1. Introduction
Reanalyses aim to construct a continuous and complete picture of the weather and climate by constraining evolving model forecasts with a large but heterogeneous mix of observations having different temporal availability, accuracy, and degree of correspondence to model state variables and fluxes. The resulting reanalysis products have proven useful in characterizing the climate (Trenberth et al. 2011) and weather (O. Reale and M. Cordero-Fuentes 2016, unpublished manuscript), though the diversity and ephemeral nature of the assimilated observations and remaining systematic biases in the forecast model each bring uncertainty to the eventual reanalysis data. During the analysis of the observations, the background forecast model is compared to the available observations, and the departure is reflected by the analysis increment. The analysis increment can vary in space and time, and even in the diurnal cycle (e.g., Bosilovich et al. 2015a). In the atmospheric water budget, this leads to an imbalance in the long term average of the physical and dynamical terms, and reflects systematic model biases.
For example, high tropical warm pool precipitation (Bosilovich et al. 2008) and trends in water vapor transport (Robertson et al. 2014) have been noted. The atmospheric water budget may also require large increments locally to preserve a closed atmospheric budget in the assimilation data (Bosilovich et al. 2015a). The analysis increment is a crucial component of the water cycle in reanalyses (Roads et al. 2002; Bosilovich et al. 2011).
Precipitation is a key meteorological observable quantity of importance to climate and prediction research, as well as to any number of societal concerns. It is closely tied to short-term local dynamics and physics and also to the large-scale atmospheric water vapor, climate circulation, and energy balance. In this way, the global water cycle spans temporal and spatial scales, while its variability is related to significant weather and climate events. However, in reanalyses, the assimilation of observations can significantly affect the weather and climate representation of all terms of the water cycle. Dynamical quantities in reanalyses have more fidelity in the upper levels and mid-to-high latitudes but become more uncertain in the tropics and lower troposphere (Rienecker et al. 2011). Trenberth et al. (2011) characterized the water cycle using a multitude of reanalyses. The variations among the reanalysis data are substantial, and the effect of the observations on the interannual variability is apparent. This demonstrates the uncertainty of the reanalyses, even for globally averaged values. However, Trenberth et al. (2011) note that moisture divergence determined from water vapor transport is more robust compared to that determined from the water flux terms (evaporation and precipitation). ERA-Interim, with a smaller estimated increment, produces tropical atmospheric divergence in line with merged evaporation and precipitation observations (Brown and Kummerow 2014). The assimilation of observations (and resulting analysis increment) plays a substantial role in each reanalysis’s water cycle, but it is also different in each reanalysis.
While the water cycle was a major consideration during the development of the Modern-Era Retrospective Analysis for Research and Applications (MERRA; Rienecker et al. 2011), a number of issues became apparent after a long integration with its water cycle and variability, including sensitivity to the changes in the observing system (e.g., Robertson et al. 2011). With developments of the GEOS-5 system in subsequent years, including some intended to address the impact of the changing observing system, an intermediate version of the reanalysis system was used to produce a new reanalysis, MERRA-2 (Gelaro et al. 2016, manuscript submitted to J. Climate). These changes have a significant effect on the interannual variability of the MERRA-2 water cycle, compared with other contemporary reanalyses. The motivation for this investigation is to explain the MERRA-2 water cycle data, especially the effects of the water vapor analysis increments on the physical terms in the water cycle. To accomplish this we will characterize the global water cycle compared to a merged and balanced observational dataset and also evaluate the temporal variations of the budget terms (section 4). From this we can determine the ocean–land transport of water in MERRA-2. The investigation will focus on the spatial and temporal variability of the analysis increment and water cycle terms, the water vapor transport from ocean to land, the controls on ocean evaporation, and the relationship between temperature and water vapor (sections 5 and 6). While we will provide some comparison with independent observations and other reanalyses, we have performed climate simulations with the MERRA-2 model, in order to assess the model climatology in the absence of the observational constraints other than lower boundary forcing.
2. Data
a. Reanalyses
Presently there are several contemporary reanalysis datasets and different configurations of reanalyses that can provide context for the changes in the water cycle from MERRA to MERRA-2. ERA-Interim is the latest from ECMWF (Dee et al. 2011), and it supersedes their previous reanalysis (ERA-40; all acronyms are listed in appendix B). Likewise, The Japanese 55-year Reanalysis (Ebita et al. 2011; Kobayashi et al. 2015) supersedes the previous JRA-25, including finer resolution, newer model physics, and improved data assimilation. The NCEP CFSR is the newest reanalysis from NCEP (Saha et al. 2010); while it serves a different purpose than the older NCEP reanalyses, which are still producing current data analyses, we will only consider CFSR because it has a higher resolution and uses the same data assimilation method as MERRA and MERRA-2. Some comparisons between MERRA and the previous generation of reanalyses (ERA-40, NCEP-2, and JRA-25) can be found in the literature (e.g., Trenberth et al. 2011; Bosilovich et al. 2008, 2011). These are not included here to simplify the discussion to the contemporary reanalyses.
In addition, we have included data from the reduced-observations century-scale reanalyses because these can provide a realistic depiction of the state fields but do not suffer spurious changes due to satellite instrument changes over time. The NOAA Twentieth Century Reanalysis is a 50-member ensemble assimilating only surface pressure observations (Compo et al. 2011). Similarly, ERA-20C assimilates the surface station pressure but also the surface winds (Poli et al. 2016), and its companion, ERA-20CM, is an ensemble of climate model simulations (Hersbach et al. 2015). While these data extend back to the late 1800s, we will only use the data that overlap with the satellite-era reanalyses. Distinct differences of the water cycles and transport of water vapor between satellite data reanalyses and reduced-observation reanalyses has been documented where the global ocean to land moisture transport increases over time in satellite data reanalyses much more than in the other estimates (Robertson et al. 2014).
Robertson et al. (2014) estimated global moisture transport from ocean to land using three multiple sources: evaporation minus precipitation (E − P) over the global oceans, reanalysis moisture flux convergence over land, and observationally constrained land surface model precipitation minus evaporation (P − E). The LSM P − E was derived from the ensemble average of four different observationally constrained estimates of P − E over land. Robertson et al. (2014) found that this collection represented a temporally stable record of moisture flux convergence over land, free of unphysical trends and discontinuities present in reanalyses and ocean area merged satellite water flux datasets. Here, we extend the use of this collected data to provide a comparison with the MERRA-2 moisture flux convergence over land.
The project overview for MERRA-2 is available from Gelaro et al. (2016, manuscript submitted to J. Climate). For MERRA-2, data (1980–2014) were acquired from several different monthly mean data collections, three-dimensional state fields (GMAO 2015a), instantaneous vertical integrals (for total column water; GMAO 2015b), time-averaged vertical integrals (GMAO 2015c), surface flux diagnostics (GMAO 2015d), and three-dimensional moisture tendencies for the analysis increments (GMAO 2015e). MERRA data collections map identically into similarly named files, as described by Rienecker et al. (2011).
b. Observations
Several merged satellite–in situ retrieved observation datasets are used to compare with the reanalyses and models here. GPCP precipitation combines gauge measurements with satellite data to cover the globe with precipitation measurements (Adler et al. 2003). The monthly data provides a continuous global precipitation estimate from 1979 through the present. While several uncertainties exist, including greater uncertainty before routine microwave observations were included in late 1987 and also some variations when those become limited, the estimates are as reliable as we have for the entire globe. Remote Sensing Systems (2016) provides total precipitable water observations using SSM/I instruments over ice-free ocean areas. Mears and Wentz (2009) describe the RSS observations of the temperature of the lower troposphere, available globally from 1979 to the present. While these merged data products and the reanalyses use many of the same instruments, the differences in the methodologies make the comparison useful. For example, while the reanalyses assimilate the SSM/I radiances, they are integrated with other satellite instruments and the background model to produce the eventual analysis fields. These can be quite different from retrievals.
3. Changes from MERRA to MERRA-2
Gelaro et al. (2016, manuscript submitted to J. Climate) describe the fundamental changes to the reanalysis system for MERRA-2 that differentiate it from MERRA (Rienecker et al. 2011) and provide more information about the system components, data, and evaluation. The prescribed SST data are described in Bosilovich et al. (2015b). For total solar irradiance, the 11-yr cycle of Lean (2000) plus the background adjustment from Wang et al. (2005), as recommended for phase 5 of the Coupled Model Intercomparison Project (CMIP5), has been implemented in MERRA-2. MERRA and MERRA-2 share the same greenhouse gas forcing, as discussed in Rienecker et al. (2008), where climatology consists of monthly zonal averages to which each of the gases are relaxed. Aerosols are assimilated and interactive with the atmospheric radiation. Some of the changes in MERRA-2 have direct effect on the water cycle. The first of these concerns is a major change in which moisture analysis increments are handled (Takacs et al. 2015, 2016). Until recently, global reanalysis systems have treated total atmospheric mass as the analyzed variable and have not explicitly enforced the invariance of the global atmospheric dry mass. As a consequence, these systems have also not enforced consistency between global E − P and changes in atmospheric total water mass storage. The MERRA-2 system now employs these constraints [see Takacs et al. (2015) for details on the formulation and implementation of these changes] by virtue of four modifications: 1) including a term for sources and sinks of water substance in the mass continuity equation,1 2) adding a term in the GSI analysis cost function that penalizes global mean departures of dry atmospheric mass from a preset constant value, 3) in the IAU methodology, adding a similar adjustment to preserve invariance of global mean dry mass during the corrector segment of the assimilation, and finally, 4) forcing the global mean increments of atmospheric water mass toward zero.
There are some significant outcomes from these changes in terms of the global atmospheric water cycle identified by Takacs et al. (2016). Changes in global atmospheric total mass are equivalent to changes in water storage—the global mean dry mass (surface pressure) now remains specified constant. For MERRA-2 this is the climatological surface pressure taken from MERRA. Locally, though, the water vapor analysis increment does not vanish and can rival that of the physical terms of vertically integrated moisture transport, P, E, and atmospheric moisture storage. The latter is typically quite small on monthly and longer time scales.
The immediate impact of the constraint can be seen in the global mean monthly time series of the water budget terms (Fig. 1). In MERRA, the unconstrained water vapor analysis leads to global precipitation and evaporation that are quite different from each other. Furthermore, because the analysis increment is sensitive to the amount and type of observations being assimilated, MERRA’s water vapor increment is seen to have changed sign when the AMSU-A radiances entered the data stream. In MERRA-2, the global average water vapor analysis increment term is essentially zero (negligible magnitude) so that monthly evaporation and precipitation difference is the total time rate of change of the atmospheric water (Fig. 1), which is typically small for global monthly averages, and so evaporation and precipitation follow each other closely. Of course, while increments of total moisture are constrained so global evaporation and precipitation balance, locally these increments can still be as important as other physical terms in the moisture conservation equations. Model physics biases interacting with evolving observing system coverage and error properties can sometimes conspire to produce increments with amplitudes comparable to physical processes. Figure 2 shows the time series of the global spatial variance of the MERRA and MERRA-2 monthly mean water vapor analysis increments. The magnitudes are generally similar, and here, it becomes apparent that locally the changes in the observing system (as indicated by jumps near 1987 and 1995) will still play a role in MERRA-2’s regional water cycle.
Global time series of monthly mean water cycle terms (mm day−1) for (left) MERRA and (right) MERRA-2: (top) global precipitation (P, red) and evaporation (E, black) and (bottom) the global E − P increment (blue), the analysis increment (ANA, green), and the E − P + ANA (pink). The MERRA-2 analysis increment is plotted but is near zero by design as described in the text. The E − P + ANA represents the total time rate of change of total water.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Global time series of monthly mean water cycle terms (mm day−1) for (left) MERRA and (right) MERRA-2: (top) global precipitation (P, red) and evaporation (E, black) and (bottom) the global E − P increment (blue), the analysis increment (ANA, green), and the E − P + ANA (pink). The MERRA-2 analysis increment is plotted but is near zero by design as described in the text. The E − P + ANA represents the total time rate of change of total water.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Global time series of monthly mean water cycle terms (mm day−1) for (left) MERRA and (right) MERRA-2: (top) global precipitation (P, red) and evaporation (E, black) and (bottom) the global E − P increment (blue), the analysis increment (ANA, green), and the E − P + ANA (pink). The MERRA-2 analysis increment is plotted but is near zero by design as described in the text. The E − P + ANA represents the total time rate of change of total water.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Spatial variance of the global analysis tendency of vertically integrated total water mass (mm day−1) in the atmosphere for MERRA (red) and MERRA-2 (black). This indicates the global variations of water increment are still similar to MERRA, even though the global mean of the increment is constrained.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Spatial variance of the global analysis tendency of vertically integrated total water mass (mm day−1) in the atmosphere for MERRA (red) and MERRA-2 (black). This indicates the global variations of water increment are still similar to MERRA, even though the global mean of the increment is constrained.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Spatial variance of the global analysis tendency of vertically integrated total water mass (mm day−1) in the atmosphere for MERRA (red) and MERRA-2 (black). This indicates the global variations of water increment are still similar to MERRA, even though the global mean of the increment is constrained.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Schröder et al. (2016) evaluate the total column water over global oceans from CFSR, MERRA, and ERA-Interim, identifying discontinuities in all three associated with satellite data changes. The effect of the mass constraint has prevented the apparent spurious temporal discontinuities on the MERRA-2 total water mass. For example, Fig. 3 shows the total column water over the ocean, where MERRA exhibits significant jumps (at 1987 for SSM/I and 1998 for AMSU-A), but MERRA-2 is much closer to the retrieved observations.
Ocean-only (60°S–60°N) 12-month running mean time series of total column water (mm) for MERRA (red), MERRA-2 (black), and RSS (SSM/I retrieved, blue).
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Ocean-only (60°S–60°N) 12-month running mean time series of total column water (mm) for MERRA (red), MERRA-2 (black), and RSS (SSM/I retrieved, blue).
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Ocean-only (60°S–60°N) 12-month running mean time series of total column water (mm) for MERRA (red), MERRA-2 (black), and RSS (SSM/I retrieved, blue).
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
After MERRA, an additional offline “replay” of the land surface component forced with gridded observation-corrected precipitation Pocor and using the near-surface meteorology from MERRA was produced under the name MERRA-Land (Reichle et al. 2011). For MERRA-2, this process was conducted “in line” using observation-corrected precipitation.2 As MERRA-2 was integrated forward in time, observation-corrected precipitation was used in the land forcing instead of model-generated precipitation Pmgen produced during the assimilation cycle. In principle, the observed precipitation corrections should lead to a better land surface hydrology than relying on model-generated precipitation. This has implications for land surface and atmospheric water balances—the vertically integrated moisture flux convergence is consistent with E − Pmgen, while E − Pocor is consistent with the land surface water budget but not the atmospheric water budget. The performance of the MERRA-2 land system in the context of land surface hydrology and GRACE observations will be reported separately (Reichle et al. 2016, unpublished manuscript); here we focus primarily on the atmospheric water balance except where specifically noted by calling out the observation-corrected precipitation Pocor.
In some comparisons, we will include a pure model experiment, based on one realization of the MERRA-2 model. This is the identical model used for MERRA-2, including the prescribed sea surface temperatures, and it will be denoted MERRA-2 AMIP (M2AMIP). The model configuration (grid and output diagnostics) is the same as MERRA-2. The GEOS-5 model changes from MERRA to MERRA-2 are documented by Molod et al. (2015).
4. Water budget climatology
a. Global water balance
In MERRA-2, the depiction of the global mean water cycle reduces to the balance of evaporation and precipitation owing to the mass constraint discussed earlier, similar to that produced from global atmospheric numerical models. Global observations of the water budget tend to have imbalances due to many factors and uncertainties—among them, heterogeneous sampling, changes in instrumentation and sensors, and time-varying calibration. Fundamental evaluation of the water cycle separates continental and oceanic domains and includes the water transport between them.
Recently, Rodell et al. (2015) used a wide variety of NASA and satellite observations and some models to develop a closed water and energy cycle dataset including consideration of the data uncertainties. One of the useful features of the MERRA-2 output diagnostics is that the tendency terms from the state budgets are written directly from the model, including vertical integrals of the model level fields. This allows balanced budgets for regions and domains to be diagnosed. Table 1 compares the NASA Energy and Water Cycle Study observational data with MERRA-2’s water cycle. While the merged observation data are provided for reference, we focus on the balanced version of the data. For the global ocean average, MERRA-2 overestimates the surface evaporation, with precipitation very close to observed. The excess ocean evaporation is compensated partly by the transport to land and also by removal due to the analysis increment. Over the land, there is far too much precipitation generated by the model. This is generally related to high-topography tropical land areas such as the Andes mountain range (Bosilovich et al. 2015b; Fig. 4). A broad area of excess precipitation across the tropical land is related to a global land positive analysis increment. Even so, the local increment tendencies are smaller than either evaporation or precipitation.
Depiction of the global water cycle developed by NEWS (Rodell et al. 2015) from NASA observing systems for 2000–10, including the collected and merged data and a balance-corrected budget, compared to MERRA-2 data (103 km3 yr−1). Uncertainty estimates were given by data providers and refined in the balanced budget. MERRA-2’s analysis increments are included in the data and are necessary to balance the water cycle. A side effect of the water vapor mass constraint is that in these complementary areas, the increment will be equal and opposite, as with water vapor transport. Numbers in parentheses indicate the NEWS reported uncertainty, and in brackets indicate the MERRA-2 observation-corrected precipitation.
Global map of 35-yr (1980–2014) mean differences between MERRA-2 and GPCP precipitation (mm day−1): (top) June–August and (bottom) December–February.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Global map of 35-yr (1980–2014) mean differences between MERRA-2 and GPCP precipitation (mm day−1): (top) June–August and (bottom) December–February.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Global map of 35-yr (1980–2014) mean differences between MERRA-2 and GPCP precipitation (mm day−1): (top) June–August and (bottom) December–February.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
The land evaporation remains close to the observations, owing in part to the use of observed precipitation for land forcing (Reichle et al. 2016, unpublished manuscript; Bosilovich et al. 2015b). The global land observation-corrected precipitation Pocor forcing in this time period is 114.1 × 103 km3 yr−1. Over the ocean, the analysis of observations is working to slow the water cycle. The general sense of this comparison is that MERRA-2 has a stronger water cycle than the observations, owing to the ocean evaporation and atmospheric transport that comes about from changes to the MERRA-2 model.
Figure 5 shows the global climatological maps of all the water budget terms. The mass constraint, discussed earlier, allows the global average of the E − P to be essentially zero; nevertheless, the analysis increment itself does have significant regional spatial pattern structure. The modeling system produces excessive precipitation in the Indo-Pacific warm pool and surrounding land regions (made apparent in Fig. 4) where the analysis is adding water to the atmosphere. The opposite occurs in the eastern tropical Pacific convergence zone and in much of the western Indian Ocean, where the precipitation is near the observed and the analysis increment is drying the atmosphere. To a significant degree these features amount to the moisture analysis increment forcing an amplification of the Walker circulation. Some large areas of very high moisture convergence occur over topography and island–coastal regions, also connected with large modeled precipitation amounts noted in Fig. 4. This is an issue in the modeling system that will be addressed in future versions.
Global maps of the components of the MERRA-2 vertically integrated water cycle (mm day−1), time averaged over 1980 through 2014, using the atmospheric model’s variable notation. Including (top left) precipitation, (bottom left) evaporation, (top right) dynamics tendencies (primarily convergence), and (bottom right) water vapor analysis increment. The total time tendency is very small in these conditions and has been omitted.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Global maps of the components of the MERRA-2 vertically integrated water cycle (mm day−1), time averaged over 1980 through 2014, using the atmospheric model’s variable notation. Including (top left) precipitation, (bottom left) evaporation, (top right) dynamics tendencies (primarily convergence), and (bottom right) water vapor analysis increment. The total time tendency is very small in these conditions and has been omitted.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Global maps of the components of the MERRA-2 vertically integrated water cycle (mm day−1), time averaged over 1980 through 2014, using the atmospheric model’s variable notation. Including (top left) precipitation, (bottom left) evaporation, (top right) dynamics tendencies (primarily convergence), and (bottom right) water vapor analysis increment. The total time tendency is very small in these conditions and has been omitted.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
b. Temporal variability
Reanalysis precipitation and evaporation are computed from the background forecast model integration in the analysis cycle and not directly analyzed from observations. Figure 6 shows the global average time series of precipitation and evaporation from MERRA-2 and a multitude of contemporary reanalyses, both satellite data reanalyses and reduced-observation reanalyses. GPCP merged satellite–gauge observation data are also included. Broadly speaking, the satellite data reanalyses have apparent interannual variations based on changes in the observing system, but also ENSO variability is present (e.g., the precipitation and evaporation peaks in the late 1990s and 2010). MERRA and CFSR precipitation, for example, increase dramatically at the introduction of ATOVS (July 1998), related to the assimilation of AMSU-A window channels. ERA-Interim, likewise, shows sensitivity to the loss of SSM/I platforms starting in the late 2000s. The reduced-observation reanalyses (20CR and ERA-20C) are useful in this context because their precipitation shows variations comparable to the GPCP observations. For MERRA-2, the precipitation variations are generally more stable than those produced in MERRA; however, the MERRA-2 significant reduction and recovery of precipitation in the early 1980s is not detected in observed precipitation datasets. The recovery of precipitation in 1988 is closely related to the time that SSM/I assimilation begins. The curves in Fig. 6 show that MERRA-2’s global precipitation and evaporation are tightly coupled, largely through the mass constraint on global water vapor increments. Note that since 2000, with the enormous increase in the number and quality of satellite observations, the MERRA-2 precipitation variability is close to that of GPCP (i.e., no major jumps in the data, despite losing SSM/I and numerous changes in the ATOVS platforms, and the addition of IASI and GPSRO).
Global average monthly mean time series of (left) precipitation and (right) evaporation for a multitude of contemporary reanalyses (mm day−1): MERRA-2 (black), MERRA (red), ERA-Interim (dark blue), JRA-55 (green), CFSR (pink), 20CR (orange), ERA-20C (light blue), and GPCP (purple). The annual cycle has been removed with a 12-month running mean.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Global average monthly mean time series of (left) precipitation and (right) evaporation for a multitude of contemporary reanalyses (mm day−1): MERRA-2 (black), MERRA (red), ERA-Interim (dark blue), JRA-55 (green), CFSR (pink), 20CR (orange), ERA-20C (light blue), and GPCP (purple). The annual cycle has been removed with a 12-month running mean.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Global average monthly mean time series of (left) precipitation and (right) evaporation for a multitude of contemporary reanalyses (mm day−1): MERRA-2 (black), MERRA (red), ERA-Interim (dark blue), JRA-55 (green), CFSR (pink), 20CR (orange), ERA-20C (light blue), and GPCP (purple). The annual cycle has been removed with a 12-month running mean.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Complementing Table 1, time series of annual mean water budget terms for ocean and land are separated for individual years of MERRA-2 in Fig. 7. The continental precipitation (this is the model-generated precipitation) in MERRA-2 is increasing over time generally following the analysis increment. The continental evaporation is insulated from this effect owing to the use of observation-corrected precipitation forcing for the land surface. The downward trend in evaporation is consistent with that of the precipitation from CPCU used in forcing the land component. Over the ocean, the covariability of precipitation and evaporation is apparent, where the variations in the precipitation are similar to the variations in evaporation (which is being governed by the prescribed sea surface temperature and the assimilation of near-surface winds and water vapor). The mean analysis increment (climatological total plus anomaly) always dries the analysis, working against a wet model bias (Molod et al. 2012), and the drying intensifies with time while the evaporation increases. But a remarkable feature here is that the moisture transport from ocean to the continents is not affected by the regional changes in the analysis increments or the increasing ocean evaporation, exhibiting little trend over the 35 yr. This is different than most other reanalyses that show noticeable trends of the moisture transport (e.g., Robertson et al. 2014, their Figs. 5 and 7).
Time series of MERRA-2’s annual water cycle anomalies (from the 1980–2014 mean) area averaged for (top) land and (bottom) ocean. The water cycle components depicted are evaporation E (red), modeled precipitation P (blue), total water vapor analysis increment (dwdtANA, green), and transport from ocean to land (dwdtDYN, derived from moisture flux convergence, purple). In addition, an experiment withholding AIRS radiances was executed to test its sensitivity and is depicted here by the three years with colored circles from 2003 to 2005. The terms and units follow Table 1 (103 km3 yr−1).
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Time series of MERRA-2’s annual water cycle anomalies (from the 1980–2014 mean) area averaged for (top) land and (bottom) ocean. The water cycle components depicted are evaporation E (red), modeled precipitation P (blue), total water vapor analysis increment (dwdtANA, green), and transport from ocean to land (dwdtDYN, derived from moisture flux convergence, purple). In addition, an experiment withholding AIRS radiances was executed to test its sensitivity and is depicted here by the three years with colored circles from 2003 to 2005. The terms and units follow Table 1 (103 km3 yr−1).
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Time series of MERRA-2’s annual water cycle anomalies (from the 1980–2014 mean) area averaged for (top) land and (bottom) ocean. The water cycle components depicted are evaporation E (red), modeled precipitation P (blue), total water vapor analysis increment (dwdtANA, green), and transport from ocean to land (dwdtDYN, derived from moisture flux convergence, purple). In addition, an experiment withholding AIRS radiances was executed to test its sensitivity and is depicted here by the three years with colored circles from 2003 to 2005. The terms and units follow Table 1 (103 km3 yr−1).
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
c. Ocean–land transport
Flux convergence and divergence of atmospheric moisture in reanalyses are typically found to be more accurate in approximating P − E than the reanalysis diagnostics of P and E. This is because the flux convergence involves analyzed state variables of wind and moisture. But there are vast data-poor regions over the tropical continents (especially central Africa) that are of critical importance. The lack of direct wind and moisture observations in these regions, especially in the lower troposphere, has critical bearing on the veracity of the moisture budget and variations across many scales. Here we focus the discussion on the MERRA-2 moisture transports and comparisons with related datasets.
As noted in section 3, for monthly and longer time scales the vertically integrated moisture flux convergence (dqvdt_dyn is the MERRA-2 variable used for VMFC) is equivalent to Pmgen − E + ANA, where ANA represents the analysis increment. Monthly climatological values area averaged over land (60°N–60°S) for MERRA-2 Pocor, Pmgen, E, P − E, and VMFC are shown in Fig. 8. Here Pmgen is the precipitation generated by the model physics during the assimilation process. The term Pocor is the observation-corrected precipitation discussed in section 3 that is used to drive the MERRA-2 land surface water and energy model during the assimilation. We compare these to the earlier MERRA and to ensemble P − E values of the collection of LSMs that, similarly to MERRA-2, are driven by gridded observed precipitation and near-surface meteorology. VMFC and the various P − E estimates have similar annual cycles peaking in boreal winter with minima in May–June (Fig. 8). This is despite the fact that P and E maximize in boreal summer. But the E maxima from MERRA and MERRA-2 peak about a month before P and have a more pronounced annual variation.
Mean annual cycle of various atmospheric water cycle components (mm day−1) area averaged over global land areas (60°N–60°S): MERRA-2 Pmgen (black) and Pocor (black dashed), VMFC (red), E (blue), Pocor − E (cyan); MERRA-Land P (blue dashed) and P − E (cyan dashed); and LSM P − E (green).
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Mean annual cycle of various atmospheric water cycle components (mm day−1) area averaged over global land areas (60°N–60°S): MERRA-2 Pmgen (black) and Pocor (black dashed), VMFC (red), E (blue), Pocor − E (cyan); MERRA-Land P (blue dashed) and P − E (cyan dashed); and LSM P − E (green).
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Mean annual cycle of various atmospheric water cycle components (mm day−1) area averaged over global land areas (60°N–60°S): MERRA-2 Pmgen (black) and Pocor (black dashed), VMFC (red), E (blue), Pocor − E (cyan); MERRA-Land P (blue dashed) and P − E (cyan dashed); and LSM P − E (green).
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
VMFC is also larger than any of the P − E estimates (Fig. 8), a result of the fact that this transport of moisture is coupled to the diabatic heating from the net precipitation produced by the assimilation, not the diabatic heating that is implied by the observed precipitation used to drive the land model. Since Pmgen is almost 1 mm day−1 greater than Pocor this significantly contributes to the magnitude of VMFC. But MERRA and MERRA-2 evaporation also have mean values of nearly 1.8 mm day−1, substantially larger than most estimates derived by LandFlux (Jiménez et al. 2011), which produces best estimates around 1.4 mm day−1, and the MERRA-2 estimates are larger than the NEWS merged observation estimates (Table 1). When paired with the excessively large Pmgen value this excess E acts to moderate the Pmgen − E values.
The LSM P − E values (an indirect estimate of VMFC) lie between the MERRA-2 VMFC and P − E values. The latter are depressed by what we judge to be excessive MERRA-2 E values. From a climatological perspective MERRA-2 thus has demonstrable climatological moisture transport biases, as noted in section 4a.
5. Effect of changing observations
a. AIRS data withholding
One of the motivations for applying the mass constraint described in section 3 was to limit the global mass imbalance of water vapor that occurs in reanalyses if left unchecked. To quantify the effect of a major observing system change, we performed a sensitivity experiment that withheld the AIRS instrument at the beginning of its availability (November 2002; every other option in the system remains identical to the MERRA-2 configuration) and continued that experiment through the end of 2005. The results of that experiment are also included in Fig. 7 as the dotted lines for 2003–05. In the continental water cycle, the immediate impact of AIRS on the water vapor increment is roughly a 50% increase, after which the increments remain at this new higher level. This is almost entirely realized as an increase in model-generated precipitation over land, as the transport changes only slightly and land evaporation does not depend on model-generated precipitation. The analysis increment becomes increasingly negative (increased drying) over the complementary ocean area as AIRS is introduced, which leads to a reduction of precipitation and a slight reduction in evaporation.
The spatial distribution of the change in the water cycle by AIRS is presented in Fig. 9. The mean land increments are generally positive (adding water) with the largest values across the Sahel and South America, while oceanic decreases in the water vapor increments are spread throughout the tropics (30°S–30°N band). The overall change to surface evaporation is an order of magnitude less than the change in any of the other water cycle components. The land precipitation increases with AIRS are distributed much like the analysis increments, while the ocean areas have small regions of positive and negative values, indicating some redistribution of the precipitation embedded in the net reduction of ocean precipitation. Contrary to the other terms, the atmospheric convergence of water vapor does not have a clear land–ocean signal. Even with the constraint on global water vapor increments, significant regional changes of the water cycle terms can occur with the introduction or change of instruments sensitive to water vapor.
Time mean (Jan 2003–Dec 2005) difference of the water cycle fields (mm day−1) (top left) precipitation, (top right) convergence, (bottom left) evaporation, and (bottom right) water vapor analysis increment, between MERRA-2 and an experiment that withholds all AIRS instrument observations; note that the evaporation differences have a smaller range than other terms.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Time mean (Jan 2003–Dec 2005) difference of the water cycle fields (mm day−1) (top left) precipitation, (top right) convergence, (bottom left) evaporation, and (bottom right) water vapor analysis increment, between MERRA-2 and an experiment that withholds all AIRS instrument observations; note that the evaporation differences have a smaller range than other terms.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Time mean (Jan 2003–Dec 2005) difference of the water cycle fields (mm day−1) (top left) precipitation, (top right) convergence, (bottom left) evaporation, and (bottom right) water vapor analysis increment, between MERRA-2 and an experiment that withholds all AIRS instrument observations; note that the evaporation differences have a smaller range than other terms.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
b. Water vapor increments
The previous section discussed some of the major characteristics of the MERRA-2 water budget and the influence of observations. Here we identify more specifically how some of the major sensor changes manifest their effects.
1) Temporal variability
Examination of vertically integrated fields, as were shown in Fig. 7, minimizes some important vertical structure through compensating errors, and so while it helps in comparisons with precipitation and evaporation, there are more complex structures embedded within the column. While the global vertically integrated moisture increments tend to be small, at any given vertical layer the contribution of the water vapor increments may not be of the same sign. Figure 10 shows a time–height cross section of the analysis increment area averaged over the global oceans and land (60°N–60°S). For the ocean, substantial vertical structure is present with persistent analysis moistening in the lower troposphere maximizing in the 850–900-hPa layer and drying in the troposphere maximizing near 500 hPa. Yet there is also temporal variability with the moistening/drying structure becoming much more pronounced after 2000. These changes are related to the dominant effects of AMSU and AIRS data. In the lowest 50 hPa near the surface, a sharp decrease in moistening around 1988 (when one SSM/I instrument is available), eventually to drying in 1992 (when multiple SSM/I instruments are assimilated), and then back to moistening in 2010 (when SSM/I is no longer assimilated) underscore strong contributions from the SSM/I sensors. For the land, a significant annual cycle variation is apparent, maximizing in the lowest 50 hPa. The deep moistening continues through the 550-hPa layer, with some drying just above that. The atmospheric moistening is consistent with the precipitation excess; that is, the analysis increments add water and the model removes it by precipitating. As with the ocean there is a significant change in the increments after 2000, but there are subtle differences between land and ocean. Over ocean the upper-level drying begins in the late 1990s, coincident with the initial use of AMSU-B channels sensitive to mid- and upper-tropospheric moisture. The GEOS-5 model used in the assimilation has a mid- and upper-tropospheric moist bias over ocean that the AMSU-B and AIRS are more effective in correcting than the earlier HIRS sensors. Over land, as noted earlier, precipitation from convection is too strong, which tends to dry the lower troposphere. In the ideal situation these increments would be randomly distributed, indicating that the model physics are unbiased and that data assimilation is not affecting the pdf of the moisture distribution in a global sense. This is obviously not true for the GEOS-5 model, nor is it true for any model used for reanalyses.
Time series of the area average vertical profile of monthly mean water vapor analysis increment (mm day−1) for (top) land and (bottom) ocean area.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Time series of the area average vertical profile of monthly mean water vapor analysis increment (mm day−1) for (top) land and (bottom) ocean area.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Time series of the area average vertical profile of monthly mean water vapor analysis increment (mm day−1) for (top) land and (bottom) ocean area.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
2) Spatial variability
To gain more insight into the spatial structure of the increments as a function of height a principal component analysis was performed on the water vapor increments at three different levels, 975, 850, and 500 hPa, focusing on ocean areas. The leading EOFs’ PCs and their time series are shown in Fig. 11 and Fig. 12, respectively. At 975 hPa the water vapor increment leading mode (EOF975 × PC975) explains 52% of the variance and shows a pronounced drying tendency near the surface produced by the SSM/I sensors from 1988 through 2009. In fact, the discrete changes induced by individual SSM/I sensors on board DSMP F8 through F15 platforms are prominent steplike increases in drying as the number of SSM/I sensors increases. As DMSP F13, F14, and F15 platforms age away in the 2000s, this drying tendency rapidly weakens. The leading EOF pattern at 975 hPa indicates drying focused in the equatorward and westward reaches of subtropical ridges. Interestingly these regions also are typified by excessive cloudiness and precipitating convection compared to observations (not shown).
Leading EOFs from a principal component analysis of MERRA-2 water vapor increment at various vertical pressure levels: (top) 500, (middle) 850, and (bottom) 975 hPa. Monthly climatological values have been removed from the time series before the analysis and are indicated on the figure. The PCs are normalized so that their product with the respective EOFs yields the contribution to the moisture increment anomalies.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Leading EOFs from a principal component analysis of MERRA-2 water vapor increment at various vertical pressure levels: (top) 500, (middle) 850, and (bottom) 975 hPa. Monthly climatological values have been removed from the time series before the analysis and are indicated on the figure. The PCs are normalized so that their product with the respective EOFs yields the contribution to the moisture increment anomalies.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Leading EOFs from a principal component analysis of MERRA-2 water vapor increment at various vertical pressure levels: (top) 500, (middle) 850, and (bottom) 975 hPa. Monthly climatological values have been removed from the time series before the analysis and are indicated on the figure. The PCs are normalized so that their product with the respective EOFs yields the contribution to the moisture increment anomalies.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
The time series of the normalized PCs of the leading EOFs presented in Fig. 11 (mm day−1): 975 (solid), 850 (dashed), and 500 hPa (dotted). Also shown for comparison are the periods of tenure for SSM/I sensors on board F8 through F15 (cyan), AMSU-B (green), and AIRS281 (purple) (excerpt from McCarty et al. 2016).
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
The time series of the normalized PCs of the leading EOFs presented in Fig. 11 (mm day−1): 975 (solid), 850 (dashed), and 500 hPa (dotted). Also shown for comparison are the periods of tenure for SSM/I sensors on board F8 through F15 (cyan), AMSU-B (green), and AIRS281 (purple) (excerpt from McCarty et al. 2016).
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
The time series of the normalized PCs of the leading EOFs presented in Fig. 11 (mm day−1): 975 (solid), 850 (dashed), and 500 hPa (dotted). Also shown for comparison are the periods of tenure for SSM/I sensors on board F8 through F15 (cyan), AMSU-B (green), and AIRS281 (purple) (excerpt from McCarty et al. 2016).
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
In contrast, at 850 hPa the dominant signature is one of moistening. The leading EOF explains less of the total variance (33%) but shows that this moistening is focused over the Indo-Pacific warm pool and Atlantic ITCZ. The PC structure indicates a dominant contribution from SSM/I sensors, at least initially. After 2000 the anticorrelation between the PCs at 850 and 975 hPa weakens with continued moistening at 850 hPa while at 975 hPa the drying has stabilized. This suggests the influence of AMSU after late 1998 and later AIRS sensors after 2002. A marked weakening of 850-hPa moistening again occurs with the loss of SSM/I sensors near 2010.
The leading mode at 500 hPa explains 32% of the variance with an EOF pattern that is very similar to that at 850 hPa but opposite in sign. The PC shows a strong correlation to the advent of AMSU-B moisture sounders whose channel response functions typically peak at mid- to upper-tropospheric levels. The direct effect of AIRS itself is best seen here as an additional jump in the PC value in 2002 indicating an increased drying.
From Table 2, which contains the percent variance explained by the first five modes at each level, it is clear that one mode at 975 hPa and at most three modes at the other levels clearly delineate the impact of the sensors. Mode-2 and mode-3 PCs and EOFs at the upper levels (not shown) serve largely to modify the seasonal cycle effects and regional details of the patterns.
6. Physical implications of data analysis
The discussion in the preceding section highlights the significant impact of changing data streams on MERRA-2 moisture budget components. Here we examine the impact on the variability of key processes that characterize interannual climate variability and trends depicted by MERRA-2.
a. Decomposition of evaporation processes
Figure 13 shows the resulting time series of evaporation and the contributions by each factor area averaged over the global oceans in three different latitude bands: 30°–60°N, 30°–60°S, and 30°N–30°S. Processes controlling evaporation changes in the midlatitude bands differ strongly from those in the tropics. Stability changes assume a prominent role in the NH extratropics. Here the average climatological SST − Tair is 1.0 K. The stability term indicates that even though SST is increasing over time SST − Tair is increasing so Tair (air temperature near the surface) is not keeping up with SST increases. The Tair anomalies are warm relative to those of SST early on but after 2000 that relationship reverses. Variations in near-surface air specific humidity Qa must be occurring at near-constant RH given the latter’s small excursions relative to SST. Since SSM/I does not directly affect Tair, it seems likely that the change from NOAA-10 to NOAA-12 temperature sounders might explain the decrease in Tair in 1991. Stronger Tair a decreases relative to SST are seen in 1998 and 2000 when ATOVS replaces TOVS. In the Southern Hemisphere midlatitudes both wind and stability changes drive evaporation increases. Winds increase after the start of SSM/I F8 data in 1988 but are rather stable after that. For the Southern Hemisphere, the climatological SST − Tair is ~0.7 K and the stability contribution to evaporation jumps in 1992 as in the Northern Hemisphere, but it is not clear why the jump in evaporation possibly associated with ATOVS in the NH is not seen in the SH. The RH changes show no significant trend indicating that Qa behaves similarly as in the Northern Hemisphere extratropics.
Decomposition of the MERRA-2 ocean surface evaporation (mm day−1) into contributing components of SST (red), wind (blue), stability and exchange coefficient (yellow), and specific humidity (green). The evaporation is in black and the residual is in purple for the latitude bands of (a) 30°–60°N, (b) 30°S–30°N, and (c) 30°–60°S. The stability and exchange coefficient term has been included. A 12-month smoothing has been applied.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Decomposition of the MERRA-2 ocean surface evaporation (mm day−1) into contributing components of SST (red), wind (blue), stability and exchange coefficient (yellow), and specific humidity (green). The evaporation is in black and the residual is in purple for the latitude bands of (a) 30°–60°N, (b) 30°S–30°N, and (c) 30°–60°S. The stability and exchange coefficient term has been included. A 12-month smoothing has been applied.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Decomposition of the MERRA-2 ocean surface evaporation (mm day−1) into contributing components of SST (red), wind (blue), stability and exchange coefficient (yellow), and specific humidity (green). The evaporation is in black and the residual is in purple for the latitude bands of (a) 30°–60°N, (b) 30°S–30°N, and (c) 30°–60°S. The stability and exchange coefficient term has been included. A 12-month smoothing has been applied.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Mechanisms governing the tropical band differ substantially from those in midlatitudes. Wind increases are a significant factor, accounting for an approximately 0.25 mm day−1 rise of the evaporation rate in late 1987. This jump accompanies the onset of SSM/I data availability and the assimilation of SSM/I wind speeds. Interestingly, stability and RH changes do not contribute substantially to any trend over the period of study but rather tend to offset each other in affecting evaporation—when Tair is high relative to SST, RH contribution is lower, and vice versa. One possible explanation for this is that as E increases (initially as a result of stronger winds), so does P driven through the parameterized convection. The ensuing decrease in Tair and subsidence drying are the net effects of convective stabilization of the subcloud layer.
Aside from these variations due to changing observing systems there are important physically based changes in the MERRA-2 evaporation. ENSO-related variations in the tropical band (Fig. 13b) are among the most obvious. SST-induced contributions to evaporation increase are seen for prominent El Niño warm events such as those in late 1997 and 2009. Smaller signals are apparent in the mid-2000s. Note that the wind speed contribution to variations in evaporation initially falls with warm events and then peak after the SST peak. Wind speed and RH contributions are anticorrelated on these interannual time scales. Cold La Niña events following these warm periods provide temporally localized evaporation reductions.
To summarize, our analysis of the mechanisms governing ocean evaporation in MERRA-2 shows that, as with other reanalyses, considerable uncertainty in time series of near-surface bulk variables arises from continual but discrete upgrades in the ability of observing systems to counter GEOS-5 assimilation model physical biases.
b. Surface energy budget and links to moisture changes
In light of the evaporation uncertainties, it is expected that these effects might propagate through the energy budget, which is tightly coupled to moisture. Here we focus briefly on surface energy flux changes and their links to the atmospheric moisture budget. Surface flux anomaly components averaged separately over land and ocean are shown in Fig. 14. Over land sensible heat and latent heat are anticorrelated, a signature that soil moisture is controlling the partitioning of surface net radiative input. Likewise, the net longwave and net shortwave radiations are anticorrelated, suggesting that when cloudiness decreases (increases) the increased (decreased) shortwave absorption is countered (augmented) by less (more) downward longwave radiation from clouds. The climatological net surface flux is small (0.6 W m−2) as are the annual anomalies in Fig. 14.
MERRA-2 annual surface energy balance terms (W m−2)—latent heat flux LE (red), sensible heat flux Hs (blue), net shortwave radiation SWn (green), net longwave radiation LWn (purple), and the net value (black)—averaged globally for (a) oceans and (b) continents. The sign convention is positive down to the surface.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
MERRA-2 annual surface energy balance terms (W m−2)—latent heat flux LE (red), sensible heat flux Hs (blue), net shortwave radiation SWn (green), net longwave radiation LWn (purple), and the net value (black)—averaged globally for (a) oceans and (b) continents. The sign convention is positive down to the surface.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
MERRA-2 annual surface energy balance terms (W m−2)—latent heat flux LE (red), sensible heat flux Hs (blue), net shortwave radiation SWn (green), net longwave radiation LWn (purple), and the net value (black)—averaged globally for (a) oceans and (b) continents. The sign convention is positive down to the surface.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Over ocean (Fig. 14) the picture is quite different, with an increasing trend of ocean energy loss to the atmosphere of −0.46 W m−2 yr−1. Evaporation increases drive this loss but are strongly coupled to surface net flux decreases. Figures 15a,c show that globally, evaporation and low-cloud increases are tightly correlated, especially in regions poleward of 30° latitude. In turn, both are also closely related to surface net flux decreases, especially on longer than interannual time scales. Note that the low-cloud increases also track the stability contributions to evaporation in Figs. 15a,c. We speculate that lower-tropospheric temperature changes induced by changes in temperature sounders (TOVS in 1992 and the change from TOVS to ATOVS in late 1998) provided enough SST − Tair decrease to elevate evaporation (at roughly constant RH) rates. In turn, these evaporation increases have elevated low-cloud fractions and reduced surface net radiation. Over the tropics parameterized convection plays a much larger role in links between moisture and radiative fluxes. The relaxed Arakawa–Schubert (RAS) convective closure in the GEOS-5 model essentially keys on the production of convective available potential energy (CAPE) by radiative, dynamical, and turbulent processes. Stabilization of the boundary layer proceeds in a quasi-equilibrium sense owing to CAPE consumption by deep convection. This constraint likely has such a strong effect on SST − Tair that assimilation of temperature data has less an effect on shallow cloudiness changes in the convecting regions of the tropics than in midlatitudes. Confirmation of these mechanisms and their roles is the subject of follow-on work.
MERRA-2 surface net radiation SWn (black; W m−2), surface latent heat flux LE (red; W m−2), and low-cloud (from surface to 700 hPa) fraction Clow (blue; %) for the latitude bands of (a) 30°–60°N, (b) 30°S–30°N, and (c) 30°–60°S. A 12-month smoothing has been applied.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
MERRA-2 surface net radiation SWn (black; W m−2), surface latent heat flux LE (red; W m−2), and low-cloud (from surface to 700 hPa) fraction Clow (blue; %) for the latitude bands of (a) 30°–60°N, (b) 30°S–30°N, and (c) 30°–60°S. A 12-month smoothing has been applied.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
MERRA-2 surface net radiation SWn (black; W m−2), surface latent heat flux LE (red; W m−2), and low-cloud (from surface to 700 hPa) fraction Clow (blue; %) for the latitude bands of (a) 30°–60°N, (b) 30°S–30°N, and (c) 30°–60°S. A 12-month smoothing has been applied.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
c. Temperature–water vapor relationship
Thus far, the MERRA-2 water cycle has been shown to be slightly stronger than previous estimates in terms of the high biased transport from ocean to land, owing to 1) larger than observed oceanic evaporation that is also increasing over the satellite era and 2) the satellite observing systems’ influence on the water cycle, though not introducing discontinuities in total column water (Fig. 3) or moisture transport (Fig. 7) as strong as those noted in other reanalyses. Yet prescribed SSTs and ocean evaporation are increasing in time. This raises a question about the representation of the temperature–water vapor relationship in reanalyses. Sun and Held (1996) tested the relationship between temperature and water vapor in GCMs by first separating the seasonal cycle and long-term trend from the interannual variability, demonstrating that the change of water vapor with temperature is stronger in atmospheric models than radiosonde observations and also stronger close to the surface than in the midtroposphere. Mears and Wentz (2009) investigated the standard deviation (in time) of the change of water vapor with temperature relationship using a similar metric, with retrieved ocean total column water (based on SSM/I) and the weighted lower-tropospheric brightness temperature. However, the MERRA-2 water vapor analysis includes information from many observed sources. Here, we test the water vapor–temperature relationship in several reanalyses,3 MERRA-2, and AMIP simulations.
Figure 16 compares RSS-retrieved TPW with the MSU–AMSU retrieved TLT in the range of 30°S–30°N (ocean only). To normalize the moisture term, we divide by the 1988–2014 global mean, to account for the climatological spatial variations of water vapor in the variability of the water vapor. Following Sun and Held (1996), linear regression of the normalized water vapor anomaly time series and the temperature anomaly time series determines the percent change of water vapor per Kelvin. In this case with the full period trend included, the RSS TPW increases at 5.4% K−1. For MERRA-2, the trend in TLT is comparable to the RSS value; however, the MERRA-2 TPW trend is much smaller, so that water vapor is increasing at 2.9% K−1. Results from the MERRA-2 AGCM model in an AMIP experiment, using the same SSTs at the boundary of MERRA-2, show that the model produces stronger trends in atmospheric temperature and moisture, and the relationship is computed at 5.0% K−1 (closer to the RSS-observed values than MERRA-2). Additional statistics for these experiments are presented in Table 3 and show that the MERRA-2 AMIP simulation has a stronger correlation between temperature and moisture than the RSS observations [an expected result considering Sun and Held (1996) and Mears and Wentz (2009)], and the MERRA-2 correlation is somewhat weaker than observed. Considering the strong model relationship of temperature and moisture, and also the reproduction of the observed trend in MERRA-2 temperature, these suggest that the water vapor assimilation is leading to the weak relationship in MERRA-2. Note that even the detrended relationship, which emphasizes ENSO interannual variability, is also weak in MERRA-2 (Table 3).
Time series of TLT tropical ocean (30°S–30°N) average anomalies (K; black) and normalized total column water anomalies (Q, percent of climate mean; red), deseasonalized with a 12-month running mean, for (a) RSS retrievals (Mears and Wentz 2009; Remote Sensing Systems 2016), (b) MERRA-2, and (c) a MERRA-2 AMIP-type present-day simulation with prescribed sea surface temperature. Linearly regressing the anomalies to determine the relationship between water and temperature in the atmosphere yields 5.4%, 2.9%, and 5.0% K−1, respectively. Table 3 provides additional statistics for these time series.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Time series of TLT tropical ocean (30°S–30°N) average anomalies (K; black) and normalized total column water anomalies (Q, percent of climate mean; red), deseasonalized with a 12-month running mean, for (a) RSS retrievals (Mears and Wentz 2009; Remote Sensing Systems 2016), (b) MERRA-2, and (c) a MERRA-2 AMIP-type present-day simulation with prescribed sea surface temperature. Linearly regressing the anomalies to determine the relationship between water and temperature in the atmosphere yields 5.4%, 2.9%, and 5.0% K−1, respectively. Table 3 provides additional statistics for these time series.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Time series of TLT tropical ocean (30°S–30°N) average anomalies (K; black) and normalized total column water anomalies (Q, percent of climate mean; red), deseasonalized with a 12-month running mean, for (a) RSS retrievals (Mears and Wentz 2009; Remote Sensing Systems 2016), (b) MERRA-2, and (c) a MERRA-2 AMIP-type present-day simulation with prescribed sea surface temperature. Linearly regressing the anomalies to determine the relationship between water and temperature in the atmosphere yields 5.4%, 2.9%, and 5.0% K−1, respectively. Table 3 provides additional statistics for these time series.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Calculations characterizing the temperature–water vapor relationship within various systems over the period of 1988–2014 area averaged for the tropical (30°S–30°N) oceans. The first two columns indicate the variable being compared. TPW is the total precipitable water (anomalies of the normalized time series), TLT is the temperature of the lower troposphere, and T850 is the 850-hPa temperature (anomalies). The trends over the period are included for dT/dt and dQ/dt (K decade−1 and % decade−1, respectively). The relationship of temperature and water (in percent change per kelvin) is computed by linear regression between the two time series, with and without trend (as in Sun and Held 1996), and the ratio of standard deviations S (as in Mears and Wentz 2009). Correlation between the two monthly mean time series is also provided.
Figure 17 compares the anomaly time series of MERRA-2 ocean E − P, water vapor increment, and TPW. The value E − P is generally increasing over the period, as the ocean surface temperatures increase. However, the vertically integrated water vapor increments are negative early in the period and generally decrease in time, becoming more negative. The steady increase in vertically integrated moisture reduction by the analysis increment is driven first by SSM/I at low levels and further by midtropospheric drying accompanying AMSU and AIRS (Fig. 10a). Before SSM/I, the analysis is adding lower-tropospheric water over the ocean, which leads to the apparent wet bias when SSM/I does eventually begin (Fig. 3). This indicates that in a column-integrated sense the analysis is working against a moist bias and the increasing surface evaporation that result in a relatively muted change in total column water.
Time series of 60°S–60°N ocean-only MERRA-2 E − P (mm day−1; black), ANA (mm day−1; blue), and TPW (cm; red), deseasonalized with a 12-month running mean.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Time series of 60°S–60°N ocean-only MERRA-2 E − P (mm day−1; black), ANA (mm day−1; blue), and TPW (cm; red), deseasonalized with a 12-month running mean.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
Time series of 60°S–60°N ocean-only MERRA-2 E − P (mm day−1; black), ANA (mm day−1; blue), and TPW (cm; red), deseasonalized with a 12-month running mean.
Citation: Journal of Climate 30, 4; 10.1175/JCLI-D-16-0338.1
To compare more reanalysis and model systems, we use 850-hPa temperature as a proxy for TLT (TLT is not a standard data product in any reanalysis dataset) in Table 3. Each of the satellite data reanalyses have water vapor trends smaller than observed, leading to small values for the temperature–water vapor relationship, though a wide range in the 850-hPa temperature trends will also play a role. In addition, the reduced-observation reanalyses can produce similar relationships to those in the AMIP models and values of the relationship closer to expected, further suggesting that the water vapor increment, how water vapor is assimilated, and controls by moist physics parameterizations require deeper investigation in reanalyses over climate periods.
7. Summary and conclusions
Given the importance of global water cycle variability in understanding the climate, there remains a need for improved accuracy of atmospheric reanalyses. Development of MERRA-2 data included several measures designed to improve the temporal variability of the climate and water cycle. The constraint on the mass field (Takacs et al. 2015, 2016) has essentially eliminated the global mean analysis increment of water vapor that imposed spurious long-term variability on MERRA’s global precipitation and total column water. Of course, forecast departures still occur, and the regional corrections made to the analysis fields must be considered in the temporal variability. For example, an AIRS data withholding experiment on the MERRA-2 system demonstrated the influence that it has on the water cycle, primarily precipitation.
For MERRA-2, the global water cycle climatology is slightly stronger than what is determined from observations (Rodell et al. 2015), owing to a higher ocean evaporation and water transport from ocean to land. While the time series of global precipitation biases have more realistic variability than MERRA, there is still more room for improvement, as the evaporation now controls much of the global precipitation variability. For the ocean, a diagnostic evaluation of the contribution from three terms of the bulk evaporation indicates that the changing wind-observing system has a significant impact on evaporation, but also noteworthy is the effect of near-surface moisture related to multiple SSM/I sensors. SST imparts a low-frequency forcing, along with the ENSO signal. Model-generated precipitation biases in MERRA-2 can pose a problem for regional studies, especially over tropical land near topographic features. Evaluation of the analysis increments indicates that their distribution and variability is as complex in MERRA-2 as in MERRA. They are not uniform in the vertical, and different levels are sensitive to different satellite instruments. However, we see that the total column water and vertically integrated moisture transport have much less pronounced apparent temporal discontinuities or spurious trends compared to previous reanalyses.
The fundamental relationship between temperature and water vapor has been well known to be stronger in atmospheric general circulation models than in observations. When considering the full period of available RSS total column water observations, the MERRA-2 relationship is found to be weaker than observed, even though the underlying model has a relationship as strong as what is observed. While the ocean evaporation is increasing over time (faster than expected), the analysis increment decreases, contributing to the limited water variability when compared to temperature variability. Further, this seems to be an issue with other contemporary reanalyses, though insufficient information is available to say for certain the water vapor analysis contributes to their reduced temperature–water relationship.
Here we have focused on detecting observing system impacts on MERRA-2. Aspects of synoptic and subseasonal variability (not shown) are less obviously affected by observing system changes. We suspect that short-term climate fluctuations such as Pacific decadal variability (Power et al. 1999; Ogata et al. 2013), the Atlantic multidecadal oscillation (Enfield et al. 2001; Zhang and Delworth 2006), or other modes of natural climate variability are likely sources of significant hydrologic cycle variations over this period. While spatial patterns of these modes may be present, these signals are masked in global land and ocean averages for MERRA-2 because of the substantial influence of changes in satellite sensors on the assimilation. This suggests due caution in the interpretation of any climate variations longer than interannual in extent in MERRA-2 and, most likely, other contemporary reanalyses.
Acknowledgments
This paper was sponsored in large part by the NASA energy and water cycle studies program (NNH13ZDA001N-NEWS). MERRA-2 data are developed by the GMAO with support from the NASA Modeling, Analysis and Prediction program. (Information on and availability of the reanalyses’ data used here can be found online reanalysis.org. Similarly, information on the observation data can be found at climatedataguide.ucar.edu.) Three anonymous reviewers provided valuable input that greatly helped the final version of the manuscript. Discussions with Max Suarez streamlined the presentation of the influence of the contributions to the evaporation variability.
APPENDIX A
Evaporation Decomposition Calculation
The error in our calculation is obtained as the residual in (4). Errors result not only from the truncation at a first-order series expansion but also because of other factors: 1) Archived evaporation is accumulated from each time step in the model and so is an exact monthly quantity. However, our use of monthly data in the nonlinear expressions results in inaccuracies. For the globally averaged diagnostics this is clearly not a problem as evidenced by the small relative size of the residual in Fig. 13. 2) In the evaporation–latent heat flux calculation within MERRA the quantity CE involves wind speed and the bulk Richardson number. We have chosen to divide out the wind speed as a separate term from CE and to subsequently display the remaining CE and stability term S together in Fig. 13. Ideally these two components related to stability should be formulated as one term.
APPENDIX B
Acronyms
| 20CR | NOAA Twentieth Century Reanalysis |
| AIRS | Atmospheric Infrared Sounder |
| AMIP | Atmospheric Model Intercomparison Project (prescribed SST) |
| AMSU | Advanced Microwave Sounding Unit |
| ATOVS | Advanced TIROS Operational Vertical Sounder |
| C-C | Clausius–Clapeyron |
| CFSR | Climate Forecast System Reanalysis |
| CPCU | NOAA Climate Prediction Center (CPC) unified precipitation dataset |
| ECMWF | European Centre for Medium-Range Weather Forecasts |
| ENSO | El Niño–Southern Oscillation |
| EOF | Empirical orthogonal function |
| ERA | ECWMF reanalysis (40 yr, Interim, 20C, or 20CM) |
| GEOS-5 | Goddard Earth Observing System, version 5 |
| GPCP | Global Precipitation Climatology Project |
| GPSRO | Global positioning system radio occultation |
| GRACE | Gravity Recovery and Climate Experiment |
| GSI | Gridpoint Statistical Interpolation |
| HIRS | High Resolution Infrared Radiation Sounder |
| IASI | Infrared Atmospheric Sounding Interferometer |
| IAU | Incremental analysis update |
| ITCZ | Intertropical convergence zone |
| JRA-25 | Japanese 25-year Reanalysis |
| JRA-55 | Japanese 55-year Reanalysis |
| LSM | Land surface model |
| M2AMIP | MERRA-2 Atmospheric Model Intercomparison Project experiment |
| MERRA | Modern-Era Retrospective Analysis for Research and Applications |
| MetOp | Meteorological Operational satellite series |
| MHS | Microwave Humidity Sounder |
| MSU | Microwave Sounding Unit |
| NASA | National Aeronautics and Space Administration |
| NCAR | National Center for Atmospheric Research |
| NCEP | National Centers for Environmental Prediction |
| NEWS | NASA Energy and Water Cycle Study program |
| NOAA | National Oceanic and Atmospheric Administration |
| PC | Principal component |
| Pocor | Precipitation that is observation corrected |
| Pmgen | Precipitation that is model generated |
| RSS | Remote Sensing Systems |
| SSM/I | Special Sensor Microwave Imager |
| SST | Sea surface temperature |
| TIROS | Television Infrared Observation Satellite |
| TLT | Temperature of the lower troposphere |
| TOVS | TIROS Operational Vertical Sounder |
| TPW | Total precipitable water |
| VMFC | Vertically integrated moisture flux convergence |
REFERENCES
Adler, R., and Coauthors, 2003: The Version-2 Global Precipitation Climatology Project (GPCP) monthly precipitation analysis (1979–present). J. Hydrometeor., 4, 1147–1167, doi:10.1175/1525-7541(2003)004<1147:TVGPCP>2.0.CO;2.
Bosilovich, M. G., J. Chen, F. R. Robertson, and R. F. Adler, 2008: Evaluation of global precipitation in reanalyses. J. Appl. Meteor. Climatol., 47, 2279–2299, doi:10.1175/2008JAMC1921.1.
Bosilovich, M. G., F. R. Robertson, and J. Chen, 2011: Global energy and water budgets in MERRA. J. Climate, 24, 5721–5739, doi:10.1175/2011JCLI4175.1.
Bosilovich, M. G., J.-D. Chern, D. Mocko, F. R. Robertson, and A. M. da Silva, 2015a: Evaluating observation influence on regional water budgets in reanalyses. J. Climate, 28, 3631–3649, doi:10.1175/JCLI-D-14-00623.1.
Bosilovich, M. G., and Coauthors, 2015b: MERRA-2: Initial evaluation of the climate. NASA Tech. Rep. NASA/TM-2015-104606, Vol. 43, 136 pp. [Available online at https://gmao.gsfc.nasa.gov/pubs/tm/docs/Bosilovich803.pdf.]
Brown, P. J., and C. D. Kummerow, 2014: An assessment of atmospheric water budget components over tropical oceans. J. Climate, 27, 2054–2071, doi:10.1175/JCLI-D-13-00385.1.
Compo, G. P., and Coauthors, 2011: The Twentieth Century Reanalysis Project. Quart. J. Roy. Meteor. Soc., 137, 1–28, doi:10.1002/qj.776.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, doi:10.1002/qj.828.
Ebita, A., and Coauthors, 2011: The Japanese 55-year reanalysis JRA-55: An interim report. SOLA, 7, 149–152.
Emanuel, K. A., 1994: Atmospheric Convection. Oxford University Press, 580 pp.
Enfield, D. B., A. M. Mestas-Nuñez, and P. J. Trimble, 2001: The Atlantic multidecadal oscillation and its relation to rainfall and river flows in the continental U.S. Geophys. Res. Lett., 28, 2077–2080, doi:10.1029/2000GL012745.
GMAO, 2015a: MERRA-2 instM_3d_asm_Np: 3d, monthly mean, instantaneous, pressure-level, assimilation, assimilated meteorological fields, version 5.12.4. Goddard Space Flight Center Distributed Active Archive Center (GSFC DAAC), accessed September 2015, doi:10.5067/2E096JV59PK7.
GMAO, 2015b: MERRA-2 instM_2d_int_Nx: 2d, monthly mean, instantaneous, single-level, assimilation, vertically integrated diagnostics, version 5.12.4. Goddard Space Flight Center Distributed Active Archive Center (GSFC DAAC), accessed September 2015, doi:10.5067/KVTU1A8BWFSJ.
GMAO, 2015c: MERRA-2 tavgM_2d_int_Nx: 2d, monthly mean, time-averaged, single-level, assimilation, vertically integrated diagnostics, version 5.12.4. Goddard Space Flight Center Distributed Active Archive Center (GSFC DAAC), accessed September 2015, doi:10.5067/FQPTQ4OJ22TL.
GMAO, 2015d: MERRA-2 tavgM_2d_flx_Nx: 2d, monthly mean, time-averaged, single-level, assimilation, surface flux diagnostics, version 5.12.4. Goddard Space Flight Center Distributed Active Archive Center (GSFC DAAC), accessed September 2015, doi:10.5067/0JRLVL8YV2Y4.
GMAO, 2015e: MERRA-2 tavgM_3d_qdt_Np: 3d, monthly mean, time-averaged, pressure-level, assimilation, moist tendencies, version 5.12.4. Goddard Space Flight Center Distributed Active Archive Center (GSFC DAAC), accessed September 2015, doi:10.5067/2ZTU87V69ATP.
Held, I. M., and B. J. Soden, 2006: Robust responses of the hydrological cycle to global warming. J. Climate, 19, 5686–5699, doi:10.1175/JCLI3990.1.
Hersbach, H., C. Peubey, A. Simmons, P. Berrisford, P. Poli, and D. Dee, 2015: ERA‐20CM: A twentieth‐century atmospheric model ensemble. Quart. J. Roy. Meteor. Soc., 141, 2350–2375, doi:10.1002/qj.2528.
Jiménez, C., and Coauthors, 2011: Global intercomparison of 12 land surface heat flux estimates. J. Geophys. Res., 116, D02102, doi:10.1029/2010JD014545.
Kobayashi, S., and Coauthors, 2015: The JRA-55 reanalysis: General specifications and basic characteristics. J. Meteor. Soc. Japan, 93, 5–48, doi:10.2151/jmsj.2015-001.
Lean, J., 2000: Evolution of the sun’s spectral irradiance since the Maunder minimum. Geophys. Res. Lett., 27, 2425–2428, doi:10.1029/2000GL000043.
McCarty, W., and Coauthors, 2016: MERRA-2 input observations: Summary and assessment. NASA Tech. Rep. NASA/TM-2015-104606, Vol. 46, 51 pp. [Available online at https://gmao.gsfc.nasa.gov/pubs/tm/docs/McCarty885.pdf.]
Mears, C. A., and F. J. Wentz, 2009: Construction of the RSS V3.2 lower tropospheric dataset from the MSU and AMSU microwave sounders. J. Atmos. Oceanic Technol., 26, 1493–1509, doi:10.1175/2009JTECHA1237.1.
Molod, A., L. Takacs, M. Suarez, J. Bacmeister, I.-S. Song, and A. Eichmann, 2012: The GEOS-5 atmospheric general circulation model: Mean climate and development from MERRA to Fortuna. NASA Tech. Memo. NASA TM-2012-104606, Vol. 28, 117 pp.
Molod, A., L. Takacs, M. Suarez, and J. Bacmeister, 2015: Development of the GEOS-5 atmospheric general circulation model: Evolution from MERRA to MERRA2. Geosci. Model Dev., 8, 1339–1356, doi:10.5194/gmd-8-1339-2015.
Ogata, T., S. P. Xie, A. Wittenberg, and D. Z. Sun, 2013: Interdecadal amplitude modulation of El Niño–Southern Oscillation and its impact on tropical Pacific decadal variability. J. Climate, 26, 7280–7297, doi:10.1175/JCLI-D-12-00415.1.
Poli, P., and Coauthors, 2016: ERA-20C: An atmospheric reanalysis of the twentieth century. J. Climate, 29, 4083–4097, doi:10.1175/JCLI-D-15-0556.1.
Power, S., T. Casey, C. Folland, A. Colman, and V. Mehta, 1999: Inter-decadal modulation of the impact of ENSO on Australia. Climate Dyn., 15, 319–324, doi:10.1007/s003820050284.
Reichle, R. H., and Q. Liu, 2014: Observation-Corrected Precipitation Estimates in GEOS-5. NASA Tech. Memo. NASA/TM-2014-104606, Vol. 35, 24 pp. [Available online at http://gmao.gsfc.nasa.gov/pubs/tm/docs/Reichle734.pdf.]
Reichle, R. H., R. D. Koster, G. J. M. De Lannoy, B. A. Forman, Q. Liu, S. P. P. Mahanama, and A. Toure, 2011: Assessment and enhancement of MERRA land surface hydrology estimates. J. Climate, 24, 6322–6338, doi:10.1175/JCLI-D-10-05033.1.
Reichle, R. H., C. Draper, Q. Liu, M. Girotto, S. Mahanama, R. Koster, and G. De Lannoy, 2016: Assessment of MERRA-2 land surface hydrology estimates. J. Climate, doi:10.1175/JCLI-D-16-0720.1, in press.
Remote Sensing Systems, 2016: Monthly mean total precipitable water data set on a 1 degree grid made from Remote Sensing Systems version-7 microwave radiometer data, V07r00, Feb 2015. Remote Sensing Systems, accessed 10 February 2016. [Available online at www.remss.com.]
Richter, I., and S. P. Xie, 2008: Muted precipitation increase in global warming simulations: A surface evaporation perspective. J. Geophys. Res., 113, D24118, doi:10.1029/2008JD010561.
Rienecker, M. M., and Coauthors, 2008: The GEOS-5 Data Assimilation System—Documentation of versions 5.0.1, 5.1.0, and 5.2.0. NASA Tech. Memo. NASA/TM-2008-104606, Vol. 27, 97 pp. [Available online at http://gmao.gsfc.nasa.gov/pubs/docs/Rienecker369.pdf.]
Rienecker, M. M., and Coauthors, 2011: MERRA: NASA’s Modern-Era Retrospective Analysis for Research and Applications. J. Climate, 24, 3624–3648, doi:10.1175/JCLI-D-11-00015.1.
Roads, J., M. Kanamitsu, and R. Stewart, 2002: CSE water and energy budgets in the NCEP–DOE Reanalysis II. J. Hydrometeor., 3, 227–248, doi:10.1175/1525-7541(2002)003<0227:CWAEBI>2.0.CO;2.
Robertson, F. R., M. G. Bosilovich, J. Chen, and T. L. Miller, 2011: The effect of satellite observing system changes on MERRA water and energy fluxes. J. Climate, 24, 5197–5217, doi:10.1175/2011JCLI4227.1.
Robertson, F. R., M. G. Bosilovich, J. B. Roberts, R. H. Reichle, R. Adler, L. Ricciardulli, W. Berg, and G. J. Huffman, 2014: Consistency of estimated global water cycle variations over the satellite era. J. Climate, 27, 6135–6154, doi:10.1175/JCLI-D-13-00384.1.
Rodell, M., and Coauthors, 2015: The observed state of the water cycle in the early twenty-first century. J. Climate, 28, 8289–8318, doi:10.1175/JCLI-D-14-00555.1.
Saha, S., and Coauthors, 2010: The NCEP Climate Forecast System Reanalysis. Bull. Amer. Meteor. Soc., 91, 1015–1057, doi:10.1175/2010BAMS3001.1.
Schröder, M., M. Lockhoff, J. M. Forsythe, H. Q. Cronk, T. H. Vonder Haar, and R. Bennartz, 2016: The GEWEX water 1 vapor assessment: Results from intercomparison, trend, and homogeneity analysis of total column water vapor. J. Appl. Meteor. Climatol., 55, 1633–1649, doi:10.1175/JAMC-D-15-0304.1.
Sun, D. Z., and I. M. Held, 1996: A comparison of modeled and observed relationships between interannual variations of water vapor and temperature. J. Climate, 9, 665–675, doi:10.1175/1520-0442(1996)009<0665:ACOMAO>2.0.CO;2.
Takacs, L. L., M. Suarez, and R. Todling, 2015: Maintaining atmospheric mass and water balance within reanalysis. NASA Tech. Memo. NASA/TM-2014-104606, Vol. 37, 46 pp. [Available online at http://gmao.gsfc.nasa.gov/pubs/docs/Takacs737.pdf.]
Takacs, L. L., M. Suarez, and R. Todling, 2016: Maintaining atmospheric mass and water balance in reanalyses. Quart. J. Roy. Meteor. Soc., 142, 1565–1573, doi:10.1002/qj.2763.
Trenberth, K. E., J. T. Fasullo, and J. Mackaro, 2011: Atmospheric moisture transports from ocean to land and global energy flows in reanalyses. J. Climate, 24, 4907–4924, doi:10.1175/2011JCLI4171.1.
Wang, Y.-M., J. L. Lean, and N. R. Sheeley Jr., 2005: Modeling the sun’s magnetic field and irradiance since 1713. Astrophys. J., 625, 522–538, doi:10.1086/429689.
Zhang, R., and T. L. Delworth, 2006: Impact of Atlantic multidecadal oscillations on India/Sahel rainfall and Atlantic hurricanes. Geophys. Res. Lett., 33, L17712, doi:10.1029/2006GL026267.
While accounting for changes in water vapor alone would be an excellent approximation in modifications made, for completeness the total water substance includes combined vapor and prognostic cloud liquid and ice water.
Daily precipitation from CPCU was first modified to have its annual climatology match that of GPCP (version 2.1), retaining the CPCU daily anomalies (Reichle and Liu 2014). The earlier MERRA model-generated precipitation was then constrained at each grid point to have the same pentad mean while retaining the full MERRA hour-by-hour variability. This record was then used to force MERRA-2 land surface.
The evaluation of MERRA and CFSR temperature water relationship is not presented because the significant jumps in their total column water vapor time series affects trends and the computation of the relationship. Schröder et al. (2016) do provide some estimates of their relationships.