|

Description

This model explores multiple time shifting options that are possible using the built-in time shifting capability of the time series element. You can run long simulations and reuse data from a short time series or run a short simulation and sample pieces from a long time series record. This model shows multiple methods that can apply to a wide array of applications. To test these options, run the model and enter one of the four submodels below. The 4 options include:

  1. Short duration simulation with a long time series (Calendar)
  2. Short duration simulation with a long time series (Elapsed Time)
  3. Long duration simulation with a short time series (Calendar time)
  4. Randomly sample multiple time series elements

Sequentially Sample from Long Time Series (Calendar Time)

This calendar time model uses the time shifting capability included in the TS element to lookup values from the historic record and map those to the month of the simulation. This model runs for 1 year (12 months) and 82 Monte Carlo realizations. On the first realization, the monthly values from the first year of the historic record are chosen and displayed in the result. On the next realization, GoldSim jumps ahead to the next year. This process is repeated for all realizations. No more than 82 realizations can be run because GoldSim will throw a fatal error when the starting time is pointing to a date outside the range of the time series.

The index used to look up values in the time series is an annual counter starting with the calendar year 1921. The equation

1921 + (Realization - 1)

is used to increment the year on each realization.

Randomly Sample from Long Time Series (Elapsed Time)

This elapsed time model uses the random "annual" time shifting capability included in the TS element to lookup values from the historic record and map those to the month of the simulation. This model runs for 1 year (12 months) and 100 Monte Carlo realizations. On the first realization, the monthly values from a random starting year of the historic record are chosen and displayed in the result. On the next realization, GoldSim jumps to another random staring year. This process is repeated for all realizations.

Sequentially Sample from Short Time Series (Calendar Time)

This calendar time model uses the time shifting capability included in the TS element to lookup values from the historic record and map those to the month and year of the simulation. This model runs for 10 years with an hourly time step. Because the simulation is longer than the time series record, when GoldSim hits the end of the time series, it wraps back around to the beginning of the series. The wrapping effect causes GoldSim to reuse the data in the time series over and over during a single realization. The time series contains 2 years of historic data so every 2 years during the simulation, the data repeats itself.

Randomly Sample from Multiple Time Series (Calendar Time)

This calendar time model uses the time shifting capability included in the TS element to lookup values from multiple historic records and map those to the month of the simulation. If the units of the time series data are not consistent, then you must use multiple time series elements to represent the data. If you need to randomly sample the time series data, then the annual periodocity option in the time series element will not align with other time series elements. In cases like this, you will need to create your own "random year" selector and link that to your time series element using the option to align the data with a start year.

This model runs for 5 years and each realization, picks a random start year and then applies that start year to multiple time series elements so that the time series all correspond to each other. In this case, the randomly sampled rainfall, is paired with the randomly sampled discharge.

 

Making Better Decisions In An Uncertain World