1. Climate change and predicting soil loss from rainfall
Peter I. A. Kinnell
University of Canberra
EGU2017

2. Modelling Soil Loss in future climates
Global Climate Change Model
Site specific climate
Plot/Hillslope Scale Erosion Catchment Erosion
USLE/RUSLE AGNPS
RUSLE2
WEPP GEOWEPP
EUROSEM LISEM
MUSLE/SWAT
EROSION 2D EROSION 3D
and so on
Downscale

3. Modelling Soil Loss in future climates
Global Climate Change Model
Site specific climate
Examine how well 3 models currently account for event soil losses from bare fallow areas in places with different climates in the USA
USLE/RUSLE The most widely used erosion model in the world
RUSLE Currently used by NRCS as a land management tool
WEPP Will take over from RUSLE2’s use by NRCS
Models that perform badly in these circumstances are UNLIKELY to perform well in modelling soil loss when climate changes in the future
Downscale

4. USLE/RUSLE Works mathematically in two steps
A1 = R K where A1 is the long term soil loss on the unit plot bare fallow area 22.1 m long on 9% slope with cultivation up and down the slope R is the rainfall-runoff erosivity factor K is the soil erodibility factor
A = A1 L S C P where L, S, C, and P are factors that account for slope lengths, gradients, vegetation and conservation practices that differ from the unit plot.
R and K are the factors that account for the geographic variation of climate and soil

5. USLE/RUSLE
R = sum EI30 / years E is total kinetic energy of the raindrops during a storm I30 is the maximum 30-minute rainfall intensity
K = total soil loss from unit plot over many years / total EI30
Event soil loss model provides the foundation for the USLE/RUSLE Ae1 = EI30 K
As calculated above, K ensures that the total predicted soil loss over the set of events considered equals the total observed soil loss for that set of events

6. Rainfall erosivity index EI30
USLE/RUSLE
Rainfall erosivity index EI30
Event soil loss model : Ae1 = EI30 K
Over prediction of low losses
Under prediction of high losses
Errors distributed randomly about the 1:1 relationship

7. Soil erodibility factor K
USLE/RUSLE
Soil erodibility factor K
Event soil loss model : Ae1 = EI30 K
K = total soil loss from unit plot over many years / total EI30
K from soil properties – the USLE nomograph equation
K values were determined from rainfall simulator experiments on bare fallow plots:
3 “storms” (1) applied to a dry surface (2) applied to the plot 24 hours later (3) applied a short time after storm 2
3 different K values : Kdry, Kwet, Kvery.wet
K = a Kdry + b Kwet + c Kvery.wet , a + b + c = 1
a, b, c chosen so that K values were appropriate for the climate in the mid west USA
Ks determined from the USLE nomograph are for a particular climate in the USA

8. Soil erodibility factor K
USLE/RUSLE
Soil erodibility factor K
Event soil loss model : Ae1 = EI30 K
The revision of the USLE provided a capacity of using soil erodibilities that in effect varied between events Ae1 = EI30 Ke
In RUSLE2, the notion that soil erodibilitry varies with the moisture status of the soil is implimented using equations that result in Ke values varying during the year with climate driven variations in rainfall and ambient temperature

9. RUSLE2 Event soil loss model : Ae1 = EI30 Ke Daily Erodibility
The temporal variations in Ke make no appreciable difference to the prediction of event soil loss
NSE(ln)
Location
USLE
RUSLE2
Holly Springs,MS
0.528
0.563
Zanesville,OH
0.646
0.644
Tyler,TX
0.342
0.372
Presque Isle, ME
0.177
0.190
Model efficiency

10. WEPP A more process based model than RUSLE2
Models rill (flow driven) erosion and interrill (raindrop driven) erosion separately
Runoff is directly considered in interrill erosivity model Di = ki I q ki = interrill erodibility, I = rainfall intensity, q = runoff rate
Rill erosion model includes limiting effect of transport capacity Dr = kr (shear stress – critical shear stress) (1 – sed load / Tc sed load) kr = interrill erodibility Tc sed load = sediment load at transport capacity

11. WEPP RUSLE2 predicts event soil loss better than WEPP
WEPP does not predict runoff and soil loss for all the erosion producing events.
WEPP runoff and erosion models calibrated to give total amounts that equaled the total observed runoff and soil loss for the set of events where WEPP predicts soil loss to occur
K for RUSLE2 calibrated for same events.
Total losses for WEPP and RUSLE2 the same as observed total
RUSLE2 predicts event soil loss better than WEPP

12. Ae1 = QR EI30 KUM , QR = runoff ratio = runoff/rain
Runoff and RUSLE2
Event soil loss model : Ae1 = EI30 Ke
Does not consider runoff as a factor in causing soil loss
Event soil loss = runoff x sediment concentration (loss per unit runoff)
Ae1 = Qe [(EI30 / Qe) Ke ] Qe = event runoff
Sediment concentration varies with EI30 per unit runoff
Analysis of bare fallow runoff and soil loss plots Sediment concentration varies with EI30 per unit RAIN
Ae1 = Qe [(EI30 / RAIN) KUM] KUM is the soil erodility associated with using Qe / RAIN x EI30 in place of EI30
Ae1 = QR EI30 KUM , QR = runoff ratio = runoff/rain

13. Runoff and RUSLE2 Event soil loss model : Ae1 = EI30 Ke
Ae1 = QREI30 KUM
Ae1 = EI30 (QR KUM)
QR KUM can be used as an alterative to Ke in RUSLE2
Largely eliminates the systematic over prediction of small soil losses
QR using observed runoff

14. QR using runoff predicted by WEPP
Runoff and RUSLE2
QR using runoff predicted by WEPP
NSE(ln)
location
RUSLE2
USLE-M with WEPP runoff
WEPP
Bethany, MO
0.325
0.317
-0.258
Holly Springs, MI
0.504
0.605
0.375
Presque Isle, ME
0.101
0.296
-0.115
Watkinsville, GA
0.505
0.362
-0.797
NSE(ln)
location
RUSLE2
USLE-M with WEPP runoff
Bethany, MO
0.325
0.317
Holly Springs, MI
0.504
0.605
Presque Isle, ME
0.101
0.296
Watkinsville, GA
0.505
0.362
Model efficiency
Reduced systematic over prediction of small losses but higher random error

15. Overview
Runoff has a major affect on soil loss generated by individual rainfall events
In theory, models that take this into account have the best capacity to account for variations in soil loss associated with changes in climate that occur in space (present day) and time (the future)
However, inaccurate prediction of the runoff is common and may result in poor performance by models that do take account of the role that runoff has affecting soil loss generated by individual rainfall events.
All rainfall erosion models are wrong, some more than others !!!!