Introduction
WEPS requires wind speed and direction in order to simulate the process of soil erosion by wind. These and other weather variables (precipitation, air temperature, solar radiation) are also needed to drive temporal changes in hydrology, soil erodibility, crop growth, and residue decomposition in WEPS.
Often it is not practical to use measured historical wind data with WEPS, since many wind records have missing data. Also, one may want to simulate wind erosion for a longer period than the length of the measured data record, e.g. for 40 years, which is the length of a typical WEPS simulation run. In addition, the measured data require much more computer disk space than wind summary statistics combined with a stochastic wind generator. Therefore, a stochastic wind generator is often more appropriate for use with WEPS than using the measured data directly.
WINDGEN was developed specifically for use with WEPS. It stochastically generates daily wind direction and hourly wind speed (van Donk et al., 2004). An earlier version of WINDGEN was described by Skidmore and Tatarko (1990). CLIGEN is the weather generator developed for the Water Erosion Prediction Project (WEPP) family of erosion models (Nicks et al., 1987). It is used by WEPS to stochastically generate daily precipitation, maximum and minimum temperature, dew point temperature, and solar radiation. Those interested in CLIGEN and how it simulates these variables should consult the WEPP documentation (Nicks et al., 1995) and the CLIGEN web site (USDA, 2004). Both CLIGEN and WINDGEN are executed under the WEPS user interface.
Statistical distributions of weather variables are needed by stochastic weather generators in order to be able to generate data. There are two steps in the stochastic generation of wind data. First, statistics need to be calculated from a historical record of measured data, describing the distributions of wind direction and speed. Second, the wind data are stochastically generated from these statistics.
Calculation of statistics used for stochastic wind generation
A quality controlled hourly wind data set (TD-6421, version 1.1), including 1304 stations in the 48 contiguous states of the USA, was obtained from the National Climatic Data Center (NCDC). Stations with less than 5 years of data were excluded, leaving 971 stations for use with a stochastic generator. Wind direction frequencies wee calculated for each of 16 directions for each month. Wind speeds less than or equal to 0.5 m/s were treated as ‘calm'. For the wind speeds that were not calm, the fraction less than or equal to certain wind speeds was calculated for each month-direction combination (12*16 = 192 combinations per station). The wind speed used were 0.5, 1.5, 2.5, ..., 20.5, 25.5, ..., 45.5 m/s. Rather than using the Weibull model, we chose to use the measured wind speed distributions themselves, without fitting to any model, but instead using linear interpolation between the measured distribution points. The reasons for this choice are described by van Donk et al. (2004).
Stochastic wind generation
First, one of the 16 cardinal wind directions or calm is selected using a random number generator with the distribution for the current month. The selected direction is applied for an entire day. Next, 24 hourly wind speeds are generated for this day. If calm was selected in the previous step, 24 wind speeds of 0 m/s are generated. Otherwise, if one of 16 directions was selected, 24 wind speeds are generated from the cumulative wind speed distribution. The distribution for the current month and wind direction is selected and a wind speed is generated from the linearly interpolated distribution, using a random number generator.
For more detail on the science behind this submodel please see the WEPS technical documentation.
References
Nicks, A.D., L.J. Lane, and G.A. Gander. 1995. Chapter 2. Weather generator. In USDA – Water erosion prediction project: Hillslope profile and watershed model documentation, eds. D.C. Flanagan and M.A. Nearing. NSERL Report No. 10, USDA-ARS, National Soil Erosion Research Laboratory, West Lafayette, IN (http://topsoil.nserl.purdue.edu/nserlweb/weppmain/docs/chap2.pdf).
Nicks, A.D., J.R. Williams, C.W. Richardson, and L.J. Lane. 1987. Generating climatic data for a water erosion prediction model. Paper No. 87-2541, International Winter Meeting ASAE, December 15-18, Chicago, IL.
Skidmore, E. L. and J. Tatarko. 1990. Stochastic wind simulation for erosion modeling. Trans. ASAE. 33:1893-1899.
USDA. 2004. Cligen Weather Generator, expanded and improved by USDA Agricultural Research Service and U. S. Forest Service. http://horizon.nserl.purdue.edu/Cligen.
Van Donk, S.J., L.E. Wagner, E.L. Skidmore, and J. Tatarko. 2004. Stochastic wind generation, comparing the Weibull model with a more direct approach. Submitted to Transactions of the ASAE.