Get reliable science

The potential benefits of incorporating seasonal climate forecasts in decision-making are numerous and span multiple sectors. However, many are reluctant to incorporate such information due to skepticism in forecast accuracy and the lack of forecasts distilled at scales or in metrics meaningful for applied use. To overcome these barriers we have statistically downscaled coarse scale seasonal climate forecasts from seven different models participating in the North American Multi-Model Ensemble (NMME) across the western United States. This downscaling allows us to obtain higher resolution (~4-km) predictions of temperature, precipitation, humidity , winds and insolation that are more directly usable than raw output. These data can be incorporated in additional modeling efforts to estimate germination success probabilities, hydrology and fire danger at lead times of up to 7 months. Secondly, we examined the historic forecast skill of these models using hindcasts from 1982-2010. Forecast skill tended to be higher for temperature than precipitation, although significant geographic and seasonal variability was found across models. Likewise, as forecast skill for individual months decreased with lead-time, statistically significant skill was found across over half of the western US for aggregated 3-6 month forecasts. These results suggest that climate forecasts do have the potential to inform decision making, albeit users should make use of forecast skill to decide on how useful they might be. Furthermore, climate forecasts should be treated differently than weather forecasts as model skill was much higher for seasonal aggregations of climate (e.g., accumulated precipitation over the next 6 months) than for specific days, weeks or even months (e.g., monthly precipitation 4 months from now).