Temporal stability and patterns of runoff and runon with different cover crops in an olive orchard (SW Andalusia, Spain)

- Observed runoff was ca. 9% higher under conventional tillage than under cover crops.
- Time to ponding with conventional tillage was much higher than that with cover crops.
- Model simulations were satisfactory (average Nash-Sutcliffe coefficient of 0.665).
- Observed and simulated runoff with cover crops had the highest spatial stability.
- Cover crops triggered higher temporal stability of runon than conventional tillage.

Conventional tillage (CT) and cover crops (CC) trigger different runoff (Q) and runon (Qin) magnitudes and patterns in woody crops. The spatial and temporal stability of these patterns is not well known yet. In this study, we run the uncalibrated DR2-2013© SAGA v1.1 model (0.5 × 0.5 m of cell size) to simulate time to ponding (Tp), runoff duration (TQ), initial runoff per raster cell (q0), Qsim and Qin in six olive plots (480 m2 per plot) during two years (108 rainfall events and 648 simulations). Two plots were managed with a mixture of plant species (CC-I), two with one single plant species (CC-II) and two with CT. Runoff yield from each plot was collected (Qobs) in gauging-stations during 27 time-integrated samples and used for modelling validation (162 control points). On average, Qobs was 9% higher under CT than under CC-I, and 8% higher than under CC-II. Topsoil saturation was simulated for the entire plots during 29 events (test-period), and Qsim appeared in another 51 and 52 events in the plots with CC and CT. Tp with CT was 2.3 times higher (59 s) than the average duration with CC and the topsoil became saturated 3.3 times faster in the inter-rows than below the trees. Values of q0 with CC were 2.3% lower than with CT and total Qsim with CC was 2% higher than with CT. However, the differences of Qsim between the different treatments were not statistically significant. The mean observed and simulated runoff coefficients were of 11 and 14%, with median values of 7 and 10%. Qsim correlated well with Qobs (Pearson ca. 0.861), and Qsim was overestimated ca. 10%. The model performed better when rainfall depth and intensity were high, and the range of variability of both Qsim and Qobs was similar. The average, best and worst Nash-Sutcliffe coefficients were 0.665, 0.791 (P6) and 0.512 (P3) and thus model simulations were satisfactory. The four plots with CC presented on average a worse performance (Kling-Gupta coefficient = 0.607) than the two plots with CT (KGE = 0.769). The lowest spatial variability of q0, Qobs, Qsim and actual available water (Waa, the sum of Qin and stored water in the soil surface) were found in the plots with CC. CT triggered higher spatial variability of runoff and higher temporal variability of runon than CC.