SCOPE 13 - The Global Carbon Cycle

ABSTRACT

Present spacecraft technology makes it possible to monitor several environmental parameters, which can be divided into two major groups: the first includes input data, such as temperature, as a means for locating high productive regions in the oceans; and the second provides direct data on conditions and/or signals from targets which can be linked to the cycling of carbon. Although many of the launched systems were conducted as experimental programmes, they have already indicated a potential use in mapping various phenomena on a global scale. The present report summarizes, in the form of maps, tables, and satellite scanning pictures, relevant data on the use of spacecraft technology for assessing the global cycle of carbon in time and in space.

14.1 INTRODUCTION

The priorities of certain experiments performed on spacecraft have been based on mission objectives; for instance, the Earth Resources Technology Satellite (ERTS) Programme (or Landsat) was established for the purpose of monitoring resources with a scanning system on a global scale. Limited by the so-called atmospheric windows (see Figure 14.1), remote sensing, especially from spacecraft altitudes, is in several cases a qualitative tool which has become useful mainly for inventories and for assessment purposes.

Remote sensing devices must consider atmospheric windows, together with the spectral characteristics of the `targets' to be investigated. Therefore, isolated spectral bands must be used for the proper monitoring process which is applied. Table 14.1 summarizes the application of the most common remote sensors used in earth-oriented investigations and also includes some of the equipment installed on satellites. Of particular interest are those instruments in the infra-red region of the electromagnetic spectrum and the multispectral system in the visible.

Remote sensing devices used in monitoring environmental data have been carried, however, mostly on aircraft. Besides being more economical, aircraft have the advantage of gaining better resolution than space-borne instruments. Satellites, on the other hand, provide a more repetitive and global coverage and can be, in many cases, a low cost effort considering the amount of data flow.

Figure 14.1 The electromagnetic spectrum and the atmospheric windows

The scanning of the field of view is performed through a scanning mirror by an angular displacement of ±2.89^°, with an instantaneous field of view on the ground of about 80 m², although the ability to distinguish adjacent objects depends mainly on their contrast. The four spectral bands of the MSS have the following band-widths:

Channel 5: 0.6 to 0.7 µm 

Channel 6: 0.7 to 0.8 µm 

Channel 7: 0.8 to 1.1 µm

In the most commonly used convention, the black-and-white image from the first band is illuminated with blue light on to a colour negative, the image from the second band with green light, and the image from the fourth band with red light (the third band is not used). Thus, the infra-red reflection is displayed as red, the red is displayed as green, and the green is displayed as blue. As a result, in a false colour composite, green vegetation, which is highly reflective in the near-infra-red, appears as various shades of red; clear deep water is black; cities are identified by a blue-grey colour; and water with sediments appears as a bright blue. Rocks and soils usually range from browns, yellows, and tans to bluish groups.

A demonstrative sample of the effect of the colour assignment is shown in Figure 14.2, where an arid region is compared with a region in the higher latitude as a false colour composite. By simple image interpretation, one is able to differentiate between vegetated areas and deserts, as well as between different surface deposits. A more detailed analysis by computer makes it also possible to separate the different types of vegetation, as will be described in a later section. Changes in vegetation can be followed by repetitive coverage of the same region and, consequently, a semi-quantitative estimate of the fluxes of carbon can be made.

Fully automated methods employ satellite data on magnetic tape, where each ERTS pixle carries data on the intensity of the radiance from one ground element.

Table 14.1 Application of remote sensors for selected disciplines

Figure 14.2 Comparison between desert and vegated surfaces as monitored by Landsat. For explanation see text

Table 14.2 International participation in Landsat-1 investigations (after Short et al., 1976)

Each pixle is identified on the tape by a triple message, indicating its x and y coordinates and light intensity level. These data are processed by a computer, through special software algorithms; the output is printed automatically in most cases, together with ground observations.

Most applications of satellite-derived data in earth-oriented investigations have been carried out in the United States and other countries are now sharing a growing interest participating in NASA (National Aeronautics and Space Administration) Programmes, as indicated in Table 14.2.

14.2 THE LANDSAT SYSTEM

The Landsat-1 spacecraft was launched on 23 July 1972, and was positioned into a nearly polar and sunsynchronous orbit, meaning that the orbit plane moves around the Earth at the same angular rate as the Earth moves around the Sun.

The first Landsat mission demonstrated the feasibility of multispectral remote sensing from space for practical earth resource management applications. The overall system aimed at the acquisition of multispectral images, the collection of data from remotely located ground platforms, and the production of photographic and digital data in types and formats most helpful to potential users. In addition, it required that the data be taken in a specific manner: namely, that repetitive observations be made at corresponding local times; that the images be overlapping, both in and across the direction of flight; and that the images be correctly located to better than 3.7 km. Periodic coverage of each area occurred at least every three weeks. Landsat carries two major instruments: the Return Beam Vidicon Camera (RBV) and the Multispectral Scanner System (MSS). The MSS was applied on a quasi-operational basis and fulfilled most of the requirements of the investigators during the programme. Therefore, in the following discussion, only data from the MSS will be considered.

The Landsat system, which makes it possible to map and to identify surface features and resources of the earth on a large scale, can be seen as a potential means of assessing the pools and fluxes of carbon. So far, no global inventory of carbon has been undertaken but the scientific base established by Landsat-1, as well as the international interest and participation in this programme, provide a foundation for a future application of spacecraft technology in estimating the carbon pools on a global scale.

The use of Landsat data for inventories is based on the different albedos of various surfaces. As shown in Table 14.3, the albedo differs significantly between targets such as soil and vegetation. In many cases, an additional separation of signals received from different species allows their identification and quantitative detection (see Figure 14.3). The Sun's elevation must also be considered, particularly for investigations of watercovered surfaces. Generally speaking, the higher the elevation, the lower the percentage of reflected light intensity.

14.3 CONTINENTAL VEGETATION

Table 14.3 Albedo of various surfaces expressed as the integral over the visible spectrum (after Schanda, 1976)

Surface		Percent of reflected light intensity;
General albedo of the earth
	total spectrum	~	35
	visible spectrum	~	39
Snow, fresh fallen			7590
Snow, old			4570
Sand, `white'			3540 (increasing towards red)
Soil, light (deserts)			2530 (increasing towards red)
Soil, dark (arable)			515 (increasing towards red)
Grass fields			530 (peaked at green)
Crops, green			515 (peaked at green)
Forest			510 (peaked at green)
Limestone		~	36
Granite		~	31
Volcano lava (Aetna)		~	16
Water: Sun's elevation (degrees)
	90		2
	60		2.2
	30		6
	20		13.4
	10		35.8
	5	~	60
	<3	>	90

Figure 14.3 Spectral reflectivity from forest cover, soils and other formations (after Krinov, 1953).

There is a need for remote sensing, especially for those regions where no data is yet available or where inventories and forecasting of crops are inadequate (see Figure 14.5). So far, inventories and/or crop forecasting have been established mainly for those crops which are of global and economic importance.

Figure 14.4 Step-by-step identification of agriculatural crops and products, as used in the region of Ohio (after Castruccio and Loats, 1974)

The potential of taking inventory of vegetative areas can be demonstrated by the ongoing Large Area Crop Inventory Experiment (LACIE) (NASA Landsat Newsletter No. 13), which is a joint project of the National Aeronautics and Space Administration (NASA), the National Oceanic and Atmospheric Administration (NOAA), and the United States Department of Agriculture (USDA), involving the testing of satellite techniques for surveying and forecasting wheat production. LACIE is showing promising results in predicting wheat production with a 90% accuracy, 90% of the time.

Plant biomass and its changes,could also be estimated from spacecraft data; however, the determination of fluxes would necessitate ground data as well.

14.3.1 Crop Discrimination and Crop Yield

The economic importance of crop yield and crop distribution has lead to many investigations in which aircraft and satellite data have been applied. Discrimination between small-grain crops as a whole and other crops from satellite altitudes is quite good, while differentiation between different types of small grains (for example, wheat from barley) is somewhat less precise. Satellite images, currently produced at scale in the area of 1:1 000 000 to 1:250 000, provide less detail than those obtained from aircraft. However, this is offset by the increased sophistication of the sensory equipment and data processing techniques. Table 14.4 gives the accuracy with which different crops and vegetative areas have been differentiated by remote sensing. For satellites such as ERTS, the errors are already very small. In addition, if needed, ERTS pictorial data can be procured already rectified, and at nominal cost. Measurement of crop area from remotely sensed data can be subject to error. Results of actual mensuration of crop area by various ERTS investigators, however, have shown that these errors can be lower than 1%, and it is expected that errors can be reduced or altogether eliminated with the development of improved processing techniques (see also Table 14.5).

Figure 14.5 Distributed of sophisticated agricultural data gathering system

Table 14.4 Accuracy (% correct identification) in discriminating remotely sensed crops and other vegetation (after Castruccio and Loats, 1974)

	Crop	Aircraft	Satellite
		(Scale 1:4000)	(ERTS)
	Small grains	100
	Row crops	96	96
	Pasture	96	84
	Trees	100	86
	Wheat	95	93
	Oats	95	85
	Water	100	100
	not available.

14.3.2 Forest Classes

Computer classification of ERTS data has led to a successful differentiation of agricultural cropland types (Bauer and Cipra, 1973; Baumgardner et al., 1973) and broad natural area categories (Kirvida and Johnson, 1973; Thomson and Roller, 1973). Lawrence and Herzog (1975) used ERTS data for geology and forestry classification and were able to classify their test site into basalt, rhyolite obsidian, pumice flats, Newberry pumice, ponderosa pine, lodgepole pine, and water. The area around Mt. Washington (Oregon, U.S.) was classified into two basalt, three forest, two clearcut, burn, snow and water classes. Lawrence and Herzog (1975) also demonstrated that both stand density and tree species have significant spectral signatures. It was also found that, in one test site, the distinction between the ponderosa pine and lodgepole pine may be controlled by stand density.

14.3.3 Tidal Marsh Classes

Quantitative data on primary production in tidal marsh ecosystems were derived from single-image photographs by Reinold et al. (1973). Based on aerial measurements, the agronomic production on a 46-ha area of Spartina marsh was estimated to be 571 g/m² year, and 698 g/m² year on a 16-ha site of Juncus, Distichlis, and Spartina. The average production for the 1142-ha Duplin Estuary marsh is 591 g/m² year. It is evidently possible to increase the accuracy of primary production estimates of large marsh areas by employing remote sensing techniques. The interpretations of physiographic features in images, the creation of vegetation maps, and the measurement of primary production over large areas of tidal marshes are already proving to be useful. Primary production estimates for large areas, obtained by coupling ground truth measurements and photographic and non-photographic images, provide essential basic data for mathematical modelling of the salt marsh ecosystem.

Table 14.5 Condition/yield measurements of crops by remote sensing (after Castruccio and Loats, 1974) 

14.3.4 Burn and Clearcut Classes

Lawrence and Herzog (1975) selected training sets for investigating clearcuts and separated them into younger and older clearcuts, on the basis of ground observations. They demonstrated that the resulting classes are found to be accurate within approximately 10% (Table 14.6) but cautioned that the areas of the individual clearcuts measured are quite small and the measurements can, therefore, be subject to considerable internal error. Since both burn and clearcut establish signatures for grass, brush, and other forms of low vegetation, they classify more than the specific features of interest. From a thorough interpretation of these low vegetation classes, many important features can be distinguished.

14.4 LAKES

Landsat data has been used in several studies in making inventories of ponds, streams, and lakes, resulting in thematic maps and statistics relating to open surface water (Work and Gilmer, 1976) and to the identification of algal blooms (Strong, 1976). Work and Gilmer used two different recognition techniques. The single waveband technique proved too simple to implement and accurately recognize prairie lakes and large ponds. Using the high information content of multiple spectral bands, it was possible to recognize a greater number of small ponds and to improve the apparent spatial resolution of the data by a factor of three.

Table 14.6 Quantitative comparison of burn and clearcut classes (after Lawrence and Herzog, 1975)

Figure 14.6 Relationship between Landsat data in channel 6 (0.70.8µm) and chloropyll concentration (after kosche, 1977)

Reflected solar radiation can be used to determine the concentration of suspended sediments and, indirectly, by relating it to chlorophyll concentration, the productivity of surface waters can be estimated. Ritchie et al. (1976) showed with linear regression analysis the relationship between reflected solar radiation and suspended sediment concentration. The best spectral region for using this quantitative relationship would be between 0.7 and 0.8 µm. The actual concentration of chlorophyll in lakes and the conversion of reflectance into chlorophyll units can be highly correlated. As an example, Figure 14.6 shows the relation between Landsat data in channel 6 (0.70.8 µm) and the actual chlorophyll concentrations.

14.5 MARINE CARBON POOLS ACCESSIBLE TO REMOTE SENSING 

14.5.1 Upwelling Regions

Due to the dynamics of upwelling, it is difficult, in several regions, to estimate primary productivity by ship observations alone. The Somali Current is a good case study based on ship observations, from which comparisons were made with the simultaneous development of the southwest component of the monsoon wind (Figure 14.7). The results show that each year the temperature gradients, during the early formation stage, are consistently in direct proportion to the wind speed. The phase lag between the development of wind and temperature gradient is surprisingly short during the development of the boundary current, ranging between 3 and 10 days. During the decay period in the late summer and autumn, the temperature gradients lag 14 to 40 days behind the wind. The fast onset of upwelling may similarly result in an increase in primary productivity. Therefore, the knowledge of temperature gradient anomalies is essential to make a better assessment of standing crop and production.

Figure 14.7 Development of the baroclinic structure of the Somali Current in response to the southwest monsoon. The data allows for a detailed assessment of the onset of upwelling (after Szekielda, 1976)

Figure 14.8 15-day average map composed of three 5-day composites (after Salomonson, personal communication)

The infra-red region between 10 and 11 µm has the highest transmittance of the atmosphere and can be used to measure sea-surface temperature making it possible, therefore, to estimate productive upwelling regions in the ocean. Figure 14.8 is a composite of 15-day average data, collected with the Nimbus High Resolution Infra-red Radiometer (HRIR). Temperature gradients, which are related to, or are a function of the major current systems are as well established as the generally used historical temperature maps, although the sampling density, especially in the high latitudes, is much better from spacecraft altitudes than from ship observations. This analysis of spacecraft data is of particular interest for studying large-scale phenomena, but it can be used only in identifying larger temperature fluctuations and gradients. For a more detailed analysis of small-scale phenomena, such as coastal upwelling, a single orbit analysis is preferable, if cloud-free conditions appear.

Table 14.7 Satellite derived sea surface temperature as established by the U.S. Department of Commerce, NOAA, National Environmental Satellite Service (NESS)

	Mean difference (°C)	Mean
Date	(Satellite-Navy SST)	S.D. (°C)	Explanation
December 1972	-0.4
January 1973	0.8		Bias in sensor calibration using
	0.1		new empircial table for
			comparison
March 1973	1.25		Warming trend in first half of
	0.35		month after radiometer
			change
April 1973	0.25		Dropped during last two weeks
	0.00		due to changing climatic
			conditions during spring
			season
May 1973	0.38		Moisture coefficient corrected
	0.50		in moist, hot air difficulty
June 1973	0.78	1.38	calibrating the sensor
July 1973	1.6		Mid-July over warm water
			regions (>25 °C) in northern
			hemisphere weakness in
			temp. retrieval and atmos.
			moisture correction processes
August 1973	near 0	1.25
September 1973	0.33	1.18
October 1973	0.21	1.13
November 1973	0.09	1.15
December 1973	-0.43	1.24
January 1974	0.30	1.17
February 1974	0.55	1.14
March 1974	0.32	1.18
April 1974	1.40		Warming increase during latter
			half of month due to calibration
			error in SR-IR data

Several methods are now applied operationally and have shown an accuracy of about 1.2 ^°C standard deviation as compared to ship data (see also Table 14.7).

Of particular importance, besides the fast changes, is heterogeneity in the distribution of parameters in upwelling regions, which may be overlooked by conventional ship observations and may lead to distorted estimates in primary production. Figure 14.9 shows the patchiness in temperature distribution along the coast of California and, in addition to being a more actual representation, may result in better estimates in primary production. Temperature and concentration of plankton are highly correlated in upwelling regions. Therefore, regions of patchiness as observed from spacecraft altitudes can be interpreted in terms of plankton or biomass. Such observations indicate that a relationship exists between the reflected energy from the ocean and the biomass, which can be used to estimate the standing stock of near coastal regions. This first crude result can certainly be improved to show details, as soon as the better sensors used for coastal investigations are put into orbit.

Figure 14.9 Upwelling along the California Coast based on NOAA-4 data from 31 July 1977

Figure 14.10 Airborne `chlorophyll' and temperature recordings perpendicular to the coast of Cape Juby

Table 14.8 Statistical evaluation of correlation between primary productivity data

Equation				r2	r
	1 m
log	Chl = 0.877 log 1 m Chl + 2.364			0.432	1.398
	20 m
log primary productivity (1 m)
	= 0.715 log Chl (1 m) + 1.054			0.6575	0.029
			1 m
log primary productivity			= 0.684 x	0.666	-0.001
			20 m
		1 m
	log Chl	+ 1.444
		20 m
	1 m
log	primary productivity = 0.931 x			0.199	0.002
	20 m
	log primary prod. at 1 m + 2.078

14.5.2 Estuaries

Apart from using particulate matter as a tracer for deriving near-coastal circulation, the latter can be determined semi-quantitatively if ground truth can be provided by Secchi disc measurements and by determination of suspended material. Figure 14.11 is an example of a turbidity map which has been produced from categorized ERTS (Landsat) images. Saginaw River shows high concentrations of suspended material (low Secchi disc readings) northward along the eastern shore of the bay, while the less turbid water of Lake Huron is mainly located along the western shoreline. Similar analyses can be established by generating maps with information on concentration of particulate matter.

14.5.3 Sea Ice

Because of the variable gas content in the different samples of ice (Table 14.9) and in sea-water, sea ice in its various stages should also be examined when considering the global cycling of carbon. The values given in Table 14.9, taken together with the fact that 9 x 10¹² t of ice variations for the Arctic and 19 x 10¹² t for the Antarctic exist per year, show the importance of monitoring sea ice fluctuations for investigations of mass balances.

Infra-red studies were undertaken to map Arctic sea ice, and improved technology for ice mapping using infra-red data monitored by the Nimbus 2 and 3 HRIR was also developed. Sea ice may be recognized in the visible, infra-red, and microwave regions of the electromagnetic spectrum but, in the highest latitudes, the application of data in the visible is restricted to the summer seasons.

Figure 14.11 Turbidity map of Saginaw Bay, Michigan (Lake Huron). Categorized from Landsat coverage of 3 June 1974 (Rogers et al., 1975)

The lack of regular oceanographic observations in polar regions has resulted in only a relatively small amount of data when compared with other oceanic areas. Oceanic processes in the Antarctic, particularly the vertical motions, are not documented in detail, even though the Southern Ocean is one of the main sources of the bottom water in all oceans. Therefore, for mass-balance. studies of sea ice, it is of interest that one can differentiate, by remote sensing with microwave: (i) the age of ice by its different brightness temperatures, and (ii) seasonal fluctuations (Figures 14.12 and 14.13).

Table 14.9 Gas composition and concentration in ice samples (Matsuo and Miyake, 1966)

		Amount
	Density	of gas	Pressure	Gas composition
	of ice,	cm 3/kg	of gas
Ice sample	g/cm3	(STP)	bar	N2	O2	Ar	CO2
Iceberg ice
1	0.906	25.2	1.9	77.7	21.3	0.94	0.084
2	0.880	51.9	1.1	77.9	21.2	0.93	0.031
3	0.894	29.2	1.0	78.0	21.0	0.93	0.098
4	0.877	34.6	0.7	78.3	20.8	0.93	0.04
5	0.885	75.4	1.8	78.1	20.9	0.93	0.10
6	0.891	67.1	2.2	78.3	20.7	0.93	0.08
7		40.8		77.9	21.2	0.93	0.030
Glacier ice
1	0.886	36.0	0.9	78.3	20.7	0.93	0.028
2	0.913	15.8	3.3	73.2	25.3	1.22	0.29
Sea ice
1	0.909	8.9	0.9	76.8	21.3	0.93	0.98
2	0.915	2.2	0.9	54.2	20.6	0.99	24.3
3		21.2		69.3	29.0	1.1	0.53
4		19.8		71.4	26.5	1.1	1.0
5		10.7		69.5	29.0	0.97	0.60
6		16.4		67.7	31.0	0.94	0.41
7		15.7		74.3	23.8	1.1	0.84
8		19.9		72.5	25.4	1.1	0.93
Pond ice
1	0.916	1.0	1.0	52.2	22.0	1.0	24.8

14.6 CONCLUSIONS

Although no infrastructure for a global assessment of carbon is yet in existence, remote sensing may give an input into the data acquisition necessary for a better estimation of the standing crop on a large scale. Investigations with satellites have shown that sea ice, upwelling, and estuaries can be monitored to a high degree, but for our purposes there is certainly a need for a more sophisticated monitoring of continental carbon pools, and spacecraft technology may open the way. Presently, the inventory of vegetation is in a very advanced mode and could theoretically be carried out on a larger scale. In this connection, it would be useful to integrate the objectives of SCOPE into the ongoing activities by other organizations to fully automate agricultural inventories from satellites.

Figure 14.12 Nimbus 5 data obtained with the electrically scanning microwave radiometer (ESMR) in a polar projection for the Antarctic region during wintertime. Source: NASA

Figure 14.13 Nimbus 5 data obtained with the electrically scanning microwave radiometer (ESMR) in polar projection for the Antarctic region during summertime. Source: NASA

REFERENCES

Baumgardner, M. F., Kristof, S. J., and Henderson, J. A. (1973) Identification and mapping of soils, vegetation, and water resources of Lynn County, Texas, by computer analysis of ERTS MSS data. In: Symposium on Significant Results Obtained from ER TS-I. NASA sp-327(1), 213-221.

Kirvida, L. and Johnson, G. R. (1973) Automatic interpretation of ERTS data for forest management. In: Symposium on Significant Results Obtained from ER TSI. NASA sp-327(2), 1075-1082.

Krinov, E. L. (1953) Spectral Reflectance Properties of Natural Formations. Aero Method Laboratory, Academy of Sciences, U.S.S.R. (trans. by G. Belkov). Nat. Res. Counc. Can. Techn. Transl. TT-439.

Matsuo, S. and Miyake, Y. (1966) J. Geophys. Res. 71, 5235.

Reinold, R. J., Gallagher, J. L., and Thompson, D. E. (1973) Remote sensing of tidal marsh. Photogramm. Eng. and Remote Sensing 39, 477-488.

Rogers, R. H., Reed, L. E., and Smith, V. E. (1975) Computer mapping of turbidity and circulation patterns in Saginaw Bay, Michigan (Lake Huron) from ERTS-Data. Prepared for: ASP-ACSM Convention Washington D.C., 9-14 March, 1975, p. 15.

Siegal; B. S. and Goetz, F. H. (1977) Effect of vegetation on rock and soil type discrimination. Photogramm. Eng. and Remote Sensing 43, 191-196.

Szekielda, K.-H. (1976) Spacecraft oceanography. In: Barnes, H. (ed.), Oceanography and Marine Biology: an Annual Review, 99-166, Aberdeen University Press, Aberdeen.

Work, E. A. and Gilmer, D. S. (1976) Utilization of satellite data for inventoring prairie ponds and lakes. Photogramm. Eng. and Remote Sensing 42, 685-694.