Weather and climate forecasts
Overview
Weather and climate forecasts address our need to know what the weather will be like over the coming hours, days, weeks, and months. Weather forecasts refer to forecast periods of 14 days ahead or less, a time frame over which the evolution of the current atmospheric pattern can be more directly predicted, and specific and high-resolution (in time and space) predictions of weather variables can be made, at least for the first 7 days. Weather forecasts over mid-latitude continental areas like the Colorado River Basin have very high skill (~accuracy) for the first 1-3 day; skill is somewhat lower for the next 3-4 days, and then it drops off dramatically (Figure 1).
Climate forecasts or outlooks are for longer periods (14 days to several months or more), and what is usually forecasted is tendencies towards anomalies (warmer/cooler temperatures; higher/lower precipitation) over weeks, months, or seasons, rather than specific quantities (Figure 1). The sources of predictability for climate forecasts are more diverse than for weather forecasts, which depend mainly on capturing the initial conditions and patterns of the atmosphere. The skill of climate forecasts is much lower than for weather forecasts in all parts of the world (Figure 1).
Incomplete knowledge about future weather conditions is leads to uncertainty in streamflow forecasts at all time scales. Most of the forecast error in April 1st seasonal streamflow forecasts is due to the weather during the April-July streamflow forecast period diverging from what is assumed in the most-probable streamflow forecast, which is the median trajectory of April-July temperature and precipitation in the recent past (1981-2015).
Relevance
Weather forecasts (out to 10 days) are routinely used by stakeholders across the basin for guiding short-term operational decision-making, such as in agriculture, recreation, and transportation, and water resource management. NOAA Colorado Basin River Forecast Center (CBRFC) incorporates weather forecasts (out to 10 days) into their short-term streamflow forecasts and peak runoff forecasts, and also into their Seasonal streamflow forecasts.
Climate forecasts and outlooks are frequently consulted by stakeholders, including agricultural producers and water managers, but because of their relatively low forecasts accuracy/skill it is challenging to actually use them to support planning and decision-making like how weather forecasts and seasonal streamflow forecasts are used. Reclamation and CBRFC are currently exploring ways to incorporate climate forecasts into basin runoff and reservoir outlooks.
Weather forecasts
Weather forecasts have relatively high skill, and continue to improve. In the Western U.S., a forecast of daily high temperature (Tmax) made 5 days in advance has a mean error of 3-4 °F, and the probability-of-precipitation (PoP) forecast at that same 5-day lead time is far more accurate than just assuming the average PoP for that season and location. In general, today's 5-day weather forecast is as skillful as the 4-day forecast was 10 years ago, or the 3-day forecast 20 years ago.
The main source of predictability and skill for weather forecasts is the accurate description of the initial conditions of the atmosphere by global networks of observations from land and ocean stations, satellites, and aircraft. Once initialized with those observations, a weather forecast model simulates the evolution of the large-scale weather patterns. The skill of weather forecasts declines as the lead time lengthens, because the information contained in the initial patterns is gradually lost to unpredictable chaotic motions of the atmosphere.
Weather forecast models are massively complex physics-based simulation models that are run on supercomputers. They divvy up the globe into a 3-D grid and calculate changes in each gridbox over time using fundamental equations, similar to Global climate models (GCMs). For weather forecasts for the U.S., the NOAA National Weather Service relies mainly on the NOAA GFS (Global Forecast System) model, supplemented with the primary European and Canadian global forecast models, and higher-resolution U.S. regional models. These models are typically run in ensemble mode, in which 10-50 different forecasts are generated using slightly different initial conditions for each forecast, to represent the uncertainty in those starting observations. Forecasters at the local NWS offices use the forecast model output as guidance in constructing the familiar twice-daily NWS official forecasts, while considering their expert knowledge of local weather and the strengths and weaknesses of the models.
The primary weather forecasting challenge for the Colorado River Basin pertaining to water resources is predicting the track and strength of the low-pressure systems (i.e., mid-latitude cyclones) coming off the Pacific that provide the vast majority of October-May precipitation. Once these systems reach the West Coast (usually 1-3 days before affecting the basin), they can be better monitored and forecasted.
Sub-seasonal climate forecasts
Sub-seasonal climate forecasts, those made for conditions 2 weeks to 3 months ahead, are a relatively recent development leveraging advances in the understanding of climate-system processes at those timescales. Slotting into the gap in between weather forecasts and seasonal climate forecasts, sub-seasonal forecasts essentially blend the methods and sources of predictability used in both (Figure 1). Like seasonal climate forecasts, official NOAA sub-seasonal forecasts are expressed probabilistically, as the likelihood of anomalous conditions.
Sub-seasonal forecasts are mainly based on the output of two overlapping classes of models: (1) weather forecast models, such as GFS and ECMWF, being run out for a longer period, and (2) climate models (including GCMs), which have coarser temporal and spatial resolution but include more components, processes, and interactions, such as ocean currents and soil-atmosphere feedbacks. Several climate models are run in a coordinated fashion to make sub-seasonal predictions through both the North American Multi-model Ensemble (NMME; 7 models) and the Subseasonal Experiment (SubX; 7 models, including 4 also in NMME); these model predictions are used in research and, increasingly, for operational forecasting.
U.S. operational forecasts at sub-seasonal scales are issued by NOAA NWS through the Climate Prediction Center (CPC). They offer an official week 3-4 (15-28-days out) probabilistic temperature forecast, an experimental weeks 3-4 probabilistic precipitation forecast, and official 1-month (1-30 days out) forecasts for both temperature and precipitation. All of these forecasts identify areas where there are tendencies towards above-normal or below-normal conditions, much like their seasonal climate forecasts (below).
In the Colorado River Basin, the skill of the weeks 3-4 forecasts is much lower than for weather forecasts (i.e., weeks 1 and 2). Skill is generally higher in the Lower Basin than the Upper Basin, especially for precipitation. Seasonally, there are few differences, except that temperature forecast skill is especially low in the Upper Basin in the spring (March-May), a season in which cold fronts dropping across the interior West can dramatically shift the 14-day average temperature away from normal.
Seasonal climate forecasts
Seasonal climate forecasts, made for conditions 3 months to 1 year or more ahead, have been produced operationally by NOAA CPC since 1995 (as "seasonal outlooks"), and their standard 'tercile' maps, with shading to indicate "tilts" in the odds away from normal conditions, are very familiar though challenging to interpret correctly. Most of the skill in seasonal climate forecasts in the Colorado River Basin--like most places on the globe--comes from the relatively consistent tendencies associated with ENSO (El Nino-Southern Oscillation) events; i.e., El Nino and La Nina. However, those tendencies are generally weak for the Upper Basin, somewhat stronger for the Lower Basin, and variable with the season. In no case is an ENSO event, even a strong one, a sure bet for wetter/drier conditions.
Seasonal climate forecasts were initially made using only statistical methods, such as regression models, and pattern-analog methods--in both cases taking advantage of the somewhat consistent climate "footprint" of ENSO events, along with the effects of other ocean-modulated variability. In the past decade, the NOAA CPC forecasts have increasingly incorporated forecast guidance from the dynamical climate models, namely the NMME ensemble. But the skill of the forecasts across the U.S., including in the basin, is still dependent on clear-cut ENSO events; there is virtually no predictability in precipitation when ENSO-neutral conditions occur. For temperature, the skill in seasonal forecasts in the western U.S. comes primarily from the strong long-term warming trend (i.e., a forecast for warmer-than-normal is always a good bet), and secondarily from ENSO.
Looking at the historical skill for CPC seasonal precipitation outlooks over the basin, there is virtually no skill for summer, fall, and early winter (July-December) in either the Lower Basin or Upper Basin, regardless of ENSO state. For late winter and early spring (January-March), there is moderate skill for the Lower Basin during ENSO events, and low skill for the Upper Basin. In late spring (April-June), there is low skill for the Lower Basin and Upper Basin during ENSO events. Temperature outlooks have moderate skill in the Lower Basin in all seasons; and in the low-moderate skill in the Upper Basin.
Data and Tools
Weather Forecasts
NWS Homepage (access to all local forecasts)
The National Weather Service (NWS) homepage has a national map that, on-click, opens up the homepage for that local Weather Forecast Office, which has a region map that on-click brings up the NWS official local weather forecasts (nowcast to 7-day) and any hazard watches/warnings.
NWS National Digital Forecast Map
This interactive map presents a seamless mosaic of all of the local forecasts (out to 7 days) issued by the Weather Forecast Offices across the U.S., for temperature, precipitation, and several dozen other variables.
NWS WPC Quantitative Precipitation Forecast (QPF)
QPFs are specific forecasts of the most likely precipitation amounts over the next 1-7 days, issued as national maps. QPFs are used as inputs to the CBRFC's streamflow forecasts made on both short-term (10-day) and seasonal timescales. The NWS WPC QPF is based on guidance from several different weather models.
CBRFC Quantitative Precipitation Forecast (QPF) Grids
This is the QPF as used by CBRFC to drive their streamflow forecast model system. Days 2-7 are taken from the NWS WPC QPF product shown above; Day 1 comes from the very similar NWS National Blend of Models (NBM) QPF.
GEFS Ensemble Plumes (for cumulative precipitation)
This is another way of displaying a QPF (here, an 8-day ensemble forecast from the GFS weather model), for a single location. This URL shows the forecasted cumulative precipitation for KASE (Aspen, CO); zoom in on the map to select any other of ~20 locations within the Colorado River Basin. The plume shows the uncertainty in the forecast given our imperfect knowledge about the initial atmospheric conditions.
NWS CPC 8-14 day Temperature and Precipitation Outlooks
Because the skill of weather forecasts for beyond 7 days out is much lower, the official NWS "week 2" (8-14-day) forecasts are issued (daily) as probabilistic outlooks, indicating enhanced likelihoods that conditions over that period will fall into the 3 categories of above normal (upper 1/3), near-normal (middle 1/3), or below normal (lower 1/3 of all historical observations). Monthly and seasonal outlooks also use these "tercile" categories.
Sub-seasonal Climate Forecasts
NWS CPC Week 3-4 Temperature and Precipitation Outlooks
These "Week 3-4" outlooks (i.e., for the period 15-30 days after the forecast date) are issued weekly and show enhanced likelihoods that conditions for the forecast period will be in one of 2 categories: above or below the historical median. The Temperature outlook is an official forecast product, but the Precipitation outlook is still considered "experimental" due to low skill/reliability.
NWS CPC 1-month Temperature and Precipitation Outlooks
These monthly outlooks are issued for each full month (January, February, etc.) at two lead times: 0.5 months lead (i.e, issued in middle of prior month) and zero lead (end of prior month). Not surprisingly, the zero-lead monthly outlooks are more skillful since they incorporate short-term weather (0-14 days) forecast guidance, in addition to longer term guidance (equivalent to the Week 3-4 outlooks).
S2S Climate Outlooks for Watersheds
Since 29 June 2022, the forecasts on this tool have not been updated. It is not known when it will become operational again. This interactive web tool, developed by NCAR and Reclamation, presents specific ("deterministic") forecasts of temperature and precipitation anomalies (departures from normal) from several weather/climate models, which can be viewed individually or as an 7-model ensemble ("NMME") average. These models are used as guidance for the NWS sub-seasonal outlooks listed above. The sub-seasonal forecasts available are for weeks 2-3 and 3-4 (CFSv2 only); and 1-month (all NMME models), at three lead times: zero, 1 month, and 2 months; the seasonal forecasts available are for 3-month seasons at three lead times: zero, 1 month, and 2 months.
Seasonal Climate Forecasts
NWS CPC Seasonal (3-month) Temperature and Precipitation Outlooks
These seasonal outlooks are issued monthly for overlapping 3-month periods (Jan-Mar, Feb-April, etc.) at lead times ranging from 0.5 months to 12.5 months. The shorter-lead outlooks are generally more skillful, but skill is relatively low for precipitation at all lead times. As with the CPC 8-14 day outlooks, the seasonal outlooks are probabilistic and based on 3 categories (above normal, near-normal, and below normal).
Additional Resources
State of the Science Report
Chapter 7 of the State of the Science report covers many aspects of Weather and climate forecasts in much greater detail.