Abstract
Estimating rainfall erosivity where sub-hourly rainfall records are unavailable is a major constraint on soil-loss modelling in West Africa. This study compared five models—multiple linear regression (MLR), degree-2 polynomial regression, Random Forest (RF), support vector regression (SVR) and an artificial neural network (ANN)—for estimating monthly rainfall erosivity at five stations in South-Eastern Nigeria. Erosivity was derived from hourly NASA POWER rainfall (2005–2025) as an hourly-intensity index (EIh). Predictors were monthly rainfall, maximum daily rainfall, number of rainy days, previous-month rainfall, station and calendar month. Models were trained on 988 station-months (2005–2021) and tested on a chronologically held-out set of 247 station-months (2021–2025). RF performed best (R² = 0.915; RMSE 36% lower than MLR), closely followed by polynomial regression (R² = 0.910), ANN (0.874), SVR (0.845) and MLR (0.792). A cubic polynomial overfitted (test R² = 0.873), and MLR produced physically impossible negative estimates. Permutation importance showed that maximum daily rainfall (0.76) and monthly rainfall (0.34) carried almost all predictive power, whereas station, month, rainy days and antecedent rainfall contributed little. The degree-2 polynomial provides near-RF accuracy in a closed-form equation suitable for regional RUSLE applications.
Keywords
Rainfall erosivity, Random Forest, polynomial regression, support vector regression, artificial neural network, permutation importance, RUSLE
References
[1] Wischmeier, W.H., and Smith, D.D. (1978). Predicting Rainfall Erosion Losses: A Guide to
Conservation Planning. Agriculture Handbook No. 537. USDA, Washington, DC.
[2] Renard, K.G., Foster, G.R., Weesies, G.A., McCool, D.K., and Yoder, D.C. (1997). Predicting Soil
Erosion by Water: A Guide to Conservation Planning with the Revised Universal Soil Loss Equation
(RUSLE). Agriculture Handbook No. 703. USDA, Washington, DC.
[3] Lee, J.-H., and Heo, J.-H. (2011). Evaluation of estimation methods for rainfall erosivity based on annual
precipitation in Korea. Journal of Hydrology, 409, 30–48.
[4] Mello, C.R., Viola, M.R., Beskow, S., and Norton, L.D. (2013). Multivariate models for annual rainfall
erosivity in Brazil. Geoderma, 202, 88–102. https://doi.org/10.1016/j.geoderma.2013.03.009
[5] Xie, Y., Yin, S., Liu, B., Nearing, M.A., and Zhao, Y. (2016). Models for estimating daily rainfall
erosivity in China. Journal of Hydrology, 535, 547–558. https://doi.org/10.1016/j.jhydrol.2016.02.020
[6] Oliveira, P.T.S., Wendland, E., and Nearing, M.A. (2012). Rainfall erosivity in Brazil: A review. Catena,
100, 139–147. https://doi.org/10.1016/j.catena.2012.08.006
[7] Nearing, M.A., Yin, S., Borrelli, P., and Polyakov, V.O. (2017). Rainfall erosivity: An historical review.
Catena, 157, 357–362. https://doi.org/10.1016/j.catena.2017.06.004
[8] Salako, F.K. (2010). Development of isoerodent maps for Nigeria from daily rainfall amount. Geoderma,
156, 372–378. https://doi.org/10.1016/j.geoderma.2010.03.006
[9] Okorafor, O.O., Akinbile, C.O., Adeyemo, A., and Egwuonwu, C.C. (2017). Determination of rainfall
erosivity index (R) for Imo State, Nigeria. American Journal of Engineering Research, 6(2), 1–13.
[10] Ekpo, A.E., Orakwe, L.C., and Nwachukwu, C.P. (2021). Determination of rainfall erosivity in Akwa
Ibom State using Modified Fournier Index. Journal of Engineering and Applied Sciences, 19(1), 427–
435.
[11] Solomatine, D.P., and Siek, M.B. (2004). Application of machine learning techniques for rainfall-runoff
modelling. [Complete the source details from the original.]
[12] Valipour, M., Shahsavani, D., and Choubin, B. (2021). Data-driven modelling approaches for
hydrological forecasting and prediction. [Complete the source details from the original.]
[13] Anyanwu, S.C., et al. (submitted). Multi-scale spatio-temporal variability and trends of rainfall erosivity
in South-Eastern Nigeria from hourly satellite-derived rainfall.
[14] Breiman, L. (2001). Random forests. Machine Learning, 45, 5–32.
Conservation Planning. Agriculture Handbook No. 537. USDA, Washington, DC.
[2] Renard, K.G., Foster, G.R., Weesies, G.A., McCool, D.K., and Yoder, D.C. (1997). Predicting Soil
Erosion by Water: A Guide to Conservation Planning with the Revised Universal Soil Loss Equation
(RUSLE). Agriculture Handbook No. 703. USDA, Washington, DC.
[3] Lee, J.-H., and Heo, J.-H. (2011). Evaluation of estimation methods for rainfall erosivity based on annual
precipitation in Korea. Journal of Hydrology, 409, 30–48.
[4] Mello, C.R., Viola, M.R., Beskow, S., and Norton, L.D. (2013). Multivariate models for annual rainfall
erosivity in Brazil. Geoderma, 202, 88–102. https://doi.org/10.1016/j.geoderma.2013.03.009
[5] Xie, Y., Yin, S., Liu, B., Nearing, M.A., and Zhao, Y. (2016). Models for estimating daily rainfall
erosivity in China. Journal of Hydrology, 535, 547–558. https://doi.org/10.1016/j.jhydrol.2016.02.020
[6] Oliveira, P.T.S., Wendland, E., and Nearing, M.A. (2012). Rainfall erosivity in Brazil: A review. Catena,
100, 139–147. https://doi.org/10.1016/j.catena.2012.08.006
[7] Nearing, M.A., Yin, S., Borrelli, P., and Polyakov, V.O. (2017). Rainfall erosivity: An historical review.
Catena, 157, 357–362. https://doi.org/10.1016/j.catena.2017.06.004
[8] Salako, F.K. (2010). Development of isoerodent maps for Nigeria from daily rainfall amount. Geoderma,
156, 372–378. https://doi.org/10.1016/j.geoderma.2010.03.006
[9] Okorafor, O.O., Akinbile, C.O., Adeyemo, A., and Egwuonwu, C.C. (2017). Determination of rainfall
erosivity index (R) for Imo State, Nigeria. American Journal of Engineering Research, 6(2), 1–13.
[10] Ekpo, A.E., Orakwe, L.C., and Nwachukwu, C.P. (2021). Determination of rainfall erosivity in Akwa
Ibom State using Modified Fournier Index. Journal of Engineering and Applied Sciences, 19(1), 427–
435.
[11] Solomatine, D.P., and Siek, M.B. (2004). Application of machine learning techniques for rainfall-runoff
modelling. [Complete the source details from the original.]
[12] Valipour, M., Shahsavani, D., and Choubin, B. (2021). Data-driven modelling approaches for
hydrological forecasting and prediction. [Complete the source details from the original.]
[13] Anyanwu, S.C., et al. (submitted). Multi-scale spatio-temporal variability and trends of rainfall erosivity
in South-Eastern Nigeria from hourly satellite-derived rainfall.
[14] Breiman, L. (2001). Random forests. Machine Learning, 45, 5–32.