In KNN-WG V1.1, you can use ensemble average for N times run.

The KNN-WG is a tool for lead time simulation of daily weather data based on K-nearest-neighbor approach. The user can load seven different variables, for example Tmin, Tmax, Rain, Srad, ETo, WSPD, and Humidity. Then, the user can load the input data and run KNN-WG. In this software the user can draw graphs and calculate efficiency criteria: d, NSE, RMSE, MBE, Pearson and Spearman. The user can compare outputs of KNN-WG with other models such as Lars-WG, SDSM, CMIP5, and etc. One of the most advantages of the KNN-WG software is that the user can select every number of variables (from 7 variables) to generate future data.

KNN-WG

What is non-parametric techniques?

Non-parametric re-sampling procedures are an alternative to generating daily weather data. Unlike parametric alternatives for time series generation, non-parametric approaches generate new values by conditionally resampling past observations using a probability rationale. Generally, a statistical method for generating daily weather sequences needs to consider the statistical dependence of the weather variables with each other on the same day. By finding this relation, we can generate future data (by similarity between days). This method can be employed on the assumption that the weather during the target year is analogous to the weather recorded in the past.

Non-parametric weather generators have evolved as a simple way to simulate climate data at multiple sites while avoiding the limitations of the parametric and semi- parametric approaches. A common drawback is that most parametric and semi-parametric models require statistical assumptions

regarding the probability distributions of climate variables.

What is KNN method?The K-nearest neighbors (K-NN) is an analogous approach. This method has its origin as a non-parametric statistical pattern recognition procedure to distinguish between different patterns according to a selection criterion. Through this method, researchers can generate future data. In other words, the KNN is a technique that conditionally resamples the values from the observed record based on the conditional relationship speciﬁed. The KNN is most simple approach.

The most promising non-parametric technique for generating weather data is the K-nearest neighbor (K-NN) resampling approach. The K-NN method is based on recognizing a similar pattern of target ﬁle within the historical observed weather data which could be used as reduction of the target year (Young, 1994; Yates, 2003; Eum et al., 2010). The target year is the initial seed of data which, together with the historical data, are required as

input ﬁles for running the model. This method relies on the assumption that the actual weather data observed during the target year could be a replication of weather recorded in the past. The k-NN technique does not use any predeﬁned mathematical functions to estimate a target variable.

Actually, the algorithm of this method typically involves selecting a speciﬁed number of days similar in characteristics to the day of interest. One of these days is randomly resampled to represent the weather of the next day in the simulation period. The nearest neighbor approach involves simultaneous sampling of the weather variables, such as precipitation and temperature. The sampling is carried out from the observed data, with replacement.

The K-NN method is widely used in agriculture (Bannayan and Hoogenboom, 2009), forestry (Lopez et al., 2001) and hydrology (Clark et al., 2004; Yates et al., 2003).