The Precipitation and Water Cycle Data website is meant to be a user-friendly web portal for accessing numerous sets of observations, forecasts, and validation data relating to water and precipitation on the global scale. For information on how these data are acquired and processed, please refer to here.

Observed Precipitation and Previous Model Forecasts
This section is provided to display and compare observed precipitation data across the world for the previous 10 days. Under "Previous Model Forecasts" there are also two models (GFS and NOGAPS) that have been analyzed for comparison with the observation methods to the left. All data are 24 hr (12Z to 12Z) totals of precipitation for each respective day (listed at the bottom of each map plot). It is possible to select multiple data sets by holding the CTRL key and clicking on each set you want.

• CMORPH
  NOAA CPC Morphing Technique (CMORPH) is a method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution. IR data are used as a means to transport the microwave-derived precipitation features during periods when microwave data are not available at a location. For more information, click here.
• MWCOMB
  NOAA CPC Merged Microwave (MWCOMB) produces global precipitation estimates from passive microwave data using several sources and normalizing them. The difference between this and CMORPH is that there is no propagation using infrared data or any morphing involved. For more information, click here.
• PERSIANN
  The PERSIANN (Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks) system uses neural network function classification/approximation procedures to compute an estimate of rainfall rate at each 0.25° ´ 0.25° pixel of the infrared brightness temperature image provided by geostationary satellites. For more information, click here.
• GPI
  The GOES Precipitation Index (GPI) uses infrared data derived from geostationary and polar orbiting satellites and measures the cloud-top temperatures. It is a useful method when observing the tropics or the warm-season extratropics, but it is a weaker method in other areas. For more information, click here.
• Gauge
  Rain gauges are considered to be the most accurate sources of actual fallen precipitation. Rain gauge data are only available for the United States and Mexico.
• Radar
  Radar data from the National Weather Service are totaled daily and combined to form an overall national image derived from radar stations across the United States.
• GFS
  The Global Forecast System (previously known as AVN) is developed by NCEP and is a global model that predicts many variables on the ground and atmosphere. In this section, only precipitation data forecast by the most recent GFS model run for each 24-hour time period are displayed. For more information about the GFS and its future forecasts, see below.
• NOGAPS
  The Navy Operational Global Atmospheric Prediction System (NOGAPS) is a global forecast model similar to the GFS but runs using various different techniques. Only precipitation data forecast by the most recent NOGAPS model run for each 24-hour time period are displayed. For more information, click here.

Validation Results
This section comprises various analyses of observations and past model data. All validation results are for the same temporal resolution (daily, for the previous 10 days) as the observations above.

• Observations-Models
  The point of subtracting a model from an observation is to see--in the simplest way possible--how well the model's prediction performed compared to the actual observation. If the subtraction produces a positive value, this indicates that the model underpredicted the precipitation in that area. Likewise if the subtraction produces a negative value, the model overpredicted the precipitation in that area. Of course, a value close to zero will indicate very close model/observation agreement.

  Example - Some interesting comparisons can be made with these plots. By selecting 3 sets of global observations with one model (e.g. "CMORPH-GFS", "MWCOMB-GFS", and "GPI-GFS") one can see that the GFS model consistently overpredicts precipitation at 40°S-60°S latitude with the exception of GPI, where it underpredicts (this is expected--GPI is very bad at latitudes greater than 40°). If one compares models with the same observation (e.g. "CMORPH-GFS" and "CMORPH-NOGAPS"), NOGAPS seems to fair much better at 40°S-60°S latitude but does a bit more poorly in the United States than the GFS.

• Improvement Plots
  This comparison method is useful for measuring the "improvement" of two related sets of observations or forecasts over each other. The plot either uses rain gauge data for the United States (considered to be the most accurate precipitation observation) as the baseline for comparison, or CMORPH for global comparisons. Then, it applies this equation to the data sets:

  |Data1-Observation|-|Data2-Observation|

  A negative result implies that the first data set (Data1) had greater accuracy than the second (Data2). Likewise, a positive result indicates greater accuracy in the second data set. A zero result indicates perfect agreement between Data1 and Data2 but does NOT necessarily imply that they were both accurate in their forecast/observation. The spatial average has been printed onto each plot.

  Example - Relating to the previous example, if one would like to better confirm the accuracy of GFS and NOGAPS, they could choose the "GFS & NOGAPS" improvement plot. Visually it seems that at 40°S-60°S latitude there is a fairly consistent red shading, indicating that NOGAPS has more accuracy when compared to the truth (or as close to the truth as we can get!). If one looks at the United States region (this time with the "GFS & NOGAPS (US Only)" plot, which uses the more accurate gauge data), in general there is a negative spatial average, which would better confirm our suspicion that GFS fairs better than NOGAPS in the United States region.

GFS Model Forecasts (Future)
The Global Forecast System (GFS) model is arguably the United States' most used model for predicting the many variables that define the state of our atmosphere and ground surface for the entire globe. For this website, specific plots relating to precipitation and the water cycle were processed from the model output, but there are many other variables that are used for forecasting the weather as well. Click here to view most of the common weather variables, or click here for more detailed information on the GFS model.

CREW attains the GFS data for the 0Z and 12Z outputs of the model, although it is also run at 6Z and 18Z. Some plots have an analysis at the 0 forecast time, while others only provide the forecasts from +6 hours and beyond. There's no capability to display more than one GFS variable at a time as some of the plots and times would not match up. Below are brief descriptions of the plots used on this website, but there are more in-depth descriptions out there if needed.

• 6-Hour Precipitation
  The precipitation total accumulated over the past 6 hours from the forecast valid time.
• Precipitable Water
  A measure of the entire atmosphere's precipitable water at the forecast valid time. Total amount of water vapor in a vertical column of air (in this case, from the top to bottom of the atmosphere), expressed as the depth of the layer of water that would be formed if all the water vapor were condensed to liquid water.
• Cloud Water
  The total concentration of ice and liquid water particles in a cloud.
• Evaporation-Precipitation
  A measure of the global water "budget" as described by the precipitation total subtracted from the latent heat flux value. To obtain comparable units the latent heat flux values have to be converted into liquid water units by the conversion factor 28.94 W/m^2 = 1 mm/day, and the 6-hour precipitation value multiplied by 4 to obtain mm/day. The average is printed onto every plot--globally, it should hover around 0 but certain regions tend to sway one way or another. Obviously, intense blue values indicate precipitation occuring while red displays areas of evaporation.
• 0-10 cm Soil Moisture
  A percentage of the water content in the soil for the first 10 cm in to the Earth's surface.
• Soil Moisture (Avgd. Levels)
  Shows the water content of 4 levels into the Earth's surface. Note: this is not a time series! Rather, these represent the average over the 180 hour forecast period for each level.
• Surface Water Runoff
  A measure of the runoff of water reaching the ground and discharging into a stream channel or other water body.
• Snow Water Equivalent
  The water equivalent of the amount of snow on the surface of the earth.

GFS Model Validation
This section takes model predictions from the past 8 days (starting 2 days ago) and verifies them against a CMORPH precipitation observation. Computed is a 24 hour precipitation total "Tot" for each 00Z initialization time for the same common valid time (daily precip. total for 2 days ago) and also the CMORPH daily precip. total "Obs" for 2 days ago so that you can see the relative improvement in the forecasts as they become more recent.

A common way to see how the forecasts improved (hopefully) over time would be to run your mouse down a vertical column of forecast times. If you notice, for example, the 156 hour forecast initialized 8 days ago is predicting the precipitation at the same time as the 132 hour forecast 7 days ago, 108 hour forecast 6 days ago etc., all the way until 2 days ago at 12 hours. The prediction initialized 2 days ago is what is used as the basis of comparison for the "Improvement Score." As was mentioned earlier, an improvement score shows the relative strength of two forecasts compared to one base observation. The general equation used is as follows:

|Reference Forecast-Observation|-|Forecast-Observation|

The reference forecast we use is the 24 hour total precip. forecast for 2 days ago and the observation is the CMORPH observation for 2 days ago. Therefore when interpreting an improvement score we can see that a negative value implies that the reference forecast was more accurate compared to the truth than the forecast we are testing. In an ideal situation, the score should get closer to 0 as the forecasts become more recent. In some instances a positive score is computed, which means that the older forecast was more accurate than the most recent reference forecast.

A problem arises when using this method since by just looking at this improvement score you are lead to believe that a score of 0 implies perfection in a forecast when in reality all it could be saying is that the two forecasts agreed--but, they could be both equally as far apart from the actual observation! This is why an "Observation-Model" score is displayed, which simply shows how much the model over- or underpredicted compared to the observation (a negative means it overpredicted, etc.). In order to better see the "accuracy" of the model, it is necessary to look at both these scores--if they are both near 0 it then may imply a higher level of accuracy. Each set of scores is areally computed for each respective region. Added 6/5/07: A root mean square error score and plot has been added, which more accurately measures the magnitude of the error. It is computed simply as (Observation-Model)²/(Observation+Model).

Climatology
The climatology sections are intended to provide a look at monthly and yearly precipitation totals, averages, and anomalies for across the world. To compute the monthly plots, the CMORPH precipitation for each month is averaged for over the past 5 years (for as far back as we have data) and compared to what CMORPH observed for the current months of this year so far. Anomalies are defined as just (Current Monthly Observation-Average Monthly Observation), or simply by subtracting the Average from the Observed plot that is shown on the page for each month. The United States has the added bonus of rain gauge data, from which the GPCP (Global Precipitation Climatology Project) average taken over 1985-2005 for each month is subtracted. Word of caution: since the GPCP data set is in a lower resolution than the rain gauge data, mountainous regions may yield bogus results.

The yearly climatology section compares each of the CMORPH precipitation totals for the past 4 years to the average CMORPH yearly precipitation. We only have data since 2003 so these anomalies will not take into account precipitation variations greater than 4 years. However, it still generally shows areas that were affected by drought or a precipitation surplus for a given year. In the near future we hope to provide similar plots using the PERSIANN method.

Seasonal Forecasts
The COLA Atmospheric Global Circulation Model (COLA AGCM) outputs monthly (aka seasonal) forecasts that are released quarterly on their GrADS Data Server (GDS). For more information on this data set and how to retrieve it, please click here.

The forecast is displayed for 7 months after the latest model run was initiated. To plot the anomalies in this section the 1982-2002 average precipitation for each month (found in the "clim" directory on COLA's GDS) is subtracted from the most recent monthly forecast. Since the grid is two degree, caution should be placed upon the accuracy of these results in mountainous regions.