Concept Development and Risk Reduction for MISTiC Winds, A Micro-Satellite Constellation Approach for Vertically Resolved Wind and IR Sounding Observations in the Troposphere

Abstract

MISTiC Winds is an instrument and constellation mission approach to simultaneously observe the global thermodynamic state and the vertically resolved horizontal wind field in the troposphere from LEO SSO. The instrument is a wide-field imaging spectrometer operated in the 4.05–5.75 μm range, with the spectral resolution, sampling, radiometric sensitivity, and stability needed to provide temperature and water vapor soundings of the atmosphere, with 1 km vertical resolution in the troposphere-comparable to those of NASA’s atmospheric infrared sounder (AIRS). These instruments have much higher spatial resolution (<3 km at nadir) and finer spatial sampling than current hyperspectral sounders, allowing a sequence of such observations from several micro-satellites in an orbital plane with short time separation, from which atmospheric motion vector (AMV) winds are derived. AMVs for both cloud-motion and water vapor-motion, derived from hyperspectral imagery, will have improved velocity resolution relative to AMVs obtained from multi-spectral instruments operating in GEO. MISTiC’s extraordinarily small size, low mass (<15 kg), and minimal cooling requirements can be accommodated aboard an ESPA-class microsatellite. Low fabrication and launch costs enable this constellation to provide more frequent atmospheric observations than current-generation sounders provide, at much lower mission cost. Key technology and observation method risks have been reduced through recent laboratory and airborne (NASA ER2) testing funded under NASA’s Instrument Incubator Program and BAE Systems IR&D, and through an observing system simulation experiment performed by NASA GMAO. This approach would provide a valuable new capability for the study of the processes driving high-impact weather events, and critical high-resolution observations needed for future numerical weather prediction.

(Not export controlled per ES-C4ISR-073019-0179, Approved for Public Release; Unlimited Distribution)

1. Introduction

1.1. Overview

To specify the current thermodynamic state of the atmosphere and to predict its future state, one must know the flow field in addition to the mass field [1]. The latent heat present in the water vapor field is one of the most dominant forms of energy storage in the atmosphere. Transport of this stored energy by the three-dimensional wind field is one of the most fundamental processes driving our weather, and is a key element in both the water and energy cycles [1].

Observations of the thermodynamic state of the atmosphere from low earth orbit (LEO) using infrared (IR) and microwave spectral sounding methods have enabled significant advances in our understanding of atmospheric processes [2]. These observations currently have a stronger impact on global numerical weather forecast model accuracy than any other type of observation [3] (p.307). However, it has long been recognized that more accurate, vertically resolved, wind field observations in the troposphere would provide a further significant improvement in our ability to characterize the processes that lead to high impact weather events and more generally, improve weather forecast accuracy [4].

The horizontal wind field has been observed from space since the introduction of the geostationary weather satellite in the 1960s using the method of atmospheric motion vectors (AMVs) [1]. This method has been refined substantially over the years, and is currently employed operationally using the observations from multi-spectral meteorological imagers worldwide. AMVs provide an important positive impact on numerical weather prediction (NWP) [5]. However, their full impact on NWP has been constrained by the limitations on height assignment accuracy inherent in multi-spectral meteorological imaging observations [6].

Over the years, several approaches that would provide vertically resolved wind field observations have been advanced, including Doppler wind light detection and ranging (LIDAR or DWL) [4], and geostationary hyperspectral infrared sounders (GEO hyperspectral) [7], with sufficient spatial resolution to enable motion-vector wind observations. Each of these approaches has received substantial investment, and technical progress has been made, but the cost of these approaches remains very high. Hundreds of millions of dollars were spent on GEO hyperspectral sounding developments during the last few decades, but these efforts in the US were halted prior to fully developing and launching an instrument. Space-based Doppler wind LIDAR has been under development for an even longer period than has hyperspectral IR sounding. After two decades of development, a single vector component/single curtain DWL [8] was recently launched by the European Space Agency and is undergoing check-out at a very high overall program cost of more than $550M.

The National Research Council (NRC), in the recently released Decadal Survey for Earth Sciences [3] (p. 10, p. 138), highlighted once again the importance of vertically resolved tropospheric winds as well as the need for thermodynamic vertical profiles at higher spatial resolution, vertical resolution, and observation tempo. The purpose of this paper is to present a new observation approach, MISTiC^® Winds (Midwave Infrared Sounding of Temperature and Humidity in a Constellation for Winds), which is highly responsive to the observation needs identified by the NRC, and to contemporary fiscal constraints. MISTiC Winds would employ miniature infrared hyperspectral sounders in a LEO microsatellite constellation that will provide vertically resolved wind velocity observations using hyperspectral and imaging methods, together with more frequent, higher spatial resolution soundings of the troposphere, at much lower cost than DWL and GEO hyperspectral implementations. A preliminary and partial presentation of MISTiC Winds was given previously [9,10,11].

The remainder of this paper is organized in the following way. Section 1.2 will briefly review the historical background of hyperspectral infrared sounding, previous attempts to bring this capability to geostationary orbit, and multispectral imaging methods for motion vector winds. Section 2 will describe the MISTiC Winds instrument and mission requirements, the observations currently provided by the program of record, and NASA and NOAA’s earlier GEO hyperspectral efforts. Section 3 will present the key results of work to reduce the risks associated with the MISTiC Winds approach, including a description of key technology risk reduction development and testing activities, and high-altitude airborne demonstrations of the observation method. Section 4 presents an analysis of MISTiC Winds requirements, their derivation, their feasibility, and planned observation capability in the context of NASA needs. A summary is presented in Section 5.

1.2. Historical Background for IR Sounding and Atmospheric Motion Vector Winds

The concept for using infrared radiance spectra to measure the vertical profile of atmospheric temperature and water vapor (infrared sounding) was first advanced by Kaplan et al. [12] in 1959. Hyperspectral infrared sounding of the temperature and moisture vertical profiles have been performed from space since the NASA atmospheric infrared sounder (AIRS) instrument was launched aboard Aqua in 2002. NASA’s AIRS instrument [13], built by BAE Systems, provided the first on-orbit demonstration of Kaplan’s method implemented with hyperspectral IR radiances. Weather and climate research based on AIRS data have made significant contributions to our understanding of the earth system [2]. Observations from AIRS and other LEO hyperspectral sounders, the infrared atmospheric interferometer sounder (IASI) and cross-track infrared sounder (CrIS), have since become some of the most important inputs into global operational numerical weather models [14,15,16]. A recent Forecast Sensitivity Observation Impact assessment by the UK Met Office [3] (p.307) ranks IR sounders as having the most significant beneficial impact of any observation in reducing the 24-h global forecast error. However, while continued incremental improvements are likely, the relative infrequency of these LEO sounding observations and the relatively large size of IR sounder spatial IFOV (now comparable to the grid spacing of global NWP models) will constrain further improvements.

Horizontal wind fields using AMVs have been derived from observations of geostationary meteorological multispectral imaging instruments (GEO MET Imagers) since the introduction of the operational geostationary weather satellite in the 1970s and ‘80s [1,17], (now the Geostationary Operational Satellite System) and are currently being provided for the US, operationally, by the GOES 16 and GOES 17 imagers and by the nearly identical Himawari instruments. GEO MET imager-based wind observations provide a positive impact to the numerical weather prediction (NWP) models, although not with as significant an impact as those provided by IR and microwave sounding.

AMVs are determined from observing the horizontal displacement of a feature in either the 2D water vapor field, or the cloud field, over a known time interval. This feature velocity, identified with the wind velocity, must be assigned a height in the atmosphere to be used in NWP, and different operational centers deriving AMVs have different methods for this height assignment. The two leading wind-velocity error terms for GEO MET Imagers are the errors in the vertical height assignment (in the presence of wind shear) and errors in the relative position change of the tracked feature, also referred to as a tracking error. The GOES ABI (advanced baseline imager) improves the accuracy of the feature tracking term relative to prior geostationary imagers, by moving from the 4 km resolution of prior GOES imagers to 2 km for the IR channels in ABI at nadir. However, the height assignment error term has a basic cause linked to a fundamental design attribute of multi-spectral imagers, the spectral channel bandwidth generally employed, which varies from 0.5 to 1.0 micrometers for IR window and water vapor channels. For water vapor channels, these bandwidths are broad enough that they include many atmospheric absorption or emission lines for water vapor, with a broad range of line strengths. As a consequence, the vertical contribution functions for such channels are very broad, extending over much of the troposphere, and accurate feature height assignment is difficult. In current instruments, these bands were chosen to emphasize higher-level water vapor and avoid surface emission. One consequence of this choice is that these wind observations are primarily weighted high in the troposphere. Cloud motion vectors are observed using atmospheric window channels. In the thermal infrared, cloud brightness temperature could indicate the cloud pressure height, but only if the vertical temperature profile was also known. In the absence of temperature profile knowledge, various approaches to height assignment have been developed, but these have shortcomings. Channel brightness temperature could also be an indicator of height for water vapor motion-vector wind tracers, but determining the height from the radiance requires knowledge of both the temperature and water vapor concentration vertical profiles.

As a result of this missing information, the AMV wind speed error associated with current multispectral imager observations remains relatively large. For the GOES 16 imager, errors near 5–6 m/s rms relative to radiosondes (near 500 hPa) are reported [18], which is approximately 2–3× larger than the wind speed error needed by NWP models [19]. An additional concern presented by the multi-spectral imaging approach for AMVs is that these wind observations are not well distributed vertically through the troposphere. Cloud motion vector winds (which, of course, require the presence of clouds for an observation) are primarily located in the lower troposphere, well below 600 hPa, whereas moisture-motion-vector winds are primarily located in the upper troposphere, above 400 hPa [18]. Very few wind observations are available near 500 hPa, where observing system simulation experiments (OSSEs) show that wind observations would have the greatest impact on weather forecast accuracy [19].

Hyperspectral infrared sounding has an inherent capacity to provide pressure-height information for the image-features observed (through the retrieved vertical profiles) that has long been recognized as beneficial for AMVs, and wind retrieval approaches employing hyperspectral techniques have been developed [20,21]. The challenge has been to provide the hyperspectral sounding observations themselves at a temporal and spatial resolution that would support wind observations-affordably.

In the US, NASA initially selected the geosynchronous imaging Fourier transform spectrometer (GIFTS), a GEO hyperspectral IR sounder, under its New Millennium Program, with plans for higher temporal rate IR soundings as well as hyperspectral AMV winds [7]. However, GIFTS was canceled, with the projected total program cost relative to the mission cost cap cited as an important factor. NOAA thoroughly evaluated the possibility of including a hyperspectral IR sounder (that would have supported hyperspectral AMVs) in the GOES-R series, supporting instrument concept development, trade studies, and technology risk reduction efforts through the Formulation Phase Program. However, NOAA determined that the cost and risk were too high in light of GOES-R budget constraints [22]. The indefinite deferral by NOAA’s GOES Program of hyperspectral observation methods, even though they have been shown to provide significant weather forecast impact, was one of the primary motivations for the development of MISTiC Winds. MISTiC Winds addresses both of these AMV wind velocity error terms, (height assignment error and tracking error) in mission implementation that is much lower in cost, more flexible, and more resilient than providing these observations from a GOES-R class IR sounder. MISTiC Winds does this while providing a measurement of the vertical temperature and moisture profile of the troposphere at unprecedented spatial resolution and refresh rate.

The approach to vertically resolved wind observations has long been envisioned as a Doppler Wind LIDAR. However, MISTiC Winds may provide tropospheric wind measurements with vertical resolution sufficient for many purposes, at a substantially lower cost and under a wide range of conditions. MISTiC Winds observations may also provide a highly effective complement to a DWL, extending the impact of the more accurate vertical profiling achievable with a LIDAR over a narrow spatial field to the broader spatial field that a mapping observation such as MISTiC Winds would provide.

A recent OSSE performed by NASA’s Global Modeling and Assimilation Office (GMAO) [23] simulating MISTiC observations of vertically resolved winds and IR hyperspectral radiances has shown significant weather forecast impact from both the assimilated wind and radiance observations. Experimental evidence of weather forecast improvement from retrieved wind observations derived using hyperspectral sounding methods has recently emerged from the AIRS Polar Winds observing system experiments [24], where these satellite-derived moisture motion-vector winds have been shown to have the largest impact of any other observation in the data-sparse polar region. These improvements were achieved, even with the disadvantages of a relatively large AIRS footprint (15 km at nadir) and the relatively long (~100 min) period between observations available with polar-crossing LEO sun-synchronous orbit satellites.

2. Materials and Methods

The development of every new instrument and mission approach is a balancing act at some level, and this is especially true for MISTiC Winds. We seek to combine the most essential aspects of infrared sounding requirements and instrument design with the most essential aspects of meteorological imagers designed to obtain the image sequences with the fidelity needed to track the wind-driven features to provide new observation capability that will enable significant advances weather process understanding and weather forecast accuracy improvements. We seek to accomplish this in an era defined by expanding observation needs but flat to declining budgets for new observations, but also one in which there will be a rich international operational system of systems for weather remote sensing observations that this new capability should complement, but need not replicate. These considerations strongly drive requirements and design choices.

2.1. MISTiC Winds Observing Concept, Requirements, and MISTiC Instrument Concept

One of the central ideas behind MISTiC Winds is the use of a constellation of micro-satellites, each hosting a miniature infrared spectral sounding instrument, to enable much more frequent observations of the three-dimensional atmosphere than a single instrument in LEO could provide. A constellation concept for these observation employs groups of three satellites in a (nearly) common orbital plane, with nodal crossing times separated by a short interval-nominally 10 min. One such wind-triplet group is shown, in blue, in Figure 1. Each group of three instruments observes a motion-vector wind triplet of hyperspectral images, analogous to three sequential images of a region by a GEO met-imager such as GOES, that are currently used to derive AMVs. In Figure 1, the light blue rectangles indicate the approximate ground-projection of the hypercube, which is acquired as the slit image of a spectrometer is scanned in a direction perpendicular to the polar orbital track (an alternative mission possibility, more frequent IR sounding, but without wind observations was also considered, and its orbits are shown in Figure 1 in goldenrod, but not discussed further).

AMV winds, such as those shown in Figure 2, represent the current derived product, with horizontal wind speed and direction being determined from tracer observations in three radiance images. In the hyperspectral AMV case, the feature tracked would either be a cloud or a geometric feature of the retrieved water vapor field. Hyperspectral AMVs will build on methods developed for multispectral AMVs, but will feature more layers, with better vertical resolution than available from current multispectral imager-derived AMVs.

A potentially more powerful, but more data-intensive wind retrieval method borrowed from the computer vision community, Optical Flow, is currently being developed by several groups [21] for use in wind field retrieval from meteorological imaging sensors. This paper will focus on the hyper-spectral extension of the current AMV method, but MISTiC observations are ideally suited for an Optical Flow-based retrieval method as well. Six MISTiC satellites in this constellation type would provide vertically-resolved wind velocity measurement, globally, with a refresh rate of six hours, together with high spatial resolution vertical temperature and moisture soundings of the troposphere at the same six-hour refresh rate as one example.

Each MISTiC instrument would collect a series of hyper-cubes as it progresses along the orbital track. A MISTiC hypercube includes 480 along-track spatial samples, more than 1000 spatial samples cross-track, and 580 spectral samples. The vertically resolved tropospheric wind is obtained through higher-level data processing of the hyperspectral image data, using methods drawn from both IR sounding and feature-tracking AMV methods, or optical flow methods, using data from all three hyper-cubes provided by the satellite triplet.

The MISTiC instrument is a dispersive imaging spectrometer, which combines features of BAE Systems’ airborne and ground-based hyperspectral imaging systems developed for the Department of Defense, with spectral resolution, range, spectral calibration stability, and spectral knowledge requirements derived from the NASA AIRS instrument requirements and GOES-R Hyperspectral Environmental Suite (HES) infrared sounding requirements. A detailed physical model view of this instrument is shown in Figure 3. The instrument will be carried aboard an ESPA-class micro-satellite in polar sun-synchronous orbit.

Key design elements of the instrument are identified in Figure 3. An internal scan mirror scans the slit-image field of view across the surface of the earth from east to west as the satellite progresses through the orbit. Like AIRS, the instrument will periodically observe both deep space, and an on-board blackbody source to provide the data used to calibrate the raw radiances, correcting for detector non-uniformity effects. Sensitive spectrometers operating in this portion of the infrared band require cooling for both the spectrometer optics and the detector array. The MISTiC spectrometer is passively cooled by using radiator panels that reject heat to space. A sun-synchronous orbit (SSO) with appropriate nodal crossing times is used so that the spacecraft and instrument are always provided with a surface for this passive radiator that is shielded from solar illumination and earth emission, without requiring spacecraft attitude adjustments. The detector is actively cooled using a miniature pulse-tube cryo-cooler to ensure both sufficient on-orbit life and to minimize vibration-induced spectral observation errors.

The overarching design goal adopted for MISTiC Winds is to provide high-quality observations of proven scientific and meteorological value, while making dramatic reductions in size, mass, power, and cost for the instruments. This goal is pursued in a comprehensive and balanced way that includes:

Tailoring requirements for the 3D winds and temperature/humidity sounding to emphasize observations from the troposphere, by adapting infrared resolving power, infrared spectral range, while minimizing instrument size, mass, power, and overall space segment cost.

Using innovative instrument opto-mechanical and thermal design methods and adopting AIRS-proven orbit and passive spectrometer cooling to minimize instrument electrical power.

Leveraging advanced instrument technology and the rapidly developing and increasingly capable CubeSat and microsat-satellite technology base including miniature coolers.

As mentioned above, relocating hyperspectral IR observations originally envisioned for GEO to a LEO constellation that can provide comparable refresh rates would provide a major reduction in required resources. A second critical requirements adjustment is to constrain spectral coverage range and resolution to those attributes needed to observe the troposphere. This choice emerges from a system-of-systems perspective, which recognizes that thermodynamic soundings of the stratosphere will be provided operationally by other sounders, allowing MISTiC Winds observations to focus on the troposphere. MISTiC spectral band selection leverage scientific and algorithmic advances of the AIRS Science Team, which uses portions of this band for vertical temperature sounding, and whose retrieval algorithms have mastered the challenges posed by non-local thermal equilibrium and surface reflectance in this band [25].

Figure 4 is an example spectrum from an IASI overpass that illustrates the planned spectral range for MISTiC observations, and key atmospheric features within this range. It includes spectral radiances strongly influenced by CO₂ emission in the troposphere (near 2385 cm⁻¹) from which the vertical temperature profile is extracted, and radiances strongly influenced by atmospheric moisture fields (between 1750 and 2000 cm⁻¹), that, together with the temperature profile, can be used to extract vertical moisture profile. This range also includes channels strongly influenced by clouds and aerosols, and by the trace gases CO, and N₂O in the free troposphere. In Figure 4, the R and P branches of the CO₂ emission dominated region centered on 2385 cm⁻¹ that are used for retrieval of the temperature profile are indicated by A. A very high transmission “super window” channel, which has utility in observing aerosols is indicated by B. Relatively high brightness channels near C are dominated by surface and boundary-layer water vapor emission. Spectral channels, such as D, are more sensitive to lower troposphere water vapor emission. Lower brightness channels, such as those indicated by E and F, are more sensitive to mid-troposphere and upper troposphere water vapor emission respectively. G indicates the spectral coverage range for the airborne MISTiC instrument. H indicates the spectral coverage range for the planned space instrument (1750 cm⁻¹ to 2450 cm⁻¹). As detailed in Appendix A, the airborne unit employed an off-the-shelf detector array with a nominal cutoff above 1850 cm⁻¹. However, employing large-area spatial averaging allowed a demonstration of the spectrometer down to approximately 1830 cm⁻¹ with the airborne unit.

The adopted spectral range specifically excludes the very long-wave spectral channels also dominated by CO₂ emission, for to include these would dramatically increase the required electrical power, instrument size, and exceed the accommodation capacity of an ESPA-class microsatellite. AIRS algorithms have advanced to the point where the temperature profile retrieval employs the shorter-wavelength CO₂ emission channels to retrieve troposphere temperature, with the long-wave channels supporting cloud-clearing and the retrieval of stratospheric temperature [25]. By mission design, observation of stratospheric temperature is not essential for MISTiC, since these variables will be well-observed by the NASA/NOAA Joint Polar Satellite System (JPSS) and EUMETSAT IASI instruments for decades to come.

An important requirement consideration for MISTiC is the selection of the lower frequency limit (upper wavelength limit) of the spectral range. This requirement strongly influences the retrievable information content and data quality for the moisture profile in the upper troposphere, and also strongly influences the size, and power demands of the instrument, due to the exponential dependencies of detector characteristics on bandgap and temperature.

Under the NASA Instrument Incubator Program (IIP), the impact on retrieved layer RMS error for both the temperature and water vapor concentration was evaluated for different values of the lower-frequency band limit, ranging from 1950 down to 1650 cm⁻¹. For this study, “truth” was selected to be a set of AIRS instrument retrieved temperature and moisture vertical profiles collected for cloud-free footprints over the Pacific Ocean, and retrieved using the AIRS Version 6 algorithm, which includes channels from the full spectral range of AIRS, including those in the very long wavelength infrared (VLWIR). A regression method was used by the NASA GSFC Sounder Research Team to perform this trade-study analysis. Spectral resolving power of 700:1 was used for the MISTiC spectral response function in this study.

The key results of this trade study are shown in Figure 5. One conclusion is that reasonably accurate temperature vertical profiles can be provided up to a pressure-height of 200 hPa, using just the CO₂ emission band centered on 4.3 μm (2383 cm⁻¹). A second key result is that a lower limit of 1750 cm⁻¹ is sufficient to provide a water vapor concentration retrieval of sufficient accuracy in the troposphere. While a lower-limit of 1650 cm⁻¹ would provide slightly better water vapor retrievals, it would also drive out the lower-frequency detector spectral cutoff wavelength, and drive up the required instrument cooling power. A higher spectral cutoff such as 1950 cm⁻¹ would degrade water vapor concentration measurement accuracy in the mid-troposphere.

Overall, the major benefit of this mission-specific tailoring is a reduction in spectral channel count by a factor of 3× to 4× for MISTiC relative to current hyperspectral sounders, and the exclusion of the most resource-intensive channels from the instrument. It has also since been recognized that the temperature vertical profile can be accurately retrieved with slightly coarser spectral resolution than used for AIRS, (adopting the effective spectral resolution of CrIS), and this fact has been taken into account in the selection of MISTiC Winds spectral resolution and sampling requirements.

The combination of requirements choices, design, and the use of contemporary technology in MISTiC Winds allow tremendous reductions in instrument size. Figure 6 shows the NASA AIRS instrument, together with the MISTiC instrument concept at the same scale. The GOES-R HES Formulation Phase studies yielded GEO sounder instrument designs from three industry teams, each with size much larger than AIRS, and very much larger than the set of three MISTiC instruments needed to perform a (triplet) hyperspectral AMV observation.

Key MISTiC instrument requirements developed under the IIP are shown in

Table 1. Key MISTiC instrument requirements.

MISTiC Key Instrument Performance Requirements
Instrument Characteristic	Value	Comments
Minimum Spectral Frequency	1750 cm−1	5.72 μm
Maximum Spectral Frequency	2450 cm−1	4.082 μm
Spectral Sampling	~2:1	~580 spectral samples
Spectral Resolving Power	> 700:1	ν/δν (similar to CrIS-Apodized)
Spectral Calibration Knowledge	1/100,000	δλ/λ
Angular Sampling (Spatial Sampling)	0.0016 radians x 0.0016 radians	1.38 km (@ Nadir)
Orbital Altitude and Orbit Type	705.3 km	Polar/Sun-Synchronous
Angular Range (cross-track)	1.570 radians	90 Degrees—Same as AIRS
Spatial Resolution	<3.0 km (geometric mean)	@ Nadir
Radiometric Sensitivity	<200 mK (Ref 250 K scene)	(<150 mK @ 2380 cm−1)
Radiometric Accuracy	<1%	@ 300 K Scene Background

. These are followed by the key MISTiC Winds Level 2 data product requirements for vertically resolved wind, temperature, and water vapor observation capability in

Table 2. Key vertically resolved wind and thermodynamic profile observation requirements.

MISTiC Winds Key Observation Requirements		Value
Vertically Resolved Motion Vector Winds (Water Vapor and Cloud Motion Vectors)	Layer Wind Speed Uncertainty	<2 m/s rms (Lower Troposphere) <3 m/s (High Troposphere)
	Layer Wind Direction Uncertainty (above 10 m/s)	<10 degrees rms
	Layer Height Pressure Height Assignment Accuracy	<30 hPa rms (assuming 850–200 hPa wind shear <20 m/s)
	Layer Effective Vertical Thickness	<100 hPa (FWHM)
	Minimum Pressure Height of Highest Level	350 hPa (WV)/500 hPa (C)
	Tracer Potential Density (Cloud-Free Conditions for Water Vapor Motion Vector, Cloud Contrast for Cloud Motion Vector)	>1 per 6 km2 per vertical layer (Water Vapor-nadir) >1 per 150 km2 per layer (Cloud AMVs @ nadir)
Temperature Vertical Profile	Layer Effective Vertical Thickness	>100 hPa (~1 km)
	Layer Temperature Accuracy	<1 .25 K (Lower Troposphere)
	Layer Water Vapor Concentration Accuracy	<15% (Lower Troposphere)
	Sounding Measurement Potential Density	>1 per 6 km2
Observation Frequency	Observation Refresh Period	<6 h (two planes, each with three instruments)

, also developed under the IIP.

The AIRS and MISTiC requirements for temperature and water vapor sounding, while similar, are not identical. The difference in vertical coverage was noted above. A requirement area where MISTiC provides considerably higher performance than current IR sounders is in the area of spatial resolution and sampling density. MISTiC is optimized to provide accurate motion-vector wind observations, which requires high spectral resolution, high spatial resolution, and appropriate spatial sampling.

The vertically resolved winds observation uses a constellation approach, and therefore requires a system-of-systems requirements framework involving instrument requirements, spacecraft requirements, and ground-processing requirements. MISTiC Winds targets both cloud and moisture-field motion. Key requirements developed under the IIP that relate to the vertically resolved vector wind observations and thermodynamic vertical profile are shown in Table 2. The purpose of these requirements is to guide the design of both the instrument and the constellation. One of the key requirements, listed first in Table 2, is the wind speed uncertainty. This requirement is assumed to include errors due to height assignment in the presence of wind shear (up to 20 m/s, as indicated in the table) well as errors in wind tracer position change. The requirements for wind vector and thermodynamic vertical profile errors are assumed to be relative to the actual atmosphere, rather than to radiosondes. The feasibility of these requirements and connections to radiosonde observations will be discussed in Section 4.

2.2. MISTiC Instrument Design and Sensitivity Performance

The core of the instrument is a field-imaging dispersive infrared spectrometer, shown in cut-away in Figure 7. A small scan mirror directs scene radiation into a small imaging optical assembly that focuses the light on the entrance slit of the spectrometer. The spectrometer optics disperse the light, pass it through optical filters, and present it to a 640×480 format infrared focal plane array. This array is cooled by an active cooler to reduce noise and detector dark currents.

The spectrometer optical assembly’s largest dimension is approximately 20 cm. The overall instrument assembly has a volume of approximately 12 U (1 U = 10 × 10 × 10 cm) and is compatible with hosting by an ESPA-class micro-satellite spacecraft with a combined (payload + spacecraft) size of approximately one cubic foot, or about 33 × 33 × 33 cm (stowed for launch). Total instrument power requirements are approximately 50 W (continuously), and total instrument payload mass is less than 15 kg, consistent with a total spacecraft mass of less than 50 kg.

BAE’s detailed radiometric performance model, adapted from models proven on AIRS and updated for GOES-R HES, has been used to estimate noise-equivalent spectral radiance (NESR) performance of the MISTiC instrument, and compare it to requirements originally developed by NASA and NOAA for GOES-R HES. MISTiC sensitivity estimates show solid performance margin against these NESR requirements. These estimates are shown for each spectral sub-band in Figure 8.

In addition to radiometric sensitivity, a key requirement for an IR sounding instrument is the stability and knowledge of its spectral calibration. AIRS has provided outstanding spectral calibration stability, a primary goal for its thermal design. Microsatellites cannot supply large amounts of electrical power. In order to minimize on-orbit power demand by the MISTiC instrument, passive radiative cooling is used for its IR spectrometer, as was done for AIRS. However, in order to be efficient, the passive cryo-radiator must have a relatively clear view to deep space, and this is best accomplished in a sun-synchronous orbit, with the additional constraint that the host spacecraft not have any structure within the cryo-radiator field of view.

An additional challenge for MISTiC, and precision measurements generally aboard a micro-satellite, is that the thermal environment is quite variable over the orbit, yet the spacecraft and instrument resources available to manage these challenges are severely constrained. This challenge was assessed through thermal modeling under the IIP, and key results are summarized in Figure 9.

For MISTiC, the worst-case thermal condition is encountered by constellation members in the sun-synchronous 1:30 orbit. The spacecraft passes into and out of the earth’s shadow during the orbit. A small spacecraft lacks a large heat capacity, so its temperature-response to changing solar illumination can be quite pronounced. Special thermal design measures for MISTiC, including an adiabatic interface with the spacecraft, are employed to stabilize the spectrometer warm interface temperature, further stabilizing spectral calibration. With these measures in place, detailed thermal modeling shows that the spectrometer temperature variation over the orbit, presented in Figure 9, is very small, and consistent with MISTiC requirements for spectral calibration stability.

2.3. Constellation, and Launch Considerations

The MISTiC instrument payload has been specifically designed to be compatible with the ESPA-class microsatellites that have emerging over the last few years. The target payload mass of <15 kg, and target payload operating-power demand of <50 W are consistent with these spacecraft, even for configurations in which solar panels are not allowed to be in the view of one (the anti-sun) instrument face. Instrument physical dimensions, shown in Figure 3, together with host spacecraft are consistent with the size and configuration requirements of ESPA-class craft. The mass, power, and size accommodation needs have been verified through detailed physical layout, opto-mechanical design, and electronics design studies under the NASA Instrument Incubator Program.

Relative to most micro-satellite payloads, MISTiC Winds would provide a very rich set of environmental observations, but these bring a higher data rate requirement as well. In continuous operation, the instrument generates about 5 Mbits/sec, following lossless onboard data compression. One option for relaying the data to the ground would employ X-band RF transmission to two ground stations, one in each polar region, with a maximum data latency of ~50 min. A lower-latency approach would employ an S-band up-link to a GEO communications satellite such as NASA TDRS, followed by downlink. In light of the substantial commercial investment in small satellite LEO networks being developed by several companies, communication options for MISTiC Winds are expected to increase substantially in the next few years.

3. Results

The key technical risks related to providing the MISTiC Winds observation capability relate to providing accurate, precise hyper-spectral radiometry within the limited resources of a micro-satellite, with instrumentation based on technology suitable for the LEO space environment. Several steps have been taken under the effort reported here to reduce the technology risks associated with the MISTiC instrument payload. These include the fabrication of a flight-like or brass-board infrared imaging spectrometer optical assembly-including the reduction to practice of the innovative elements in its optical design, key performance tests of the spectrometer, and the high-energy proton total-dose testing of the selected infrared focal plane array detector. Key results from this instrument risk reduction work, as well as some instrument hardware adaptations needed for the airborne demonstration, are reported in Appendix A. In addition to the instrument hardware risks, the hyperspectral method for vertically resolved wind observation itself has had only limited demonstration prior to this work. To further reduce risks associated with the MISTiC Winds observation method, airborne observations using an instrument incorporating this spectrometer, and flown aboard a NASA ER2 were performed, with comparisons to independent observations of the wind, temperature, and moisture fields. Example results for these tests are provided below.

MISTiC Airborne Instrument Observations from the NASA ER2

Airborne versions hyperspectral IR sounding instruments have been used many times as part of validation campaigns for AIRS [26] and for other IR sounders. In anticipation of a NASA GIFTS Program, the NPOESS atmospheric sounder testbed instrument (NAST) was used to make hyperspectral observations from an ER2 during repeat-pass observations of a moisture field from which motion vector wind observations were derived [20], with a repeat-period of approximately 45 min. However, the specific use of repeat-pass flight trajectories with an air platform to simulate space-borne acquisition of a sequence of images from which to derive AMVs with image repeat periods of 10–20 min appropriate for changing cloud formations do not appear to have been reported previously. A key benefit of repeat-pass imagery on this time scale is that the primary changes between images are due to the translational motion of both clouds and moisture features as they move with the wind (advection), rather than due to changes in the spatial distribution (shape) of the features themselves. A second practical benefit is that the probability of observing an atmospheric feature on multiple passes from the relatively low vantage point of an aircraft platform is much higher if the time-separations used are not too large. Some atmospheric observations using this airborne instrument on the ER2 are described below. The path shown for this flight is shown in Figure 10. More details on the airborne observation demonstration flight on 4 December, including methods, weather conditions, and a comparison with an IASI satellite observation are provided in Appendix B.

An example of the types of airborne MISTiC observations made using the repeat-pass hyperspectral imaging/sounding method for cloud AMV identification, together with approximately co-located observations from a radiosonde launched from the Edwards AFB during the ER2 overflight on 4 December are shown in Figure 11, Figure 12, Figure 13, Figure 14 and Figure 15. In Figure 11, two MISTiC “window-band” radiance images taken at the latitude and longitude positions listed in the text in Figure 11 at times separated by approximately 1000 s. The images shown in Figure 11 have brightness temperature contrast selected to highlight the separation of the thermal emission of clouds, which is much lower than that from that of the surface. Common points in the cloud field in the two images were identified using a manual process for this airborne test. An involved but straight forward geometric construction is used to determine the absolute position of the common point of the selected cloud in each image, using the platform (3D) position, velocity, and heading, relative (angular) position of the cloud within each image, and airborne instrument design parameters (pixel size, focal length). The specific position and angle data for each image are listed in the image text at left. The AMV angle and speed were determined from the difference in absolution cloud positions, and the time at which the cloud was seen in each image. Additional information on this airborne observation approach is provided in Appendix B. One practical challenge in the airborne repeat-pass imaging is that at the timing of collection of each image was not synchronized with the platform position as it moved around the orbit several times, as shown in Figure 10. In this case, even a stationary feature appears in different relative positions in the two images. Another practical challenge is that the wind direction and speed were far from ideal for observing a feature twice. The cloud features were at the northern-most edge in the first image, and the southern-most edge of the second.

A cloud (circled in red the figure) was one of several common points identified and located in both of the two images (in this case, given the wind direction and speed, most of the clouds in the field are seen in only one of the two images.) As shown in Figure 10, the ER2 ground track is essentially the same for the two images, but the nadir points of the images are offset in the along-track direction.

A cloud motion wind vector is determined from the angular position of the cloud in each of the images along with the instrument’s GPS-position and IMU attitude data collected by the instrument during each overpass, and a rough height estimate. The upper two panels in Figure 12 show the wind speed and direction from the lower-altitude portion of the radiosonde profile collected during the MISTiC overpass period. The cloud-feature-based wind speed and direction obtained from the data in Figure 11 are indicated in red in Figure 12. These wind-speed and wind angle estimates agree to within 2.5 m/s and five degrees azimuth, if the cloud lies just below the pronounced temperature inversion near 2500 m (AGL) apparent in the lower-left panel of Figure 12. This height for the cloud is physically sensible, and also consistent with the cloud top radiance range (corrected for cloud reflectance at that wavelength) observed by MISTiC in this spectral window, as shown in Figure 13 and indicated by the red dashed arrows in temperature profile in Figure 12.

Figure 13 also shows spectra for areas both in and near this cloud. The image in this figure is a small portion of a single spectral channel image for a spectral channel near 5 μm, lying near one local minimum in spectral radiance, indicating somewhat stronger water vapor absorption and re-emission at a lower temperature for some areas surrounding the cloud. On physical grounds, a water vapor field surrounding this low-level cloud is expected. There is also evidence of a lower-level water vapor field in the radiosonde data, shown in the lower right quadrant of Figure 12. The vertical position of this water vapor field is consistent with being trapped by the same temperature inversion that trapped the clouds observed in Figure 12.

Images of the same scene, but for spectral channels near 4.894, 4.952, 5.017, and 5.079 μm, are shown in Figure 14. The infrared brightness for these channels is progressively lower, indicating that these channels are weighted higher in the atmosphere. The images show a broad region of higher water vapor concentration extending over much of the image, in addition to the moisture field around the lower-level clouds. Figure 15 shows infrared radiance spectra in the MISTiC water vapor band for regions both inside and outside of this moisture field, showing lower infrared radiances, especially for the radiance minima at longer wavelengths-features which indicate that this moisture field is higher in the atmosphere. Figure 15 shows the water vapor from the same Edwards 4 December sounding whose data were shown in Figure 12, but with an expanded altitude scale, together with the water vapor retrieval from the nearest cloud-free AIRS sounding less than one hour later.

These both show the presence of a higher-level water vapor field, near 560 hPa over the Edwards AFB area, consistent with what was observed in the MISTiC images shown in Figure 15.

As noted previously, due to the cost constraints under the IIP, the airborne version of the MISTiC instrument employed a previously produced version of the IRFPA, which does not have the full spectra range planned for the MISTiC spectrometer and its custom-grown HgCdTe composition for the detector array. Specifically, it lacks the full spectral channel set needed to fully retrieve the water vapor features at 500 hPa and at higher altitudes. However, as shown in Figure 4, the full MISTiC spectral channel set, with the planned lower wavenumber limit of 1750 cm⁻¹ includes the channels needed to retrieve both the temperature and water vapor through the troposphere, with layer accuracies comparable to AIRS, and would be expected to support water vapor retrievals similar to those provided from AIRS in Figure 16.

4. Discussion

4.1. Discussion of MISTiC Winds Requirements and Their Relation to Other Instrument Requirements

MISTiC Winds provides an opportunity to develop the requirements and design for a remote sensing capability that are fundamentally driven by the specific needs of the vertically resolved wind observation. This situation differs from the historical case with GOES or especially for AIRS, where the instrument requirements did not explicitly address the needs of wind observation, and where the capabilities of the resulting instrument were adapted to wind observations after the fact. Two questions that arise in the present case relate to requirements sufficiency and requirements feasibility. The driving requirements of wind speed accuracy are considered from both the sufficiency and feasibility perspectives.

The LAWS Program previously identified wind speed uncertainty requirements for a future Doppler Wind LIDAR as 2 m/s for the lower troposphere (0–3 km) and 3 m/s for the upper troposphere (3–15 km) [4]. These are also the values more recently identified by the World Meteorological Organization [27] (requirement IDs 383 and 385 in the OSCAR database) as breakthrough values for the horizontal wind uncertainties for the lower and upper troposphere, respectively, based on weather forecast impact for high-resolution regional and global NWP. More recently, the NASA Weather Focus Area Workshop in 2015 [28] identified a wind speed observation requirement of 2 m/s rms uncertainty for the remote sensing of vertically resolved wind field during its deliberations, as sufficient for advancements in weather remote sensing research and weather forecast improvement (although specific numerical requirements of this type were not included in the final report of this workshop). The wind speed uncertainty requirement adopted for MISTiC Winds, shown in Table 2, is informed by these requirements.

The MISTiC OSSE [23] has shown that vertically resolved observations provide significant weather forecast impact. OSSEs can provide constraints on the accuracy that must be provided in order for these new observations to have the desired impact on forecast accuracy. Within this MISTiC OSSE, the total wind speed uncertainty employed was approximately 5 m/s rms. However, this included both the observation errors and errors associated with the 3DVAR data assimilation, so the allowed contribution from observation errors would again be in the 2–3 m/s range, consistent with Table 2. An updated MISTiC OSSE may provide more specific constraints on observation error requirements. Finally, a <2 m/s wind speed observation uncertainty was reported to be an important aspect of more recent DWL OSSEs [19,29].

Requirements feasibility can be assessed by considering the implications of the requirements in light of physical or cost constraints. Under the MISTiC Winds IIP, an initial partitioning of the 2 m/s wind speed error requirement was performed, and this partitioning includes flow-down requirements for the error due to tracked feature height assignment (<1.5 m/s) and errors in horizontal feature tracking (<1 m/s) as leading terms in the error budget. These errors are resumed to be un-correlated.

The AMV height assignment error contribution to wind speed error is only significant in the presence of significant wind-shear. To support an initial assessment, a wind speed difference of 20 m/s, between the 850 and 200 hPa pressure levels is assumed as a reference case, with the further assumption of linear wind speed change with pressure height. Under these assumptions, a 30 hPa feature height assignment error (see Table 2) would yield a 1.5 m/s wind speed error. Infrared spectral sounding vertical resolution in the troposphere is approximately 100 hPa. The AIRS moisture retrieval shown in Figure 16 provides a good illustration of this, in comparison with the moisture vertical profile obtained from the near-by radiosonde shown in Figure 15. The radiosonde shows a sharply defined higher-concentration water vapor layer (in dew-point temperature) over a range of approximately 80 hPa in pressure height, centered near 560 hPa. The AIRS retrieved water vapor profile is somewhat broadened, but it is still sufficiently sharp so that the position of the peak water vapor concentration can reasonably be located to a precision of 30 hPa. The pressure height of the peak in water vapor concentration obtained in the AIRS retrieval is in good agreement with the radiosonde. The vertical resolution of water vapor retrievals obtainable from MISTiC is expected to be similar to those obtained by AIRS, as shown in by trade-study results in Figure 5. Full-scale development of this observing system would include a deeper exploration of this capability, for a variety of thermodynamic and wind speed profiles so that these requirements and feasibility assessments can be refined.

Although the above example discussed the feasibility of height assignment improvements for water vapor tracers, hyperspectral observation should improve cloud AMV height assignment also. It enables this improvement in two ways: By providing access to the observed local temperature vertical profile co-located with the AMV feature, and by providing access to highly transparent narrow spectral channels for the brightness temperature measurements that are more accurate than those using the broader multi-spectral imagers.

To ensure that feature tracking error portion of the total wind-speed error remains small (within a 1 m/s allocation), the current levels of image-sequence quality provided by GOES 16 ABI and its sister instruments on Himawari 8 and 9 are also needed for MISTiC. In this feature-tracking application, image quality is driven most strongly by two attributes-ground sampling distance (GSD) and the spatial oversampling ratio. The GOES ABI instruments provide ~2 km GSD at nadir for thermal IR bands, and approximately 3.5 km at edge of scan. MISTiC is designed to have a 1.4 km GSD cross-track at nadir, and also approximately 3.5 km at edge of its scan. GOES ABI requirements (as have those for all prior GOES instruments) also called for sufficient spatial over-sampling (i.e., the ratio of the spatial point spread function FWHM to the sampling distance). For feature-tracking through a sequence of images ( as is done for AMVs or for image loops from GOES), it is important that the tracked features do not change in apparent size, shape, or centroid, as the feature it moves from image to image, as such changes will cause a wind velocity observation errors. The MISTiC spatial over-sampling ratio is approximately 1.5:1 along-track-similar to GOES 16 ABI, and more than 2:1 in the along-scan direction. Feature position change-tracking to half the GSD, or finer, is enabled under such sampling conditions, and feature brightness fluctuations are minimized. In addition, MISTiC Winds observes feature position changes with two separate instruments on different platforms in order to measure feature velocity. However, the attitude knowledge provided by contemporary micro-satellite-compatible star trackers and ephemeris knowledge are sufficiently accurate, therefore, these potential error terms in AMV tracer velocity are quite small. These considerations, together with the radiometric sensitivity requirements identified in Table 1 and shown in Figure 8 should be sufficient to support the <1 m/s tracking error capability for both cloud field and water vapor tracer features, for the planned observation separation period of 10 min.

Optical Flow methods are under consideration by several groups, as a tracking method potentially capable of improvements in the density and accuracy of passive wind observations. However, large-displacement optical flow estimation algorithms assume a relatively high degree of grey value constancy, gradient constancy, and image smoothness in the image data from image to image. These image quality attributes need to be ensured for these wind retrieval methods as well, and in order to accomplish this, the spatial resolution and spatial over-sampling considerations described above for AMV tracking with MISTiC are essentially the same as those needed to support optical flow-based wind field retrieval methods.

Spatial resolving power is directly related to the potential density of AMVs that can be observed and tracked. Useful water vapor tracers may be as small as the spatial response function of the instrument. For MISTiC, the SRF FWHM is between one and two GSDs, and the entry in Table 2 relates to the number of SRF-scale targets that could be tracked. Cloud AMVs require more spatial information to ensure that the same cloud is being tracked between images, and so the potential density of cloud AMVs is much smaller, as shown in Table 2. The actual density of AMVs depends on the specific atmospheric conditions. In the future, Optical Flow methods may allow a higher density of tracking points in both cloud and clear conditions.

Validation of current AMV observation methods widely employs comparisons with radiosonde observations as “truth”, and validation of the MISTiC Winds observation method would employ this approach as well. However, there are additional error terms that arise in this comparison, that are due to potential errors in the radiosonde data, and to the displacements in either space or time between radiosonde observations. One important difference will be the simultaneous observation of the vertical thermodynamic field profiles both by the MISTiC Winds system, and independently by the radiosondes, which should improve the ability to understand the causes of differences between radiosonde and satellite-derived wind observations.

The GOES-ABI-like image sequence requirements adopted for MISTiC differ in important ways from the traditional way IR sounder spatial response attributes are specified. IR sounders are envisioned as providing spatially un-correlated, independent “soundings” of the atmosphere. Such sounding instruments are designed so that the individual ground footprints (and atmospheric columns) observed do not overlap. In some sounding instrument designs such as IASI, there is a substantial gap between spatial samples. In image-quality terms though, the images such instruments produce are significantly under-sampled spatially. While spatial under-sampling is fully acceptable for single-frame sounding observations, spatial under-sampling is not desirable at all for image-sequence tracking or optical flow estimation, due to the radiance and radiance gradient errors it introduces.

The finer spatial resolution of MISTiC relative to current hyperspectral sounders, although selected in order to support AMV wind observations, will also provide benefits for observing the spatial structure of the atmospheric fields. This capability will enable a larger fraction of cloud-free observations than current operational sounders observe, along with an improved capability to observe multi-level moisture field gradients.

The most significant spectral range difference between the requirements proposed for MISTiC and those of hyperspectral sounders currently in orbit is that MISTiC spectral range does not include the longwave infrared (LWIR) atmospheric window or very longwave (VLWIR) CO₂ emission-dominated regions of the infrared spectrum. If it did, the size, mass, and power accommodation needs of the instrument would approximately triple, requiring a substantially larger spacecraft host. The MISTiC spectral range includes 4.3 μm (2390 cm⁻¹) region channels, which are in the range used by the AIRS science team algorithms for retrieving the temperature vertical profile. The moisture-band channels used for MISTiC come from the shorter wavelength side of the atmospheric moisture band-as was planned for the NASA GIFTS instrument. These channels provide the same function for moisture sounding as those on the long wave side of the moisture band used by AIRS.

The functionally similar temperature-sounding spectral channel range near 15.4 μm (650 cm⁻¹) in AIRS have somewhat broader vertical weighting functions, but are valuable for sounding the stratosphere, and (when used together with the 4.3 μm region channels) for identification of partly cloudy footprints, and for the creation of cloud-cleared radiance products. For MISTiC, improved cloud observations, enabled by its 5x higher spatial resolution, will at least partially compensate for the lost ability to use spectral cloud-clearing methods with the VLWIR band. Moreover, by mission design—MISTiC requirements are selected to support tropospheric winds and sounding—not stratospheric sounding. Operational IR sounders will continue providing stratospheric soundings in the VLWIR band for decades to come, so MISTiC will not plan to replicate this capability. Finally, radiance forward-modeling challenges in the 4.3 μm region that limited its use by earlier sounder teams have been addressed within the contemporary forward model used by the AIRS science team.

4.2. Significance of MISTiC Winds for Weather Research

While tremendous advances have been made during the last two decades on space-based observations of the mass field with infrared and microwave sounders, the tropospheric wind vertical profile remains the number one un-met observation needed to advance weather forecasting [24]. MISTiC Winds will observe the vertical wind profile in the troposphere, as well as the moisture and temperature vertical profiles simultaneously, with unprecedented spatial coverage, spatial resolution, and temporal refresh rate. The most recent NAS Decadal Survey [3] (p. 11) identified atmospheric winds as a targeted observable, specifically citing the need for “3D Winds in the troposphere/PBL for transport of pollutants/carbon/aerosol and water vapor, wind energy, cloud dynamics and convection, and large scale circulation.” The MISTiC spectral range also includes channels that provide observation of nitrous oxide and carbon monoxide pollutants in the free troposphere, as well as some highly transparent (super-window) channels useful in observing aerosols. When combined with MISTiC ‘s vertically resolved wind observations, the atmospheric transport of these materials can also be observed. The planetary boundary layer (PBL) has also been identified in the Decadal Survey as a targeted observable. PBL observations are recognized as challenging, and are placed in an incubation stage by the NRC. However, hyperspectral infrared sounding observations are noted as among the essential techniques that will be needed for PBL studies. While the vertical resolution needed for PBL characterization is a challenge for traditional vertical profile retrieval approaches, differential approaches (on-line/off-line methods) together with infrared skin-temperature observations should be effective in important situations, including over-ocean observations, where the surface emissivity is well characterized [30]. The substantially higher spatial resolution of MISTiC observations relative to other infrared sounders would be particularly valuable in supporting studies of the lower troposphere and PBL in partly cloudy fields. The relatively higher frequency of observation, and potentially even the short-burst hyperspectral snap-shot sequence (10-min sampling) of the scene observed by each element in a wind triplet could provide valuable insight into PBL evolution.

The MISTiC Winds combination of observations is expected to be particularly advantageous for improved NWP. The differential equations solved numerically to provide a weather forecast constitute an initial-value problem, which requires initial conditions provided by atmospheric observations for solution. Assimilated MISTiC Winds observations would provide valuable constraints on both the initial atmospheric state, and its first time-derivative. The value of this approach for weather forecast accuracy improvement has been demonstrated through an OSSE performed by NASA’s Global Modeling and Assimilation Office evaluating the MISTiC observations on a global weather forecast [23]. This simulation showed significant potential for weather forecast improvement from the new MISTiC observations, with the vertically resolved wind observations being most significant, but with significant improvement also provided by assimilating IR radiances. Of particular note are the Forecast Sensitivity of Observation Impact (FSOI), where the MISTIC vertically resolved water vapor motion and cloud motion vector wind observations of a MISTiC constellation showed the largest impact on the 24-h forecast of any other observation. When Himawari-sized errors are included for the wind observations in the GMAO OSSE system’s 3D-Var assimilation, the MISTiC FSOI impact remains very high-comparable with the impact of the full microwave sounder constellation. An update to this MISTiC OSSE is currently under way at the Jet Propulsion Laboratory, using a more representative error model for hyperspectral AMV winds. These initial OSSEs were performed at GMAO using the 7 km GEOS5 Nature Run with a ¼ degree Forecast Model, a ½ degree Analysis model, and 3D-Var assimilation with Himawari-sized errors in wind speed. It is expected that the fine temporal and spatial scale observations MISTiC Winds will provide will have an even greater impact on forecast accuracy for the fine-grid convection-permitting NWP models being developed.

4.3. Differences in Observed Spectral Resolving Power for the Airborne and Laboratory cases

Spectral observations from the MISTiC airborne instrument during the ER2 flights demonstrated somewhat lower spectral contrast than expected from the instrument design, based on comparison with observations from IASI and based on analysis using radiance forward models informed by the atmospheric vertical temperature and moisture profiles provided by radiosondes. In contrast, the laboratory measurements of the brass-board spectrometer, described in Appendix A, show a spectral resolving power in very close agreement with design expectations. Additional information on this difference and its most likely cause are provided in Appendix B.

5. Conclusions

A new space-borne weather remote sensing approach, MISTiC Winds, employing miniature infrared spectral imaging sounders in a microsatellite-hosted constellation is described in this paper. This observation approach would provide individual high-resolution infrared spectral soundings in a dense array, with image quality characteristics of spatial sampling, spatial resolution, and sensitivity that enable high-quality atmospheric feature tracking. These attributes, together with the small size of these instruments enable a small number of them to be deployed in a common orbital plane in LEO short with time-separation, providing a feature-track wind observation, with the improved feature height assignment accuracy provided by hyperspectral infrared sounding. The vertically resolved tropospheric wind observations enabled by MISTiC Winds are directly responsive to the need identified by the National Academy in its 2018 Decadal Survey for Earth Sciences and identified as a targeted observable. Such observations would enable research for an improved understanding of key atmospheric processes, especially for severe weather, and improved numerical weather prediction.

This proposed observation builds on the experience of the NASA AIRS Program, as well as the requirements, design, and technology developments of the NASA GIFTS Program and the NASA/NOAA GOES-R HES Formulation Phase Program. The specific requirements, mission concept, and instrument design concept were developed under the NASA Instrument Incubator Program and BAE Systems internal funding. A program of critical risk reduction testing was also completed. Specific accomplishments under these efforts include:

Development of key instrument and mission requirements, observing system architecture and instrument detailed concept design tailored to focus on dynamic weather characterization in the troposphere with miniature instruments in a constellation:

Cloud and water vapor motion vector winds with rms wind-speed errors <2 m/s at 6+ levels in the troposphere,

High-resolution IR soundings of temperature and vertically resolved moisture gradients in the troposphere,

15 kg/50 W instrument is 60x smaller in volume, 10x lower mass, and 4x lower power demand than NASA’s AIRS instrument,

Performance of critical technology risk reduction through laboratory and airborne testing:

APD-mode 640x480 IRFPA with Proton total dose tolerance for four-year LEO mission life demonstrated,

Ultra-low distortion brass-board IR spectrometer assembly demonstrates >700:1 spectral resolving power and high spectral calibration temperature stability (dλ/dT <2.5% of spectral response function FWHM per 100 mK),

Conducted several airborne observation demonstration flights on ER2 (~30 h) multi-pass observations over land sites in southern California, and over the nearby ocean,

Analyzed the airborne data to show vertically resolved moisture gradients, and hyperspectral cloud AMVs matching radiosonde wind speed observations to within 2.5 m/s under challenging high wind-shear conditions.