1. Introduction
Recently, the National Centers for Environmental Prediction (NCEP)’s Environmental Modeling Center (EMC) completed the Climate Forecast System Reanalysis (CFSR) (Saha et al. 2010). The CFSR is a global coupled atmosphere–land–ocean assimilation system with a horizontal resolution of spectral T382 truncation, which is about 35 km in midlatitudes. In comparison to the NCEP Climate Data Assimilation System (CDAS), the new system has improved physics and consistent ocean and atmosphere circulations. The Global Land Data Assimilation System (GLDAS), based on the Noah model, is running along with the CFSR. The GLDAS is driven by analyzed precipitation (P) from observations. The land variables, such as soil moisture, are injected into the analysis–forecast cycle at 000 UTC each day; P, evapotranspiration (E), soil properties, sensible heat, and radiation terms are taken from the 6-h forecasts. In this paper, we evaluate the land surface variables related to long-term extreme P events and address whether the CFSR products can be used to improve drought monitoring and classification.
Currently, we use drought indices to monitor drought development (Mo 2008). For meteorological droughts, the standardized precipitation index (SPI) is used to measure precipitation deficits (Hayes et al. 1999). The SPIs can be derived from any monthly-mean precipitation time series and cover all time scales longer than one month. The standardized runoff index (SRI) derived from the monthly-mean runoff, similar to the SPI, is used to measure the deficiency of runoff. It is used to classify hydrological droughts (Mo 2008; Shukla and Wood 2008). Soil moisture monthly-mean anomaly percentiles measure the soil moisture (SM) anomaly deficit and are used to classify agricultural droughts (Maurer et al. 2002; Andreadis et al. 2005). SPI can be derived from the P analysis. The SRI and soil moisture anomaly percentiles are based on the North American Land Data Assimilation System (NLDAS) because few observations are available.
Before using the CFSR for drought classification, we need to evaluate the CFSR products. The difficulty is that there are limited observations available for these variables. That is, P and SPI from the CFSR can be compared with those from the Climate Prediction Center (CPC) unified P analyses (Xie et al. 2010). There are soil moisture measurements from Illinois and Oklahoma, while E and sensible heat data are available from the AmeriFlux station flux towers. There are also streamflow data from the U.S. Geological Survey (USGS; Lohmann et al. 2004), and surface radiation fluxes derived from satellite measurements by the National Aeronautics and Space Administration (NASA) Earth Radiation Budget Experiment (ERBE; Barkstrom et al. 1989). Unfortunately, there are no long-term gridded observations available for soil moisture or runoff. Therefore, the SRIs and SM percentiles derived from the CFSR are compared with those derived from the ensemble NLDAS and the North American Regional Reanalysis (NARR; Mesinger et al. 2006).
Currently, the NLDAS products are used to monitor drought. The NLDAS uses offline land surface models to produce SM, runoff, and E (Y. Xia et al. 2010, unpublished manuscript). All models are driven by the same forcing, which consists of P, radiation, and low-level circulation fields (Cosgrove et al. 2003a,b). There are four land surface models in the NCEP NLDAS system: Noah (Mitchell et al. 2004), Mosaic (Koster and Suarez 1992, 1994), Variable Infiltration Capacity (VIC; Liang et al. 1994, 1996), and Sacramento (SAC; Burnash et al. 1973). Because the SAC model takes potential evaporation from Noah and does not have radiation terms, it is not included in this evaluation. All models have the horizontal resolution of 0.125°.
There are large differences among the outputs from different NLDAS models even though they are driven by the same forcing. Each model treats the land surface hydrologic components differently, and land properties, such as SM and E, largely depend on the soil properties, vegetation information, model parameterization, and water-holding capacity. Dirmeyer et al. (2006) shows that the ensemble means are more representative than individual model outputs overall. Therefore, the ensemble means, defined as the equally weighted means of the three model outputs (Noah, Mosaic, and VIC) are evaluated here.
Because all models in the NLDAS ensemble are driven by the same forcing, the uncertainties of the NLDAS products may be underestimated. If drought indices from the CFSR are realistic, then the CFSR can be added to the NLDAS ensemble to increase spreads among the members. The consistency among the atmospheric variables and land properties makes the CFSR a valuable resource to monitor the development and evolution of droughts. The objectives of this paper are 1) to evaluate the land surface properties and radiation terms derived from CSFR, NLDAS, and NARR over the United States, and 2) to quantify the uncertainties of drought indices derived from three analyses. Datasets used for evaluation and brief descriptions of the NLDAS and NARR are given in section 2. An evaluation of the ensemble NLDAS against the AmeriFlux data is given in section 3. A comparison of the SPIs is given in section 4. An evaluation of E is given in section 5. The uncertainties in SRI and soil moisture percentiles are discussed in section 6 and section 7, respectively, and conclusions are given in section 8.
2. Data
a. Observations
1) Soil moisture from the Oklahoma Mesonet
The Oklahoma Mesonet soil moisture is measured at depths of 5, 25, 60, and 75 cm below the surface over 72 sites across Oklahoma (Brock et al. 1995; Basara and Crawford 2000, Robock et al. 2003). The data were collected every 30 min and transmitted to a central data ingest system at the Oklahoma Climatological Survey at the University of Oklahoma. The monthly-mean soil moisture used for verification here is calculated as time–space averages over all Oklahoma sites for the period 1997–2003.
2) Soil moisture from the Illinois State Water Survey
The soil moisture measurements are obtained from the FTP site (available at http://www.isws.illinois.edu/data.asp) for the Illinois Climate Network (ICN; Hollinger and Isard 1994; Robock et al. 2000). The ICN consists of 11 soil layers from the surface to 2 m. The data are available from 1981 onward. The measurements are given twice per month over 17 stations in Illinois. In this study, the monthly means averaged over the 17 sites are used.
3) ERBE data
The validations of the surface radiation fluxes are carried out using a dataset derived from the satellite observations obtained through NASA’s ERBE dataset (Barkstrom et al. 1989). The satellite-to-surface algorithm used to calculate shortwave flux is a physical parameterization of Darnell et al. (1988, 1992), which accounts for atmospheric transmissions, cloud attenuation, aerosol extinctions, and Rayleigh scattering. The satellite-to-surface algorithm for longwave flux by Gupta et al. (1992) is a parameterization function of effective atmospheric temperature, column water, fractional clouds, and cloud-base height. The dataset has been extensively compared with surface measurements and GCMs (Gupta et al. 1999). The error analysis of the monthly data suggests that the longwave radiation accuracy is within 6 W m−2, with a root-mean-square (RMS) error of 10 W m−2, while the shortwave radiation accuracy is within 5 W m−2 globally and 15 W m−2 for the midlatitudes. For consistent data quality, the 6 yr from January 1985 to December 1990 were selected for validation when ERBE scanning radiometers were functioning.
There are more recent data available from NASA’s Cloud and the Earth Radiant Energy System (CERES; Wielicki et al. 1996) for the period from March 2000 through October 2005. The differences between CERES and ERBE are insignificant when compared to the differences between the CFSR and ERBE; thus, only the results from ERBE are presented here.
4) Station data from the AmeriFlux network
The AmeriFlux network was established in 1996. The site information, instrumentation, principal investigators (PIs), and documentation are given online (at http://public.ornl.gov/ameriflux/). These datasets provide observations of ecosystem-level exchanges of CO2, water, energy, and momentum on an hourly basis. In this paper, we selected 21 stations over the United States for verification (Fig. 1). Their locations, vegetation, and PIs, who obtained and quality controlled the data, are listed in Table 1; 13 stations are the Global Energy and Water Cycle Experiment (GEWEX) stations and are maintained by the Atmospheric Turbulence and Diffusion Division of Oak Ridge Associated Universities. The data cover the period 1997–2007, but not all station reports are available at the same time.
It is always difficult to compare gridded analysis with station data; therefore, stations are grouped according to their geographic locations. The western region stations include Wind River Crane, Vaira Ranch, and Sky Oaks. The southeastern stations are Chestnut Ridge, Austin Cary, Duke, Canaan Valley, Goodwin Creek, and Walker Branch Watershed. The Southern Great Plains stations are those of the Atmospheric Radiation Measurement Program (ARM), Niwot Ridge, and Okmulgee. The northern plains stations are Fort Peck Indian Reservation, Black Hills, Brookings, Cottonwood, Sioux Fall, Bondville, and Colombia. We can also group stations based on the vegetation types, which have a large influence on land properties. The vegetation for each station is listed in Table 1. Note that the gridded analyses are set as undefined values when the station reports are missing.
b. Analyses
1) NARR
The NARR (Mesinger et al. 2006) is a coupled mesoscale atmospheric and land surface data assimilation system over North America with a horizontal resolution of 32 km and a temporal resolution of 3 h. It assimilates P over the entire domain. The P data are obtained from the CPC unified P analyses (Higgins et al. 2000) with the Parameter-elevation Regressions on Independent Slopes Model (PRISM; Daly et al. 1994) correction over land and the CPC Merged Analysis of Precipitation (CMAP) data over the ocean before 2007. After 2007, the CPC morphing technique (CMORPH) global precipitation analyses data are used (Joyce et al. 2004) instead of the CMAP data. The land surface model is the Noah model of 2003 (Ek et al. 2003; Mitchell et al. 2004). It covers the period from 1979 to the present. The atmospheric hydrologic cycle is realistic overall as evaluated by Mo et al. (2005). Sensible heat, E, and radiation terms were taken from 3-h forecasts in the assimilation cycle.
2) NLDAS
Three land surface models—Noah, VIC, and Mosaic—are adopted in this study. The model domain is the continental United States. The documentation and evaluation of the early version of the NCEP NLDAS system can be found in Mitchell et al. (2004). Updates and a description of the system and real-time maps are given online (at http://www.emc.ncep.noaa.gov/mmb/nldas).
The same forcing is used to drive all three land models. The forcing (Cosgrove et al. 2003a,b) fields are P, 10-m wind (both u and υ components), 2-m temperature and specific humidity, surface pressure, downward longwave radiation, and shortwave radiation. The P data used are the observed CPC/Office of Hydrologic Development (OHD) ⅛° daily rain gauge data. This dataset is derived from three sources: the National Climatic Data Center (NCDC) daily Cooperative Observer Program (COOP) stations (1948 to present), the River Forecast Center first-order stations (1992 to present), and daily accumulations from the hourly precipitation dataset (1948 to present) (Higgins et al. 2000). As with NARR, the PRISM (Daly et al. 1994) is used to adjust for topographical influences on P. The monthly-mean P from this dataset and the CPC unified P analyses (Xie et al. 2010) for the study period are similar. The other forcing terms are obtained from the NARR. To account for the surface elevation differences between the NARR and the NLDAS grid, a terrain height adjustment is applied to the air temperature and surface pressure. Because the downward shortwave radiation from the NARR has large positive biases, the climatology was corrected using the satellite-derived data from Pinker et al. (2003).
The three land surface models have different soil structure and physics. The Noah model has four vertical soil layers, located at 0–10, 10–40, 40–100, and 100–200 cm. The documentation can be found in Mitchell et al. (2004) and Ek et al. (2003). The VIC model is a macroscale hydrologic model. It is able to represent the observed nonlinear soil moisture dependence of the partitioning of precipitation into direct runoff and infiltration via subgrid parameterizations of the effects of spatial variability in soils, topography, and vegetation (Liang et al. 1994, 1996). The top soil layer is fixed at 10 cm, while the other two soil layer depths vary. The Mosaic land surface model was developed for use in the NASA global climate model (Koster and Suarez 1992, 1994). It has the same soil layers as the Noah model. The most recent activities and updates can be found online (available at http://ldas.gsfc.nasa.gov/).
The equally weighted ensemble means were computed. For each model, anomalies are defined as departures from its own climatology. The ensemble-mean anomaly is the equally weighted mean of anomalies from the three models. Both the climatology and anomalies are examined here.
Because CSFR and NARR use different versions of the Noah model and Noah is also one of the ensemble members of the NLDAS system, a description of the Noah model and its evolution are given in the appendix. For intercomparison, all datasets (NARR, CFSR, and NLDAS) were interpolated to a horizontal resolution of 0.5°. The base period is from 1979 to 2008 unless otherwise stated. To classify drought (floods) based on the SM anomaly percentiles, the percentiles need to be less than 20% (greater than 80%) (Svoboda et al. 2002; Andreadis et al. 2005). To classify drought (floods) based on SPI and SRI, indices need to be less than −0.8 (greater than 0.8) (Svoboda et al. 2002).
3. Comparison between the ensemble NLDAS and the AmeriFlux data
The outputs from the ensemble NLDAS are compared with the AmeriFlux data. For each variable, the time series of the ensemble mean was obtained at the 21 station grid locations from 1997 to 2008. The monthly climatology is the mean averaged over all 21 stations and over the period 1997–2007 when the station reports were available. The mean climatologies for both AmeriFlux (solid line) and the ensemble NLDAS (dashed line) are plotted in Fig. 2. The RMS difference between the NLDAS values and the station reports for each month averaged over all available years is also given (open circles, Fig. 2).
The net radiation has small RMS values. The maximum is just about the level of uncertainties of the AmeriFlux estimated by δ. The RMS values for the upward radiation and net shortwave radiation are small. For the net longwave radiation, the RMS difference for winter is about 30–40 W m−2, which is about 50% of the mean. For latent and sensible heat terms, there are two ways to attribute the uncertainties of the AmeriFlux measurements. We can assume that all errors come from the latent heat (Fig. 2b, dark circles and solid line) or we can assume that errors are equally distributed among latent and sensible heat terms based on the Bowen ratio (dark circles and dashed line). The latent heat from NLDAS compares well with the AmeriFlux data. Errors are within the uncertainties of the flux measurements. The sensible heat, however, does not compare well. The NLDAS values (dashed line) are larger than the AmeriFlux (solid line) in summer and the RMS value is about 45% of the mean. The ensemble NLDAS shows double maxima, but the AmeriFlux does not. The ground flux values are very small in comparison to the other terms, but the RMS values are as large as the mean. The NLDAS shows a maximum in spring (March–May), but the AmeriFlux indicates that the maximum is in June.
Figure 3 shows the comparison of energetic terms between the AmeriFlux (solid line), three NLDAS models, and the equally weighted ensemble mean of the three models (dashed line). There are large differences among individual models. Errors among models are averaged out so the ensemble mean has the best performance. For example, Mosaic (red line) has the highest latent heat values in comparison with other models and the AmeriFlux. Both Noah (green) and VIC (blue) values are smaller than the AmeriFlux in spring. VIC has a maximum in June, and Noah has a maximum in July. The ensemble-mean balances the differences and compares well with the AmeriFlux (Fig. 2b). For sensible heat, Mosaic has a maximum in July and Noah has a maximum in May. The ensemble mean still has a bimodal structure, but the values are closer to the AmeriFlux. For longwave radiation, the ensemble mean is closer to Mosaic and Noah. Values from VIC are too high. The shortwave radiation is a forcing term and is adjusted for biases, so they compare well with the AmeriFlux.
4. Meteorological drought
The severity of the meteorological drought is measured by SPI derived from P. Because P is assimilated in the NARR, P from the NARR and NLDAS is similar to the P analyses. The CFSR does not assimilate P. Instead, P is taken from the 6-h forecasts in the analysis–forecast cycle. In this section, we compare the CFSR with the NLDAS. Since the NLDAS is driven by the P analyses, the comparison between the CFSR and the P analyses will give similar results. Differences in precipitation between the CFSR and the NLDAS are given in Fig. 4 along with the differences in precipitation variances. It illustrates that the CFSR is wetter than NLDAS over the Pacific Northwest, the western mountains, and the Ohio Valley in winter [December–February (DJF)] and spring [March–May (MAM)]. The differences can be as large as 1.5–2 mm day−1. In summer, the CFSR is drier over the Great Plains with differences as large as 1 mm day−1 while it is wetter over the Southeast with differences as large as 3 mm day−1 in Florida. The differences are not limited to the climatology. Over the Southeast, the CFSR has higher variances than NLDAS in summer, but it has less variability in winter. Over the West Coast, the CFSR has more variability than NLDAS.
These errors in P have an impact on the SPI drought indices. The 3- and 6-month SPI indices computed with P from the CFSR are compared to the indices from NLDAS (Fig. 5). We recognize that the data length is short. There were also no intense drought events such as those in the 1930s and the 1950s during this period. With these limitations in mind, we compare SPI indices derived from the NLDAS with those derived from the CFSR. The correlations of SPI3 and SPI6 between them are above 0.75 over the areas west of 110°W, southern Texas, the central plains, and the Northeast. The correlations are lower over the north-central region and the area between 105° and 110°W. These are also regions where the RMS values are above 0.8. For these areas, the differences are too large for drought classification. Two examples are given to illustrate the differences. For the 1993 floods, the CFSR shows the wetness over the northern plains and the upper Missouri River basin. However, SPI6 from the CFSR is just above 2.4, while NLDAS correctly indicates severe flooding with a larger area, with the SPI6 higher than 2.4. SPI6 from the CFSR is also lower than the NLDAS over the Dakotas and Montana. Similarly, the CFSR does show dryness over the northern plains, with SPI6 nearly −2.4 for the 1988 severe drought; however, the SPI6 from the NLDAS has a larger area, with the index lower than −2.4.
To examine whether the CFSR and the NLDAS are able to detect the same drought events, we count the number of drought events from SPI6 derived from CFSR and the NLDAS. The threshold for the meteorological drought is when SPI6 is below −0.8. (Svoboda et al. 2002). At each grid point, the number of months (NUM) that the SPI6 from either the CFSR or NLDAS is below the given threshold was determined. These events can be separated into three cases. Months in which both the CFSR and the ensemble NLDAS satisfy the criterion belong to case 1. Months that the CFSR (NLDAS) satisfies the criterion but not the NLDAS (CFSR) belong to case 2 (3). The ratio between the number of months in each case and NUM is displayed in Figs. 6a–c. Consistent with Fig. 5, there are more dry months in common between the two systems over the regions where the correlations are high (Fig. 5b) and the RMS values are lower. Even for these regions, the percentages that both systems classify the same drought events are about 40%–50%. If we consider the three-category classification of drought (SPI6 < −0.8), flood (SPI6 > 0.8), and the neutral (−0.8 < SPI6 < 0.8) months, then the Heidke score (Fig. 6d) is about 0.5 except for the West Coast and the Northeast, where the score is about 0.6 or higher. This means that the CFSR does not classify one-third of months correctly in comparison to the NLDAS.
Figures 6e–g show the percentage of areas under drought conditions for the western (105°–125°W), central (85°–105°W), and eastern (65°–85°W) regions, indicated by the CFSR (red) and the NLDAS (black line). In comparison with the NLDAS, the areas under drought conditions are reasonably well captured by the CFSR.
5. Evapotranspiration
Evapotranspiration from the three systems is compared with E from the AmeriFlux (Fig. 7). Stations were grouped according to their locations (Figs. 7a–d) and vegetation (Figs. 7e–h). Evapotranspiration is at a minimum in winter, and it starts to increase in spring when vegetation starts to grow. It usually reaches a maximum in summer [June–August (JJA)] and then starts to decrease in fall [September–November (SON)]. The ensemble NLDAS (green) compares well with the AmeriFlux data (black) for all four locations (Figs. 7a–d) except over the West, where the NLDAS is slightly higher. The NLDAS also compares well for all vegetation types except over the corn field, where E from the NLDAS is higher in summer. Evapotranspiration from the NARR (red) is larger than the AmeriFlux reports for all locations and all vegetation groups except for the shrub land areas, where values are comparable. Evapotranspiration from the CFSR (blue) is higher over the Southeast and the West. Over the southern Great Plains, the CFSR has an early maximum in spring rather than in summer.
Because the ensemble NLDAS compares well with the station data, E from the CFSR and the NARR are compared with the ensemble NLDAS (Fig. 8). The differences in fall and winter (DJF) are small (not shown), so we focus on spring (MAM) and summer. The NARR is always higher than the ensemble NLDAS with largest differences (1–2 mm day−1) over the Southeast for both seasons. In JJA, the NARR is also higher than the ensemble NLDAS over the West and New England. The differences over the Great Plains are small, but the variance is also smaller in comparison to the NLDAS. One major reason the NARR has higher E is that the leaf area index (LAI) in the NARR does not vary with time. The LAI is the ratio of the total one-sided green leaf surface of vegetation per unit ground surface. The dynamic seasonal-varying LAI was implemented in the NLDAS version of the Noah model later (H. Wei et al. 2010, personal communication; appendix).
Another reason for high E is that the NARR has high positive downward shortwave radiation biases in comparison to ERBE. The largest differences are in JJA. Figure 9 indicates that the NARR values are about 60–80 Wm−2 higher than these from ERBE, with a maximum located over the West and the Southwest (Fig. 9d). The model significantly underestimates the cloud cover. That leads to high downward shortwave solar radiation (Markovic et al. 2009). The radiation is also one of the forcing terms that drive the NLDAS surface land models. For NLDAS, the positive biases were corrected based on satellite-derived solar radiation. Therefore, errors are smaller (Fig. 9c).
For MAM, the CFSR has higher E than the ensemble NLDAS over most of the United States except for the Southwest (Fig. 8a). In the Southwest, vegetation starts to grow after the monsoon starts in late June or early July. Only after the monsoon onset does E start to increase (Watts et al. 2007). Therefore, E and differences in E are small before the monsoon onset. The CFSR also has high E in MAM because of the high potential evaporation (PE) (Fig. 10). PE is defined as the evapotranspiration when soil is saturated. The transpiration term of E is proportional to a potential transpiration reduced by a soil wetness function. For both MAM and JJA, PE from the CFSR is almost twice as large as the PE values from the ensemble NLDAS. Over the Southwest, PE from the CFSR is about 20–24 mm day−1 in MAM, while the ensemble NLDAS has PE less than 12 mm day−1. The high PE is in part related to errors in radiation. The incoming shortwave radiation is a contributor to PE. Similar to the NARR, the CFSR has larger shortwave radiation in comparison to that in ERBE in both spring and summer. In JJA, the CFSR is about 40–60 W m−2 higher than ERBE (Fig. 9b).
The CFSR has higher PE than the ensemble NLDAS for both MAM and JJA. However, it has less E over the Great Plains in JJA (Fig. 8b). The reason is that the CFSR has less precipitation in JJA (Fig. 4c), which means less soil moisture in the root zone in comparison to the ensemble NLDAS (Figs. 10c and 10f). Therefore, the CFSR has less E over the Great Plains in summer in comparison to the ensemble NLDAS.
6. Hydrological drought
The hydrological drought is measured by the deficit in runoff or streamflow. Means of P for periods longer than a year are balanced by E and runoff. For NLDAS, runoff from each model is routed to calculate streamflow for both major and minor river basins. This approach is described in Lohmann et al. (1998, 2004). Figure 11d shows the annual-mean difference in runoff between the ensemble NLDAS and observations over river basins for the period 1997–2006. Observations were obtained from the USGS through their Web site (available at http://www.usgs.gov). Overall, the comparison is favorable. The differences greater than 1 mm day−1 are located over the Great Plains, where the ensemble NLDAS has higher runoff. Over the western interior mountains, the ensemble NLDAS has less runoff than observations; however, the differences are mostly less than 0.5 mm day−1.
The NARR produces too little runoff over the entire United States. The annual-mean runoff is less than 0.3 mm day−1 east of 105°W. This indicates that the NARR does not partition E and runoff correctly. It has too much E and too little runoff. For the CFSR, the comparison is favorable. For most of the United States, the differences are less than 0.5 mm day−1. Over Florida, the CFSR is dryer; however, it is wetter than the ensemble NLDAS over the Pacific Northwest.
Figure 12 shows the comparison between the SRIs derived from the CFSR and the ensemble NLDAS. Because runoff from the NARR is not realistic, it is not included in this comparison. The correlations for SRI are lower than those for SPI. The correlations between the CFSR and the NLDAS for SRI3 and SRI6 are lower than 0.5 over the central United States and the RMS differences are higher than 0.8 over the areas west of 95°W except the West Coast. Data over that region are sparse, and the runoff values are small (Fig. 11). Therefore, small differences in runoff can cause large differences in the SRI. In contrast to the SPI, both the CFSR and the ensemble NLDAS reproduce the 1993 floods and the 1988 drought and their intensity.
An important question is whether the CFSR and the NLDAS select the same hydrological drought events using the SRI6 as drought indices. The process to select drought events using SRI is the same as the process to select drought events using SPI6. The criterion used to classify drought (floods) is −0.8 (0.8) for SRI6. The results are given in Fig. 13. Over the areas east of 90°W and the West Coast, there is a 40%–60% chance that both the ensemble NLDAS and the CFSR capture the same drought events (Fig. 13a). The correspondence is weaker for areas between 95° and 115°W. If SRI6 is used to identify drought, floods, and neutral events, then the Heidke score between the CFSR and the ensemble NLDAS (Fig. 11d) is lower than 0.5 everywhere except the West Coast and the Southeast. For the interior western region, the Heidke score is less than 0.4, which means that the CFSR misclassifies 41% of the months. For example, the CFSR overestimates the percentage of areas in the West under drought from 1987 to 1995; however, it underestimates drought in this region from 2000 to 2003 (Fig. 13e). For the central and eastern regions, the agreement in the percentage of areas under drought between the CFSR and the ensemble NLDAS is good.
7. Agricultural drought
The agricultural drought is measured by the SM anomaly percentiles. Figures 14a and 14c show the comparison of soil moisture anomalies from the three systems with observations from the Illinois State Water Survey and from the Oklahoma Mesonet for the period 1997–2003, respectively. Overall, the SM anomalies from the ensemble NLDAS (green line) compare well with the observations. The RMS value between the NLDAS and the observations is 19.5 mm for Illinois and 6.25 mm for Oklahoma. The correlation is 0.73 for Illinois and 0.87 for Oklahoma. The CFSR (blue) does not compare as well. The RMS difference between the CFSR and the observations is 33.35 mm for Illinois and 13.85 mm for Oklahoma. The correlation is only 0.58 for Illinois and 0.68 for Oklahoma. The NARR (red) and the CFSR are similar, but the NARR has larger variability with larger maxima or minima.
The spreads among the three NLDAS members for Illinois and Oklahoma show that Noah (green) is very close to the ensemble mean. Mosaic (red) tends to have strong variability. The spread for Illinois is 17.8 mm among the NLDAS, and the RMS difference between the ensemble-mean NLDAS and the CFSR is 20.2 mm. For Oklahoma, the spread among the NLDAS is about 5.6 mm, and the RMS difference between the ensemble-mean NLDAS and the CFSR is about 8.1 mm.
The correlation map of SM percentiles between the ensemble NLDAS and the CFSR (Fig. 15a) resembles the correlation map of SRI6 (Fig. 12b), but correlations are higher and the RMS errors are smaller. Both the CFSR and the NARR show high correlations over the areas east of 95°W, the West Coast, and Arizona. Over the interior western mountains, the correlations are about 0.5. Over the western interior region, the RMS values between 25% and 30% are close to the spread of the NLDAS members (not shown). The CFSR captures the wet soil during the 1993 floods and the dryness of the 1988 drought. Both the pattern and magnitudes of the anomalies compare well with those of the ensemble NLDAS (Fig. 15).
The process to select drought events using the SM percentiles is the same as the process to select drought events using SPI6. The criterion used to classify drought (floods) is 20% or less (80% or more) for SM percentiles. The CFSR can detect agricultural droughts better than meteorological droughts. There is a 40%–50% chance over the United States that both systems select the same drought events (Fig. 16). Over the areas west of 100°W, there is a 40% chance that only one of the two systems will identify a drought event. The CFSR overestimates the percentage of areas under drought over the West during the period 1990–94, but it underestimates the areas under drought from 2000 to 2004. For the central regions, the correspondence is good. The CFSR overestimates areas under drought from 1993 to 1997 in the eastern region.
The Heidke score for soil moisture between the CFSR and the ensemble NLDAS (Fig. 16d) is higher than SRI6. Over the areas east of 100°W and the West Coast, the scores are higher than 0.6. The soil has a long memory, and the total soil moisture has a long decaying time. SM from the GLDAS is inserted into the CFSR at 0000 UTC during the analysis–forecast cycle. The GLDAS is driven by P analyses. This may be the major reason why the CFSR and the ensemble NLDAS compare well for soil moisture. The scores are lower over the interior region, which is the region where the observation stations are sparse. The spread among the NLDAS members is large, so the uncertainties of the NLDAS are also large.
8. Summary and conclusions
In this paper, we examine whether the CFSR products can be used for drought classification and monitoring. Drought indices are derived from precipitation, runoff, and total soil moisture anomalies taken from the CFSR 6-h forecasts. These indices are compared with the indices derived from the ensemble NLDAS and the NARR. There are three members in the NLDAS ensemble: Noah, Mosaic, and VIC. All land surface models are driven by the same forcing, which includes the precipitation analyses based on observations and solar radiation after the positive biases are corrected with satellite-derived radiation. Soil moisture, runoff, and E from the CFSR, the ensemble NLDAS, and the NARR are also compared with limited observations. Observational data include E from the AmeriFlux flux tower, soil moisture from Oklahoma Mesonet and the Illinois State Water Survey, and streamflow data from the USGS.
The CFSR has positive P biases over the western mountains, the Pacific Northwest, and the Ohio River valley in winter and spring. There are large negative biases over the Great Plains and positive biases over the Southeast during summer. These errors are not limited to the climatology, and they influence the SPIs used to measure the meteorological drought.
Evapotranspiration from the ensemble NLDAS compares well with the flux tower data from the AmeriFlux, and errors in streamflow are also small in comparison with data from the USGS. Largest errors in streamflow are about 1 mm day−1 over the central United States. In comparison with the ensemble NLDAS and the AmeriFlux, the NARR does not capture the partitioning between E and runoff. In the NARR, E is too high and runoff is too low almost everywhere over the United States.
The CFSR has positive E biases in MAM but negative biases west of 100°W in JJA except for the West Coast. One reason for large positive E biases is that the CFSR has very high PE for both MAM and JJA. PE from the CFSR is double the value from the ensemble NLDAS. The NCEP models tend to underestimate the cloud cover, leading to high downward solar radiation, which in turn contributes to the high PE and high E. The NLDAS has corrected the solar radiation biases using satellite-derived radiation (Pinker et al. 2003), while the CFSR does not. The negative biases over the Great Plains are due to less P and soil moisture in the root zone.
The evaluation on drought statistics is limited to the short data period from 1979 to 2008. There were no long-lasting severe droughts like the 1930s or 1950s in this study period, which may skew the statistics. The comparison between drought indices from the CFSR and the NLDAS is more favorable over the eastern United States and the West Coast. It is less favorable over the western interior states, where the data coverage is sparse. For the classification of drought, floods, and neutral cases, the Heidke score for SPI6 derived from the CFSR P is about 0.5, which means the CFSR misclassifies about one-third of months. The Heidke score for SM anomaly percentiles is above 0.6 except for the western interior states, where the scores are lower.
For the meteorological drought, the real-time P analyses over the United States are readily available. The P dataset starts from 1950, so the record length is longer than the CFSR data record. Therefore, the best choice for monitoring the meteorological drought is to use SPIs derived from P analyses alone instead of the CFSR or the combination of the P analyses and the CFSR. For runoff, the NLDAS compares well with the station data from the USGS; the runoff from the ensemble NLDAS will be a good choice for monitoring. For SM, uncertainties are larger because SM depends on the model parameterization and input data. Over the area east of 90°W and the West Coast, the Heidke scores are higher than 0.6. For the western interior states, the Heidke score is low; however, the uncertainties of the NLDAS-derived SM are also high. Therefore, the CFSR may be added as one member of the NLDAS ensemble for soil moisture.
Acknowledgments
The authors would like to acknowledge the PIs of the AmeriFlux datasets. These PIs are T. Meyers, G. Katul, R. Oren, B. Drake, R. Monson, W. Oechel, D. Baldocchi, Kyaw Tha Paw U, M. Fischer, and M. Torn. The first 14 stations are the GEWEX stations and are maintained by the Atmospheric Turbulence and Diffusion Division. The AmeriFlux data were created at Oak Ridge National Laboratory by the AmeriFlux and FLUXNET data management groups. These groups were supported by the U.S. Department of Energy and National Aeronautics and Space Administration. Thanks to Matthew Switanek from the University of Arizona for working on the AmeriFlux data in summer 2009. EMC would like to acknowledge support from the NCPO CPPA and CTB programs. This project is supported by the NOAA (CPPA) Grant NA09OAR4310189.
REFERENCES
Andreadis, K. M., Clark E. A. , Wood A. W. , Hamlet A. F. , and Lettenmaier D. P. , 2005: Twentieth-century drought in the conterminous United States. J. Hydrometeor., 6 , 985–1001.
Barkstrom, B., Harrison E. , Smith G. , Green R. , Kibler J. , and Cess R. , and and the ERBE Science Team, 1989: Earth Radiation Budget Experiment (ERBE) archival and April 1985 results. Bull. Amer. Meteor. Soc., 70 , 1254–1262.
Barnston, A. G., 1992: Correspondence among the correlation, RMSE, and Heidke score. Wea. Forecasting, 7 , 699–707.
Basara, J. B., and Crawford T. M. , 2000: Improved installation procedures for deep-layer soil moisture measurements. J. Atmos. Oceanic Technol., 17 , 879–884.
Brock, F. V., Crawford C. , Elliott R. L. , Cuperus G. W. , Stadler S. J. , Johnson H. , and Eillits M. D. , 1995: The Oklahoma Mesonet: A technical overview. J. Atmos. Oceanic Technol., 12 , 5–19.
Burnash, R. J., Ferrel R. J. , and McGuire R. A. , 1973: A generalized streamflow simulation system: Conceptual modeling for digital computers. Tech. Rep. of the Joint Federal–State River Forecast Center, Department of Water Resources, State of California, and the National Weather Service, 204 pp.
Chen, F., and Coauthors, 1996: Modeling of land surface evaporation by four schemes and comparison with FIFE observations. J. Geophys. Res., 101 , 7251–7268.
Chen, F., Janjic Z. , and Mitchell K. E. , 1997: Impact of atmospheric surface layer parameterizations in the new land-surface scheme of the NCEP mesoscale Eta Model. Bound.-Layer Meteor., 85 , 391–421.
Cosgrove, B. A., and Coauthors, 2003a: Land surface model spin-up behavior in the North American Land Data Assimilation System (NLDAS). J. Geophys. Res., 108 , 8845. doi:10.1029/2002JD003316.
Cosgrove, B. A., and Coauthors, 2003b: Real-time and retrospective forcing in the North American Land Data Assimilation System (NLDAS) project. J. Geophys. Res., 108 , 8842. doi:10.1029/2002JD003118.
Daly, C., Neilson R. P. , and Phillips D. L. , 1994: A statistical-topographic model for mapping climatological precipitation over mountainous terrain. J. Appl. Meteor., 33 , 140–158.
Darnell, W. L., Staylor W. F. , Gupta S. K. , and Denn F. M. , 1988: Estimation of surface insolation using sun-synchronous satellite data. J. Climate, 1 , 820–835.
Darnell, W. L., Staylor W. F. , Gupta S. K. , Ritchey N. A. , and Wilber A. C. , 1992: Seasonal variation of surface radiation budget derived from ISCCP-C1 data. J. Geophys. Res., 97 , 15741–15760.
Dirmeyer, P. A., Gao X. , Zhao M. , Guo Z. , Oki T. , and Hanasaki N. , 2006: GSWP-2: Multimodel analysis and implications for our perception of the land surface. Bull. Amer. Meteor. Soc., 87 , 1381–1397.
Ek, M. B., Mitchell K. E. , Lin Y. , Rogers E. , Grunmann P. , Koren V. , Gayno G. , and Tarpley J. D. , 2003: Implementation of Noah land surface model advances in the National Centers for Environmental Prediction operational mesoscale Eta model. J. Geophys. Res., 108 , 8851. doi:10.1029/2002JD003296.
Gupta, S. K., Darnell W. L. , and Wilber A. C. , 1992: A parameterization for longwave surface radiation from satellite data: Recent improvements. J. Appl. Meteor., 31 , 1361–1367.
Gupta, S. K., Ritchey N. A. , Wilber A. C. , Whitlock C. H. , Gibson G. G. , and Stackhouse P. W. , 1999: A climatology of surface radiation budget derived from satellite data. J. Climate, 12 , 2691–2710.
Hayes, M. J., Svoboda M. D. , Wilhite D. A. , and Vanyarkho O. V. , 1999: Monitoring the 1996 drought using the standardized precipitation index. Bull. Amer. Meteor. Soc., 80 , 429–438.
Higgins, R. W., Shi W. , Yarosh E. , and Joyce R. , 2000: Improved United States precipitation quality control system and analysis. Atlas 7, NCEP/CPC, 47 pp. [Available online at http://www.cpc.ncep.noaa.gov/research_papers/ncep_cpc_atlas/7/index.html].
Hollinger, S. E., and Isard S. A. , 1994: A soil moisture climatology of Illinois. J. Climate, 7 , 822–833.
Joyce, R. J., Janowiak J. E. , Arkin P. A. , and Xie P. , 2004: CMORPH: A method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution. J. Hydrometeor., 5 , 487–503.
Koren, V., Schaake J. , Mitchell K. , Duan Q. , Chen F. , and Baker J. , 1999: A parameterization of snowpack and frozen ground intended for NCEP weather and climate models. J. Geophys. Res., 104 , 19569–19585.
Koster, R., and Suarez M. , 1992: Modeling the land surface boundary in climate models as a composite of independent vegetation stands. J. Geophys. Res., 97 , 2697–2715.
Koster, R., and Suarez M. , 1994: The components of a SVAT scheme and their effects on a GCM’s hydrological cycle. Adv. Water Resour., 17 , 61–78.
Liang, X., Lettenmaier D. P. , Wood E. F. , and Burges S. J. , 1994: A simple hydrologically based model of land surface water and energy fluxes for GCMs. J. Geophys. Res., 99 , 14415–14428.
Liang, X., Lettenmaier D. P. , and Wood E. F. , 1996: One-dimensional statistical dynamic representation of subgrid spatial variability of precipitation in the two-layer variable infiltration capacity model. J. Geophys. Res., 101 , 21403–21422.
Livneh, B., Xia Y. , Mitchell K. E. , Ek M. B. , and Lettenmaier D. P. , 2010: Noah LSM snow model diagnostics and enhancements. J. Hydrometeor., 11 , 721–738.
Lohmann, D., and Coauthors, 1998: The Project for Intercomparison of Land-Surface Parameterization Schemes (PILPS) phase 2(c) Red–Arkansas River basin experiment: 3. Spatial and temporal analysis of water fluxes. Global Planet. Change, 19 , 161–179.
Lohmann, D., and Coauthors, 2004: Streamflow and water balance intercomparisons of four land surface models in the North American Land Data Assimilation System project. J. Geophys. Res., 109 , D07S91. doi:10.1029/2003JD003517.
Mahrt, L., and Ek M. , 1984: The influence of atmospheric stability on potential evaporation. J. Climate Appl. Meteor., 23 , 222–234.
Mahrt, L., and Pan H. L. , 1984: A two layer model of soil hydrology. Bound.-Layer Meteor., 29 , 1–20.
Markovic, M., Jones C. G. , Winger K. , and Paquin D. , 2009: The surface radiation budget over North America: Gridded data assessment and evaluation of regional climate models. Int. J. Climatol., 29 , 2226–2240.
Maurer, E. P., Wood A. W. , Adam J. C. , Lettenmaier D. P. , and Nijssen B. , 2002: A long-term hydrologically based dataset of land surface fluxes and states for the conterminous United States. J. Climate, 15 , 3237–3251.
Mesinger, F., and Coauthors, 2006: North American Regional Reanalysis. Bull. Amer. Meteor. Soc., 87 , 1–18.
Mitchell, K. E., and Coauthors, 2004: The multi-institution North American Land Data Assimilation System (NLDAS): Utilizing multiple GCIP products and partners in a continental distributed hydrological modeling system. J. Geophys. Res., 109 , D07S90. doi:10.1029/2003JD003823.
Mo, K. C., 2008: Model-based drought indices over the United States. J. Hydrometeor., 9 , 1212–1230.
Mo, K. C., Chelliah M. , Carrera M. L. , Higgins R. W. , and Ebisuzaki W. , 2005: Atmospheric moisture transport over the United States and Mexico as evaluated in the NCEP regional reanalysis. J. Hydrometeor., 6 , 710–728.
Pan, H.-L., and Mahrt L. , 1987: Interaction between soil hydrology and boundary-layer development. Boundary-Layer Meteor., 38 , 185–202.
Pinker, R. T., and Coauthors, 2003: Surface radiation budgets in support of the GEWEX Continental-Scale International Project (GCIP) and the GEWEX Americas Prediction Project (GAPP), including the North American Land Data Assimilation System (NLDAS) project. J. Geophys. Res., 108 , 8844. doi:10.1029/2002JD003301.
Robock, A., Vinnikov K. Y. , Srinivasan G. , Entin J. K. , Hollinger S. E. , Speranskaya N. A. , Liu S. , and Namkhai A. , 2000: The Global Soil Moisture Data Bank. Bull. Amer. Meteor. Soc., 81 , 1281–1299.
Robock, A., and Coauthors, 2003: Evaluation of the North American Land Data Assimilation System over the southern Great Plains during the warm season. J. Geophys. Res., 108 , 8846. doi:10.1029/2002JD003245.
Saha, S., and Coauthors, 2010: The NCEP climate forecast system reanalysis. Bull. Amer. Meteor. Soc., 91 , 1015–1057.
Schaake, J. C., Koren V. I. , Duan Q. Y. , Mitchell K. E. , and Chen F. , 1996: Simple water balance model for estimating runoff at different spatial and temporal scales. J. Geophys. Res., 101 , 7461–7475.
Shukla, S., and Wood A. W. , 2008: Use of a standardized runoff index for characterizing hydrologic drought. Geophys. Res. Lett., 35 , L02405. doi:10.1029/2007GL032487.
Svoboda, M., and Coauthors, 2002: The drought monitor. Bull. Amer. Meteor. Soc., 83 , 1181–1190.
Watts, C. J., Scott R. L. , Garatuza-Payan J. , Rodriguez J. C. , Prueger J. H. , Kustas W. P. , and Douglas M. , 2007: Changes in vegetation condition and surface fluxes during NAME 2004. J. Climate, 20 , 1810–1820.
Wielicki, B. A., Barkstrom B. R. , Harrison E. F. , Lee R. B. III, Smith G. L. , and Cooper J. E. , 1996: Clouds and the Earth’s Radiant Energy System (CERES): An Earth Observing System experiment. Bull. Amer. Meteor. Soc., 77 , 853–868.
Xie, P., Chen M. , and Shi W. , 2010: CPC unified gauge-based analysis of global daily precipitation. Preprints, 24th Conf. on Hydrology, Atlanta, GA, Amer. Meteor. Soc., 2.3A. [Available online at http://ams.confex.com/ams/90annual/techprogram/paper_163676.htm].
APPENDIX
Noah Model and Its Evolution
The Noah land surface model (LSM) is targeted to be a computationally efficient model of intermediate complexity for use in operational numerical weather prediction and seasonal climate forecast models. Taken from the original Oregon State University (OSU) LSM (Mahrt and Pan 1984; Mahrt and Ek 1984; Pan and Mahrt 1987), the Noah LSM uses a Jarvis–Stewart canopy conductance approach and a linearized (noniterative) solution to the surface energy balance; it carries multiple soil layers (nominally four) with predicted states of soil moisture, ice, and temperature (using soil moisture diffusion and heat conduction equations) along with intercepted canopy water (Chen et al. 1996; Ek et al. 2003).
The Noah LSM is formulated with a single-layer snowpack (which includes snow density) and frozen soil physics (Koren et al. 1999; Livneh et al. 2010) and includes the effect of snow cover patchiness on surface fluxes. The infiltration scheme follows that of Schaake et al. (1996) for the subgrid variability of precipitation runoff and soil moisture. Surface exchange coefficients (and thus surface fluxes) are determined via the surface layer parameterization described in Chen et al. (1997).
The Noah models in the NARR, NLDAS, and CFSR have major differences.
a. NARR (2003 version of the Noah model)
The NARR has the 2003 version of the Noah model, which is the earliest of the three versions here. Patchy snow cover effects were considered only for sensible and soil (ground) heat flux. The surface exchange coefficient yields lower sensible heat flux and greater surface skin temperatures (compared to the CFSR), though it might be hard to quantify the improvements because there are other physics changes included.
b. CFSR (2009 version of the Noah model)
The CFSR used the 2009 version of the Noah model, which is based on Noah in the Global Forecast System (GFS), which is closest to Noah version 2.7.1 (available online at http://www.emc.ncep.noaa.gov/mmb/nldas/), and version 3.0 in the Weather Research and Forecasting Model (WRF)/Noah. Here patchy snow cover effects were considered for ALL surface fluxes, and bare soil evaporation falls off more rapidly with drying soil compared to the NARR.
The surface exchange coefficient (Ch) used to calculate surface turbulent sensible and latent heat fluxes is greater than in the NARR (and NLDAS) since Ch has a GFS origin; therefore, in general, sensible heat fluxes are higher and skin temperatures are lower compared to the NARR and NLDAS, although it might be hard to quantify the improvements because there are other physics changes included.
c. NLDAS
This version is similar to the CFSR version, but it also including refinements to snow albedo. This also has a Ch value such that sensible heat fluxes are lower (higher) and skin temperatures are higher (lower) compared to the NARR (CFSR). The NLDAS version also includes the improvements from the CFSR version, which are listed below.
1) For warm season
Replace constant LAI treatment with a seasonally and spatially varying LAI;
Allow a seasonally varying efficiency of root-zone water uptake (via soil temperature);
Change the function for the vapor-pressure deficit term in the canopy resistance;
Change the upper threshold of soil moisture at which the vegetation feels soil moisture deficit;
Change the minimum stomatal resistance parameter for a few vegetation classes;
Change the treatment for the diurnal variation of surface albedo;
Reduce the Czil parameter in the formulation for the roughness length for heat (to increase the daytime aerodynamic conductance);
Increase total runoff to match observed streamflow;
Reduce large positive bias for evaporation by calibrating against observations from the ARM/CART site and reduce negative bias for skin temperature.
2) For cold season and hydrology
Improve Noah snow albedo formulation with different snow albedo for accumulation and depletion phase and raise maximum snow albedo: both work to increase snow water equivalent (SWE) and overcome early snowmelt;
Change the saturated hydraulic conductivity (Ksat) and Kdt to improve runoff; Ksat is a function of soil texture class and a highly nonlinear function of soil moisture; Kdt is an intermediate quantity and a function of Ksat (Ek et al. 2003);
Fix numerical instability during large rain and drainage rates to overcome extremely large runoff and soil moisture values in some grid points;
Apply Drew Slater modification to Ch surface exchange coefficient in stable regime to reduce sublimation and increase SWE.
Locations of the AmeriFlux stations.
Citation: Journal of Hydrometeorology 12, 2; 10.1175/2010JHM1310.1
(a) Monthly-mean uncertainties of the flux measurements defined in Eq. (1) averaged over all AmeriFlux station reports. (b) Monthly-mean latent heat averaged over all AmeriFlux station reports (solid line), the ensemble NLDAS (dashed line), the RMS (open circles), and uncertainties of the AmeriFlux data if δ is attributed to the latent heat term only (dark circle and solid line) or δ is divided between sensible and latent heat according to the Bowen ratio (dark circles and dashed line). (c) As in (b), but for monthly-mean sensible heat. (d) Monthly-mean ground flux averaged over all AmeriFlux station reports (solid line) and the ensemble NLDAS (dashed line) and the RMS values (open circles); the unit is W m−2. (e) As in (d), but for net radiation. (f) As in (d), but for net longwave radiation. (g) As in (d), but for upward longwave radiation. (h) As in (d), but for net shortwave radiation.
Citation: Journal of Hydrometeorology 12, 2; 10.1175/2010JHM1310.1
(a) Monthly-mean latent heat (W m−2) averaged over all AmeriFlux station reports (solid black line): Noah (green), Mosaic (red), VIC (blue), and the ensemble NLDAS (black dashed line). (b) As in (a), but for monthly-mean sensible heat. (c) As in (a), but for monthly-mean longwave radiation. (d) As in (a), but for monthly-mean shortwave radiation.
Citation: Journal of Hydrometeorology 12, 2; 10.1175/2010JHM1310.1
Difference between precipitation from CFSR and NLDAS averaged from 1979 to 2008 for (a) DJF, (b) MAM, (c) JJA, and (d) SON. Contour interval is 1 mm day−1. Zero contours are omitted. Contours −0.5 and 0.5 mm day−1 are added. (e)–(h) As in (a)–(d), but for the differences of variances.
Citation: Journal of Hydrometeorology 12, 2; 10.1175/2010JHM1310.1
(a) Correlation between SPI3 derived from the CFSR and NLDAS. Contour interval is 0.25. (b) As in (a), but for SPI6; (c) SPI6 for July 1993 from the CFSR. Contour interval is 0.8. (d) As in (c), but for June 1988. (e),(f) As in (a),(b), but for RMS values. Contour interval is 0.4. (g),(h) As in (c),(d), but from NLDAS.
Citation: Journal of Hydrometeorology 12, 2; 10.1175/2010JHM1310.1
(a) Ratio × 100 between the number of months that both CFSR and NLDAS declared drought and the number of months that either CFSR or NLDAS declared drought. Drought is defined when the SPI6 is −0.8. Contour interval is 10. (b) As in (a), but for number of months that NLDAS declared drought, but not CFSR. (c) As in (a), but for number of months that the CFSR declared drought but not NLDAS. (d) The Heidke score. Contour interval is 0.1. (e) Percentage of areas under drought over the West (105°–125°W) for CFSR (red line) and NLDAS (black). (f) As in (e), but for the central United States (85°–105°W). (g) As in (e), but for the eastern United States (65°–85°W).
Citation: Journal of Hydrometeorology 12, 2; 10.1175/2010JHM1310.1
Evapotranspiration from AmeriFlux stations (black line), ensemble NLDAS (green line), NARR (red line), and CFSR (blue line) averaged over the AmeriFlux sites for the (a) West, (b) Southeast, (c) southern Great Plains, and (d) northern Great Plains, and for (e) grassland vegetation, (f) corn, (g) forest, and (h) shrub land.
Citation: Journal of Hydrometeorology 12, 2; 10.1175/2010JHM1310.1
Difference of E between the CFSR and the ensemble NLDAS averaged from 1979 to 2008 for (a) MAM and (b) JJA. Contour interval is 0.5 mm day−1. Zero contours are omitted. Values >(<)1 mm day−1 (−1 mm day−1) are shaded dark (light). (c) As in (a), but for the difference of variance between the CFSR and the ensemble NLDAS. (d)–(f) As in (a),(c), but for the NARR.
Citation: Journal of Hydrometeorology 12, 2; 10.1175/2010JHM1310.1
Downward solar radiation averaged for JJA from (a) ERBE, (b) CFSR, (c) ensemble NLDAS, and (d) NARR. Contour interval is 20. Values >280 (320) W m−2 are shaded light (dark).
Citation: Journal of Hydrometeorology 12, 2; 10.1175/2010JHM1310.1
(a) PE averaged for MAM from CFSR. Contour interval is 5 mm day−1. Values >10 (20) mm day−1 are shaded light (dark). (b) As in (a), but for JJA. (c) Volumetric SM at root zone. Contour interval is 0.1. Values >0.2 (0.3) are shaded light (dark). (d)–(f) As in (a)–(c), but for the ensemble NLDAS.
Citation: Journal of Hydrometeorology 12, 2; 10.1175/2010JHM1310.1
Annual-mean runoff (mm day−1) for (a) ensemble NLDAS. Contours are indicated by the color bar. (b) As in (a), but for the runoff difference between ensemble NLDAS and CFSR. (c) As in (b), but for NARR. (d) Streamflow differences between the ensemble NLDAS and observations.
Citation: Journal of Hydrometeorology 12, 2; 10.1175/2010JHM1310.1
As in Fig. 5, but for SRI.
Citation: Journal of Hydrometeorology 12, 2; 10.1175/2010JHM1310.1
As in Fig. 6, but for SRI.
Citation: Journal of Hydrometeorology 12, 2; 10.1175/2010JHM1310.1
(a) Total SM anomaly averaged over Illinois stations from observations (black line), NARR (red line), CFSR (blue line), and ensemble NLDAS (green). (b) As in (a), but for each individual ensemble member: Mosaic (red), VIC (blue), and Noah (green), and the equally weighted ensemble-mean NLDAS (dark black line). (c),(d) As in (a),(b), but for SM at 60 cm from Oklahoma.
Citation: Journal of Hydrometeorology 12, 2; 10.1175/2010JHM1310.1
(a) Correlation between SM anomaly percentiles derived from CFSR and the ensemble NLDAS. Contour interval is 0.25. Values >0.5 (0.75) are shaded by light (dark). (b) As in (a), but for NARR. (c) SM percentile for July 1993 from CFSR. Contour interval is 10%. Values >80% (90%) are shaded by light (dark). (d) As in (c), but for June 1988. Values <10% (20%) are shaded by dark (light). (e),(f) As in (a),(b), but for RMS values. Contour interval is 5%. (g),(h) As in (c),(d), but from the ensemble NLDAS.
Citation: Journal of Hydrometeorology 12, 2; 10.1175/2010JHM1310.1
AmeriFlux stations.
