The NIES TM is an off-line model driven by Japanese reanalysis data, which covers more than 30 years from 1 January 1979 to present (Onogi et al., 2007). The period of 1979–2004 is covered by the Japanese 25 year Reanalysis (JRA-25), used by Belikov et al. (2013), and is the product of the Japan Meteorological Agency (JMA) and Central Research Institute of Electric Power Industry (CRIEP). After 2005, real-time operational analysis, employing the same assimilation system as JRA-25, has been continued as the JMA Climate Data Assimilation System (JCDAS). The JRA-25/JCDAS dataset is distributed on a Gaussian horizontal grid T106 (320 × 160) with 40 hybrid σ–p levels. The 6 hourly time step of JRA-25/JCDAS is coarser than the 3 hourly data from the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) and Global Point Value (GPV) datasets, which were used in the previous model version (Belikov et al., 2011). However, with a better vertical resolution (40 levels on a hybrid σ-p grid vs. 25 and 21 pressure levels for GFS and GPV, respectively) it is possible to implement a vertical grid with 32 levels (vs. the 25 levels used before), resulting in a more detailed resolution of the boundary layer and UTLS region (Table 1).

The 2-D monthly distribution of the climatological heating rate, used to calculate vertical transport in the θ-coordinate domain of the hybrid sigma-isentropic coordinate, is prepared from JCDAS reanalysis data, which are provided as the sum of short- and long-wave components on pressure levels.

The HIPPO study investigated the carbon cycle and greenhouse gases at various altitudes (from 0 to 16 km) in the western hemisphere through the annual cycle. HIPPO is supported by the National Science Foundation (NSF) and its operations are managed by the Earth Observing Laboratory (EOL) of the National Center for Atmospheric Research (NCAR). Its base of operations is the EOL Research Aviation Facility (RAF) at the Rocky Mountain Metropolitan Airport (RMMA) in Jefferson Country, Colorado. The main goal of HIPPO was to determine the global distribution of CO2 and other trace atmospheric gases by sampling at several altitudes and latitudes (from 0 to 16 km, 87.0∘ N to -67.2∘ S) in the Pacific Basin.

The dataset used in this paper includes the merged 10 s data product of meteorological, atmospheric chemistry, and aerosol measurements from three HIPPO Missions 1 to 3. The three missions took place from January 2009 to April 2010; HIPPO-1 (9–26 January 2009), HIPPO-2 (2–22 November 2009), and HIPPO-3 (24 March–15 April 2010), ranging from 128.0∘ E to -84.0∘ W, and 87.0∘ N to -67.2∘ S (Table 2). All data are provided in a single space-delimited format ASCII file (https://www.eol.ucar.edu/field_projects/hippo).

HIPPO measured atmospheric constituents along transects running approximately pole-to-pole over the Pacific Ocean and recorded hundreds of vertical profiles from the ocean/ice surface up to the tropopause five times during four seasons from January 2009 to September 2011. HIPPO provides the first high-resolution vertically resolved global survey of a comprehensive suite of atmospheric trace gases and aerosols pertinent to understanding the carbon cycle and challenging global climate models. The 10 s merge product applied in this study was derived by combining the National Science foundation (NSF)/NCAR GV aircraft navigation and atmospheric structure parameters including position, time, temperature, pressure, and wind speed reported at 1 s frequency, with meteorological, atmospheric chemistry and aerosol measurements made by several teams of investigators on a common time and position basis.

The standard model was run with the three HIPPO missions to study atmospheric tracer transport and the ability of the model to reproduce the column-averaged dry air mole fractions and vertical profile of atmospheric CO2. The model was run with a 6 h time step and 1∘ space step at a horizontal resolution of 2.5∘×2.5∘ and 32 vertical levels from the surface to 3 hPa using tracer CO2.

The CO2 simulations were began on 1 January 2009, 1 November 2009 and 1 March 2010 for the three HIPPO missions 1 to 3, respectively, with individual initial 3-D tracer distributions using the Level 4A global fluxes of biosphere–atmosphere, fossil fuel, air–ocean exchange, biomass burning, and inverse correction, obtained by monthly global CO2 flux estimated from FTS (SWIR) level 2 XCO2 data.

Figure 1 show the scatters diagram of modeled results vs. total column of HIPPO-1, 2, 3. The majority of points are within a 95 % confidence interval of total CO2 column concentration. Modeled HIPPO-1's precision successively exceeds 2 and 3, inferring the simulation results with the relevant either seasonal changes or data quality.

The simulation results of CO2 concentration time-varying for HIPPO-1 using the standard model display good performance and weak dispersion of concentrations. The validation results (Fig. 2a) show that approximately 69.2 % of the absolute biases are within 1 ppmv, approximately 92.3 % are within 2 ppmv, and only 7.7 % exceed 3 ppmv. Furthermore, as shown by the root-mean-square error (RMSE) with time, during most days in January the model values were stable compared with the observed values, apart from the first few days of the month. According to the simulation results of the HIPPO-1 observed and simulated latitude-varying CO2 concentration data, the comparison values always underestimate the atmospheric XCO2, and the differences are all within 1.5 ppmv in the Southern Hemisphere, and vice versa in the Northern Hemisphere with 85.8 % of the differences under 1.1 ppmv. Figure 2b shows that the larger biases usually occur in the Northern Hemisphere high latitudes. The RMSE also reflects the instability of the simulated values in the Northern Hemisphere high latitudes.

For HIPPO-2 data from 2 to 22 November 2009,the absolute biases of observed and simulated time-varying are all within 2 ppmv, and 77.8 % of the differences are less than 1 ppmv (Fig. 2c). Approximately 5/6 of the data over the month show comparative stability. Similarly with HIPPO-1, the simulation results are always underestimates in the Southern Hemisphere and overestimates in the Northern Hemisphere. As shown in Fig. 2d, the complete simulation displays good performance, apart from one day in the Northern Hemisphere high latitudes. In the same manner, the RMSE shows good stability in the Southern Hemisphere, in particular for the low-to mid-latitudes of the Southern Hemisphere. The model also simulates well in the Northern Hemisphere, especially from 45 to 70∘ N.

Based on HIPPO-3 data from 24 March to 15 April 2010, the model simulation overestimates in March and underestimates in April. As shown in Fig. 2e, in March, the biases over several days were over 2 ppmv, and one of these days exceeded 3 ppmv. However, the absolute biases were all within 2 ppmv in April, and 75 % of the absolute biases were less than 1 ppmv, which suggests relatively good performance by the model simulation. As shown by the RMSE, the data for the last days in March were not stable. However, 81.8 % of the data in April showed comparatively good stability. The absolute biases are all under 1.5 ppmv in the Southern Hemisphere, and are also within 2 ppmv for the low- and mid-latitudes of the Northern Hemisphere (Fig. 2f). However, a relatively large difference occurs at the Northern Hemisphere high latitudes, at one point exceeding 3 ppmv. Furthermore, the RMSE become greater with latitude from the Southern to Northern Hemisphere, inferring the simulation results are increasingly unstable with increasing latitude.

The GV flight plan performed 787 vertical ascents/descents from the ocean/ice surface/land surface to the tropopause. Two maximum altitude ascents were planned per flight to the tropopause/LS; one in the first half and the other in the second half of the research flight. In between, several vertical profiles from below the PBL to the mid-troposphere (1000–28 000 ft) were flown. Profiles were flown approximately every 2.2∘ of latitude with 4.4∘ between consecutive near-surface or high-altitude samples. Rate of climb and descent was 1500 ftmin-1 (457 mmin-1). During these profiles, the GV averaged a ground speed of approximately 175 ms-1, or 10 kmmin-1.

Most of a flight was conducted below the international Reduced Vertical Separation Minimum (RVSM), usually 29 000 ft or 8850 m, to allow the GV to descend and climb constantly to collect data at different altitudes throughout the troposphere. All flight plans were subject to modifications depending on local atmospheric conditions and approval by air traffic control. Most profiles extended from approximately 300 to 8500 m altitude, constrained by air traffic, but significant profiling extended above approximately 14 km.

One of the aims of this paper was to validate the model column-averaged concentration against the typical HIPPO flight plans, and we therefore examined the variability of CO2 concentrations with HIPPO merged 10 s meteorology, atmospheric chemistry, and aerosol measurements from Mission 1 to 3. For each mission, several hundred vertical profiles were produced. We have only selected the vertical profiles from near-surface to LS to compare the simulations using the standard model with observations. Each mission can be divided into six parts for analysis; the low-, mid- and high-latitudes in the Southern and Northern Hemispheres, respectively.

For HIPPO-1, the total simulation value is always less than the observation value in the Southern Hemisphere and vice versa in the Northern Hemisphere. The bias is less than 2 ppmv for the entire profile from the near-surface to the LS; however, it increases from 2 to 4 ppmv above 10 km covering the Northern Hemisphere high latitudes.

Figure 3 shows the comparison of simulation results and observations for data from the near-surface to the LS in the low-, mid- and high- latitude. In the low-latitudes, as shown by Fig. 3c and d, the simulation performed very well compared with observations. With the exception of the biases of approximately 2 ppmv in the tropopause in Fig. 3d, the biases are all within 1 ppmv. In the mid- and high-latitudes, it is different in both hemispheres. In the Southern Hemisphere, the majority biases are within 2 ppmv but the LS zone in Fig. 3a and 2 to 6 km region in Fig. 3b. In the Northern Hemisphere (Fig. 3e and f), the simulated vertical profiles show good performances, apart from UTLS, and the biases are less than 2 ppmv. Some large biases occurred in the UTLS exceeding 4 ppmv when the potential temperature gradient increased rapidly with height.

HIPPO-2 data showed overall similarity with HIPPO-1 data based on the distribution of positive and negative bias. However, an anomaly occurred at approximately -60∘ and 75∘ latitude, showing positive and negative biases, respectively, some exceeding 6 ppmv. Figure 4a is the vertical profile of the Southern Hemisphere high latitudes, which clearly shows that the simulation matches well with the observations from the near-surface to the tropopause. However, large biases occur above 8 km; Fig. 4b also shows this phenomenon above 10 km. In the low latitudes (Fig. 4c and d), the simulations match well with observations. The potential temperature gradient is smooth and the biases are less than 1 ppmv from near-surface to the UT, which indicates good performance. For the mid-latitudes of the Northern Hemisphere, Fig. 4e shows relatively good simulation performance. However, as shown in Fig. 4f, the high latitudes did not perform well in the near-surface or the low- and mid-troposphere. Compared with observations, the simulation profiles do not appear to reflect the original shape.

As shown by HIPPO-3 data the biases increase abruptly with flight height for the mid- to high-latitudes of the Northern Hemisphere with values reaching 7 ppmv. In the high-latitudes of the Southern Hemisphere (Fig. 5a) the simulation underestimates the observations, and the absolute biases are isostatic from the near-surface to the LS, which are less than 3 ppmv. The Southern Hemisphere low latitudes (Fig. 5c) indicate good performance of the simulations, where all the biases are less than 1 ppmv. In the Northern Hemisphere low latitudes (Fig. 5d), the entire simulation appears to match well with observations. However, some locations do not reproduce the precise shape through the entire height. For the mid- to high- northern latitudes (Fig. 5e and f), the simulations performed relatively well from the near-surface to the UT. Larger bias in simulations is found in the winter lower stratosphere in the northern high-latitudes. The problem appears because between tropopause and 350 K level model uses vertical wind provided by reanalysis instead of using radiative heating rate, which is more accurate in stratosphere. The positive bias can reach level of 4 ppm for CO2. This problem only affects simulations for observation made in lower stratosphere in high latitudes in cold season when the tropopause level is low. However the number of in-situ observations made in this altitude is very limited. The satellite observations of the total column such as GOSAT are also reduced considerably in high latitudes in cold season (Yoshida et al., 2013). Thus this lower stratosphere bias is not likely to deteriorate the transport model performance in the inverse modeling applications.