Skip to main content
ARS Home » Plains Area » El Reno, Oklahoma » Grazinglands Research Laboratory » Agroclimate and Natural Resources Research » Research » Publications at this Location » Publication #336131

Research Project: ADAPTING SOIL AND WATER CONSERVATION TO MEET THE CHALLENGES OF A CHANGING CLIMATE

Location: Agroclimate and Natural Resources Research

Title: Cascade rainfall disaggregation application in U.S. Central Plains

Author
item Garbrecht, Jurgen
item Gyawali, Rabi
item Malone, Robert - Rob
item Zhang, Xunchang

Submitted to: Environment and Natural Resources Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/8/2017
Publication Date: 10/27/2017
Citation: Garbrecht, J.D., Gyawali, R., Malone, R.W., Zhang, X.J. 2017. Cascade rainfall disaggregation application in U.S. Central Plains. Environment and Natural Resources Research. doi:10.5539/enrr.v7n4p30.
DOI: https://doi.org/10.5539/enrr.v7n4p30

Interpretive Summary: Hourly rainfall data are increasingly used to simulate integrated hydrological processes at a finer resolution. Long-term records of daily rainfall are common, but long continuous records of hourly rainfall are rare and must be developed. A statistical model (MRC) is proposed to disaggregate observed daily rainfall time-series into a long continuous hourly rainfall record. The model was calibrated and validated with 22 years of observed daily and hourly rainfall at the Oklahoma Mesonet Weather Station near El Reno, in Central Oklahoma. Model performance was measured as the difference between observed and disaggregated hourly rainfall characteristics. Results indicated that variations in disaggregated hourly rainfall due to the random component of the MRC model were mostly within 10% of the mean value. Only the “longest wet spell duration” and “maximum hourly rainfall” displayed higher variations when simulated with different sets of random numbers. The MRC model produced fewer large rainfall clusters compared to the number of clusters in the observed rainfall record. Also, the MRC disaggregation produced a large number of trace hourly rainfall events smaller than 0.1 mm/hr. However, interpreting trace rainfall events as “rain-free” did not improve the performance of the disaggregation model. Altogether, the overall good performance of the model leads to the conclusion that most characteristics of observed hourly rainfall were reproduced by the disaggregated hourly rainfall. As such, MRC disaggregation of daily to hourly rainfall is a promising approach to augment the short record length of existing observed hourly rainfall. However, the large number of probability combinations of the MRC model requires several years of observed hourly rainfall data to ensure that the derived/calibrated probability values are stable and reliable.

Technical Abstract: Hourly rainfall are increasingly used in complex, process-based simulations of the environment. Long records of daily rainfall are common, but long continuous records of hourly rainfall are rare and must be developed. A Multiplicative Random Cascade (MRC) model is proposed to disaggregate observed daily rainfall time-series into a long continuous hourly rainfall record. The MRC model structure consists of four time resolutions, two rainfall amount classes, four position classes, and a variable number of cascade generation states. The model was calibrated and validated with 22 years of observed daily and hourly rainfall at the Oklahoma Mesonet Weather Station near El Reno, in Central Oklahoma. Model performance was measured as the difference between observed and disaggregated hourly rainfall characteristic. Performance measures consisted of 17 statistical rainfall characteristics, including wet spell features, and rainfall cluster properties. Results indicated that: a) variations in disaggregated hourly rainfall due to the random component of the cascade model were mostly within 10% of the mean value. Two exceptions were the “longest wet spell duration” and “maximum hourly rainfall” which displayed up to 64% and 49% variation, respectively, when different sets of random numbers were used; b) the MRC model produced fewer large rainfall clusters compared to the number of clusters in the observed rainfall record; c) the MRC disaggregation produced a large number of trace hourly rainfall events (< 0.1 mm/hr). However, interpreting trace rainfall events as “rain-free” did not improve the disaggregation model performance. Nevertheless, the overall good performance of the model leads to the conclusion that most characteristics of observed hourly rainfall were reproduced by the disaggregated hourly rainfall. As such, MRC disaggregation of daily rainfall is a promising approach to augment a record of observed hourly rainfall. However, the large number of “time-rainfall-position-state” probability combinations of the MRC model requires several years of observed hourly rainfall data to ensure that the derived/calibrated probability values are stable and reliable.