Weather generation using LARSWG

At present, output from GCMs is of insufficient spatial and temporal resolution and reliability to be used directly in hydrologic models. A stochastic weather generator, however, can serve as a computationally inexpensive tool to produce multiple-year climate change scenarios at the daily time scale, which incorporates changes in both mean climate and climate variability (Semenov & Barrow 1997). Stochastic weather ensures that daily values of variables are realistic, consistent, site-specific, and preserve both values and variability predicted to change from current to future climate scenarios by GCMs.

LARS-WG is a stochastic weather generator that can be used for the simulation of weather data at a single site (Semenov et al. 1998). LARS-WG is based on the series weather generator, which utilizes semi-empirical distributions for the lengths of wet and dry day series, daily precipitation and daily solar radiation. According to Semenov et al. (1998), the wet/dry time series are better represented in LARS-WG than in WGEN (Richardson & Wright 1984) and other similar weather generators. WGEN, which is the weather generator included in HELP (the hydrologic model used in this study for estimating recharge), has been known for inadequate modelling of persistent wet or dry periods (Wilks & Wilby 1999). In contrast, the serial weather generators (e.g. LARS-WG) avoid this shortcoming. These models determine sequences of dry and wet series of days, and then generate other climatic variables.

A comparison of LARS-WG and WGEN was undertaken as part of this study and, ultimately, the LARG-WG was found to more accurately reproduce the historic dataset. The base case is here defined as the average of the entire historical period, assuming that it is representative of pre-climate change conditions. Then, climate change scenarios were generated by perturbing the generated weather using the change factors to modify the base case. Each scenario consists of 100 years of generated weather, noting that although generated weather runs of 1000 years converge better to specified climate 'normals', there are diminishing returns of performance after 100 years. The length of generated weather time series is not meant to model actual changing climate year-to-year, but rather to model climate change step-wise for each scenario, and to generate a long enough weather time series to preserve and properly represent statistical properties for the site and the specified climate for the scenario. Averages were computed for monthly and annual data.

Simulated precipitation data compare favourably to observed normals for precipitation mean monthly amounts and precipitation variability (Fig. 5a), although variability in May and July were under-predicted. The LARS-WG reproduced air temperatures very precisely compared to the observed records (Fig. 5b). However, in winter months, LARS-WG produced 0.5 to 1.0°C cooler minimum temperatures than observed, when comparing variability in monthly values. Solar radiation was similarly very well reproduced using LARS-WG (Fig. 5c). Modelled mean solar radiation values were within 1% of observed values. Daily variability in daytime solar radiation was also reasonably well preserved in the stochastic weather model, although daily values were under-predicted by 5 to 10% compared to observed. This under-prediction might cause small error in evapotranspiration estimates in the HELP recharge model, once the LARS-WG weather is input into HELP.

Weather generation for climate change

As stated earlier, the base case is defined as the average of the entire historical period, assuming that it is representative of pre-climate change conditions. Then, climate change scenarios are generated by modifying the base case climatic time series (perturbing the weather) within LARS-WG. The following parameters were modified in LARS-WG according to the downscaled GCM results: precipitation relative change (future/base) or (base/base); wet spell length relative change; dry spell length relative change; absolute temperature change; standard deviation of relative temperature change; and absolute change in solar radiation. Each scenario consists of 100 years or more of simulated weather, from which monthly averages are calculated.

Was this article helpful?

0 0

Post a comment