variables that are more rarely available and/or more difficult to measure.