Case Study: EnMAP
When EnMAP was envisioned by the German Aerospace Center in 2006 its aim was spectacular. It was planned to bring Germany on the forefront of hyper-spectral imaging. Today more than a decade later the satellite is not yet launched. Why? Let's have a look!
Introduction
Before looking at the EnMAP satellite we will establish what hyper spectral instruments are, why they are used, how they are implemented and what challenges they pose.
What is hyperspectral?
In imaging payloads there is usually a distinction by how many channels they have. A single channel instrument is called panchromatic, an camera containing multiple color channels is called multi spectral. Most space payloads have one panchromatic high resolution channel and 4-5 lower resolution multi spectral channels.
In addition, there is a different class of instrument, the hyper spectral. In this case the instrument is not primarily defined by its resolution but the number of spectral channels. Typically any instrument with more than 30 channels is called hyper spectral in the case of EnMAP this can reach more than 250 channels.
Why hyperspectral [3]?
The electromagnetic spectrum is continuous. All objects, including plants and minerals have certain specific spectral fingerprints. With the 4-5 channels of a multi spectral instruments it is often impossible to differentiate between those.
The abundance of channels across a wide range enables a hyper spectral instrument to do new science and applications
With the 260 channels of EnMAP ranging from 400-2450nm it is possible to cater to a host of different applications, from mineral mapping, plant identification (e.g. illegal crops), precision agriculture & forestry, soil management to water quality.
How to do it?
A hyper spectral instrument consists in general of the following main elements: an optic, the dispersive element and the detector.
Optic: the optic could be a diffractive elment (glass lens) or a reflective element (mirror) or any combination of the two. Since glass elements show residual chromatic errors which are especially problematic in an instrument that focuses on spectral quality, most imaging optics for high performance hyper spectral payloads are made of mirrors. To correct the most prominent optical aberrations usually a three mirror anastigmatic design (TMA) is chosen.
Dispersing Element: the dispersing element is usually either a refractive prism (glass element) or a grating (diffractive element). As with the imaging optics all glass elements introduce color aberrations and are usually avoided for modern designs.
Alternative Designs: beyond this standard implementation there have been a number of alternative designs being proposed in the last few years to implement hyper spectral instruments. These usually forgo the separate dispersive element and the collimation optics and replace it with a more simple lens, filter, sensor design. The filter can either be placed on the sensor itself or using a filter glass in front.
Main difference to slit based hyper spectral instruments is that the spectral domain for one location not collected all at once but usually at different times which means that moving elements will have spectral shadows.
Challenges?
There are several challenges involved when implementing hyper spectral instruments.
1) There is is never enough light.
The reason for that is that the width of the spectral band for each channel is only about 5-10% of a regular multi spectral and only 1% of a pan chromatic channel. In addition, in sensor signal accumulation (TDI) does not work due to how the spectrum is generated. As a result hyper spectral instruments are usually of much lower resolution that comparable multi spectral instruments. In other words, at a given aperture the achieved resolution is typically by a factor of 3-10 lower.
2) There is always too much data
Due to the increased number of channels the raw data of the device scales with the number of channels. Therefore, a hyper spectral instrument with 300 channels has 75x the data of an instrument with just RGB & NIR of the same resolution. As a result and despite the low resolution of 20-30m both the image processing chain as well as the data transmission often rival systems with very high resolution (<1m).
3) System Complexity
In comparison to a normal panchromatic / multi spectral instrument the hyper spectral instrument is with few exceptions (e.g. wedge filter) more complex. In addition to the imaging optics a spectroscope optics is attached. These components need to be aligned and maintain there alignment over the entire life time in orbit. In the case of EnMAP complexity is further added by the wide spectral range from UV to SWIR. This complexity shows; a multi spectral (RGB NIR) instrument of similar resolution (30m) and swath (30km) could easily fit inside 1U restriction (1 liter volume, 1kg). The EnMAP payload on the other hand fills the 1.5m3 volume of the LEO-SAT 1000 payload.
In the case of ENMAP these requirements made even the spacious payload bay of the LEOS-SAT 1000 look small. Indeed one of the problems during the EnMAP design was that the satellite was constantly on the verge of busting its mass budget which was capped at 1000kg both due to the platform as well as due to the chosen rocket.
One Giant Leap
EnMAP [1],[2] was envisioned by the German Aerospace Centre as Germany's calling card when it comes to hyper spectral. As one can see from the illustration below its capabilities were targeting a new quantity and quality of spectral information.
Unfortunately, in a very German fashion the EnMAP was aiming for nothing short of perfection. The most hyper spectral channels across the widest spectral range and high resolution. Several leaps of technology were required to implement these specs and at several junctions during the development of the platform the engineers were facing insurmountable difficulties.
Prime
One of the aims of the EnMAP platform was to bring a new mission prime to the German space ecosystem. The Munich based Kayser Threde (KT) [18] wanted to build beyond their traditional strength of developing state of the art payloads and establish itself as the third mission prime after Airbus and OHB. All they needed was a stage to prove what they could do. The EnMAP mission was slated to be that stage. Unfortunately KT's agressive growth both in quality and quantity was overstretching the company and to avoid bankruptcy the company was sold to OHB system for the modest sum of 6 MEUR [17] plus credit guarantees for the large new buildings that KT had bought prior. Considering that KT at this time had 200 employees and a turnover of 38 MEUR annually this seemed to be a firesale similar to what happened to RapidEye a few years later
Platform [1]
The platform for EnMAP was chosen to be the "LEO-Bus 1000" [1] with heritage from SAR-Lupe [20] mission built by OHB. The bus can support missions with up to 1000kg mass and 370kg of payload.
Payload [1]
The payload of EnMAP is a marvel of engineering. Unfortunately, it also is a beast of a machine and it took some time to tame it. Envisioned as a 4 year development project the payload was one of the main reasons why EnMAP was 10 years late.
The optical design of the telescope is that of a three mirror anastigmat (TMA) with 500mm focal length. Behind the slit there is a beam splitter to separate the VNIR and the SWIR channel. The spectrometer is a modified Offner type whereas instead of a grating a prism was used. Several patents have been applied for the payload.
The cost of perfection
First things first, I am not sure how much exactly EnMAP has cost up to now. I am still working on this number and I would be grateful for any little bird who could help me.
EnMAP started as a mission for 90MEUR and 4 years. Now 10 years delayed it has likely cost more than 300MEUR.
That said, EnMAP started out as 90 MEUR contract in 2008. Since the project was not as originally intended launched in 2012 on PSLV but is now planned to be launched in 2022 on Falcon 9 [2] I think it is fair to assume that this project until now has consumed 3x the original budget. My personal bet is somewhere in the range of 300 MEUR.
Secondary Effects
It is important not to forget the secondary effects of delay by perfectionism.
Canceled Projects: since budgets are limited it means that if one mission takes longer others can not be done. For example the premature death of DEOS can in part be attributed to the delay and cost overruns of missions like EnMAP.
Reduced Innovation: if you asked anyone in Germany that proposed a new mission to DLR in the last 10 years, you always heard one thing: there is no money. If you asked where the money went the answer very likely involved EnMAP.
Reduced Science: the researchers including the PI at GFZ Potsdam are waiting for the data since 2012. Some of the careers including the one of the first EnMAP PI (Prof. Kaufmann) have already ended in retirement in 2014 [21] and his successor (Luis Guanter) has left the GFZ to pursue a Professorship in Barcelona in 2019 [22] all before the launch of EnMAP!
领英推荐
Time moves on
In 2006 when EnMAP envisioned only few hyper spectral missions existed but during the 14 year time it was built the world moved on. What can be observed is that today, using advances in technology, a satellite 1/10 the size of EnMAP can achieve similar specs.
Also while the time from 2010-2020 can be considered a time to prototype, the technology has now arrived in the era of commercial utilization.
EnMAP was designed as cutting edge hyper spectral mission. When it will finally launch it will meet a world in which commercial hyper spectral missions are the new normal.
Companies like Northstar , Satellogic , and Scanworld have announced plans to build constellations with tens of small satellites with hyper spectral payloads.
What EnMAP offers would have been great in 2012 but is now lagging severely behind. Indeed, if it would have been available as per original schedule it would have brought Germany into throwing distance of building a new business case. Unfortunately, due to the delay it is now other nations that take the lead.
What would I have made different
Innovation is faster in iterations than in one giant leap
My general recommendation would be to not do programs that require a giant leap - at the very least if you have applications depending on them. Let us imagine that EnMAP in an alternative world would have been built in a less sexy way that would have allowed to conclude the mission as planned in 2012 then for a decade earlier we would have had data.
A less perfect EnMAP could have been up in 2012 and we would have data since then.
Based on the lessons learned we could very likely have expedited the launch of an Enmap 2 to 2015 and Enmap 3 in 2018. With all that heritage and experience our hypothetical Enmap 4 to be launched in 2021 would very likely be more performant than what we will have now. I am also willing to wager that these smaller jumps would have cost us for four satellites that we now paid for one. In addition, this strategy would have enabled commercialization to a much greater extend than what EnMAP will.
Instead of one giant leap we should have build one EnMAP satellite every 2-4 years.
If you still want perfect, set aside some funds for quick, too
In addition, my recommendation if indeed perfect is what the mission has to be then we should strife to set aside a small portion of the budget for something at reduced capacity that could be build and brought into operation fast.
This approach at the very least would have provided the decision makers with better options when the main project should get delayed. Since an initial capacity is in place it is much easier to make the choice to kill the main project and start from scratch instead of throwing good money after bad.
The original EnMAP should have been cancelled long since. That there were no alternatives made this decision difficult.
Both multiple simple EnMAPs (SE) as well as quick EnMAP (QE) strategy would have enabled Germany to participate in the benefits of more modern technology: better optics, better filters, better detectors all at lower cost.
To set aside a small budget for fast implementation or risk mitigation is highly recommended to avoid future problems
It is therefore my suggestion to set aside a small budget inside every large project to achieve some capacity quick or to test and introduce new technologies instead of one giant leap.
How can you help:
This text is part of a?series ?of articles in which the author sets the framework to start a discussion about the wrongs of the space industry. If you have experienced similar things, leave a comment. Other views and opinions are very welcome, too, as they may present a way forward. Please be kind to each other.
Disclaimer
The author’s views are his own do not represent the views of Berlin Space Technologies.
Sources
[2] EnMAP at DLR
[3] DESIS on Muses - hyper spectral payload on ISS from EOportal
[4] ASI Prisma - Hyperspectral mission from Italy from EOportal
[5] ELOIS Payload flyer - from AMOS website 06.11.2021
[9] CHIME cost - 450MEUR for 2 satellites
[10] Hysis (India) - EOportal
[11] Hysis (India) - Wikipedia
[12] GomX-4 Eoportal
[13] Hyperscout 2
[18] Kayser Threde Wiki
[19] SarLupe Wiki
[20] SarLupe EOPortal
[21] GFZ News Prof. Kaufmann retires as EnMAP principal investigator in 2014 his successor is Prof. Luis Guanter
[22] GFZ News Prof. Luis Guanter is succeeded by Sabine Chabrillat in 2019
[23] Svalbird proposal for a low cost hyper spectral mission based on TUBSAT C bus
Excellent primer on hyperspectral and the dangers of over engineering
Interdisciplinary Optical System Engineer
3 年Hey Tom, nice article. Just a small correction, the table comparing the hyperspectral payloads has inaccurate spectral ranges. Enmap, Prisma, Elois, Chime all target up to 2400+ nm and not ~1000nm as you mention. Thanks for the whole series, a big fan!