ESTIMATING AERIAL BIOMASS IN SEMI-NATURAL VEGETATION FROM SPECTRAL REFLECTANCE MEASUREMENTS .1. PRELIMINARY EXPERIENCES

SUMMARY 
 
A technique of calculating aerial green biomass from spectral measurements in the red and near infrared bands is increasingly being applied in vegetation studies. It is fast, non-destructive and can be directly applied from aircraft or satellite mounted sensors to large areas. Literature data illustrating the potential of the approach refer sofar generally to rather ideal situations in low, open, herbaceous vegetation. 
 
 
 
Before general application can be recommended a body of reference including various situations should be available. The present contribution indicates some disturbing factors such as the presence of flowers or shade and evaluates the applicability of the technique for a number of vegetation types. Results of tests in 11 semi-natural vegetation types are presented, ranging in height from 30–150 cm and in cover from 20–100%. Measurements were made with – as later appeared – non-optimal equipment and no correction could be made for variation in incoming radiation. Multiple sampling at three to five dates between April and July resulted in correlation coefficient values of 0.96, 0.95, 0.92, 0.91, 0.91, 0.88, 0.85, 0.82, 0.67, and -0.06 (in the last case Holcus lanatus dominant) for aerial biomass dry and the IR/R ratio. These relatively high values are critically examined. The data provide evidence for a rather general applicability of the technique, but also for a cautios approach and a mandatory calibration per (floristically and/or physiognomically) different vegetation type.

Estimates of the amount of above-ground vegetation, referred to as aerial biomass or standing crop, are used in many kinds of vegetation studies. Where the vegetation is studied as an exploitable renewable natural resource in itself, estimates of the amount of harvestable plant material are essential. Grazing potential research and the evaluation of rangelands are obvious examples (Stoddart et al. 1975). Traditionally clipping and weighing of vegetation in sets of sample plots has provided the estimates. The method is time consuming and, where manpower is expensive, costly (Brown 1954). It is destructive as well, preventing repeated measurement of the same sample plots to estimate vegetation productivity. Of the equipment developed for non-destructive measurement, the use of electronic capacitance instruments is best known, not only in herbaceous but also in shrub vegetation. The technique has disadvantages and is not suitable for determinationof aerial biomass over large areas Currie et al. 1973;Morris et al. 1976).
An approach of using spectral reflectance measurements in selected narrow wavelength bands has been developed in the last decade and is now widely being tested. Such measurements are rapid, non-destructive and can be made from different altitudes, related to small plots as well as to large areas. The sensor(s) can be hand-held, but also mounted in aircraft or earth orbiting satellites.
Particularly this last option has induced extensive further research.
The development of the approach is well documented in Miller & Pearson (1971), Pearson & Miller (1972) and Tucker et al. (1975). The technique is based on the characteristic reflectance properties of green living plants. The plant pigments (mainly chlorophylls) absorp solar irradiance in the red band of the spectrum. The more green plant material present, the higher the amount of pigments and absorption and the less the reflected (measured) energy in this band. Reflection in the near infra-red section of the spectrum increases by increased amount of biomass. In this band absorption is low and multiple reflection causes a measurable increased reflection at higher amounts of green vegetation. For theoretical considerations and experimental evidence reference can be made to Knipling (1970), Woolley (1971), Sinclair (1973), Tucker et al. (1975) and Bunnik (1978). A literature review is given in Beck (1979). reflected light-energy values measured in the red (R: 0.650-0.700 pm) and near infra-red (IR; 0.775-0.825 pm) bands, used in combination, particularly as ratio, are therefore a promising measure for the amount of green vegetation.
A relatively simple, handy instrument developed on this principle is the so called biometer (bio-mass meter), described in Miller (1973) andPearson et. al. (1976). The potential of spectral mapping of biomass following this principle from aircraft was used by McNaughton(1976) in East Africa and an airborne multispectral scanner application was reported in Pearson & Miller (1972) and Pearson et al. (1976). The feasibility of band ratio biomass estimation from space platforms (Landsat) was demonstrated amongst others by Carneggie et al. (1975), Haas et al. (1975) and in Maxwell (1976). Recent research in this field also deals with separate estimation of live wet (green) biomass and dry (brown) plant material (Tucker 1977a(Tucker . b. 1980 (Pechanec & Pickford 1937;Brown 1954;Kinsinger & Strickler 1961;Mason & Hutchings 1967;Thalen 1979 Homogeneous cover of mainly Carex acuta and C. aquatilis. Reference: described in detail in Boedeltje(1976

Open vegetation of Glyceria maxima
A rather high vegetation of only this species. Growth development appeared to be not optimal during the season,indicated by brown leaftips and retarded growth compared to adjacent areas.
Reference: Area 20 in Leemburg (1974) Open vegetation of Glyceria maxima A rather high vegetation of only this species. Growth development appeared to be not optimal during the season,indicated by brown leaftips and retarded growth compared to adjacent areas.

RESULTS AND DISCUSSION
3.1.

Sun angle and shade
The reflection coefficient for most types of vegetation changes with the angle of the sun. Minimum values are found when the sun approaches its zentih and higher values as the sun descends. In this last situation there is less multiple scattering in the canopy. This has an effect on the ratio values. Figure 2 shows the course of the values of ratio between 9.00 and 17.00 hrs for three sets of conditions. In fig. 2a   the "diffuse" character of the reflection changes in a more or less "mirror" character (Suits 1972  Warren-Wilson (1965) and Bunnik (1978) have investigated this point and an observation angle of 50-55 seems optimal (see also Beck 1979).

Wet versus dry vegetation surfaces
The influence on the reflectance ratio of leaf surfaces being wet can be seen from table 2. No major differences were found except for the Urtica dioica stand. The impression was that these differences could at least partly be explained by changes in leaf position under the weight of the water. It should be noted that the  High correlations were found for the open Sphagnum-Anthoxanthum stand and the Typha-Phragmites stand. It should be noted that the moss layer was not included in the clipping. An equally green (bright green for the Sphagnum spp. and dark green for the Scorpidium cover) or water underground has apparently little disturbing effect, as long as it is included in the calibration. The harvested amounts of biomass were of the same magnitude. Reflection ratios for the second stand, however, were generally more than twice as high as for the first one. The picture for the third stand, a dense Phragmites australis cover, is not yet fully clear due to lacking data for a period of very fast growth between early May and early June. The correlation coefficient only expresses a general positive stand. The stand of the low Carex spp. had a low biomass (less than 300 g/m 2 dry weight in July) but nevertheless a positive correlation was found between the clusters. In the dense stand of Holcus lanatus no correlation at all was found between the ratio and the aerial biomass. The soft hairy surface of this species, giving it a greyish appearance, apparently disturbs the picture that can be seen for the other species and species combinations.

Effect of ripening (drying)
For some of the measured stands a marked increase in non-green vegetation parts was seen between June and July (N.B. apart from the flowers which were always removed before measurement). Stems and leaves turned yellowish to light brownish. Separation of the green material was not possible in these cases and the reflection ratio decreased. For two stands, the "species-rich meadow" and the "low grasses and forbs" this effect even resulted in a negative correlation, possibly according to the principle: the olderthe less greenthe higher the aerial biomass amountthe lower the reflectance ratio. In fig. 4  (ii) The digital reading of the instrument for which switching to differentscales of sensitivity was required, proved a slight source of error.
(iii) Readings were taken between 10.00 and 16.00 hrs. This may cause variation due to the differences in sun angle (see 3.1.).
(iv) The observations were made vertically and not under the now recommended angle of 50-55 .
(v) An effect of partial cloudiness and therefore changing conditions, could not always be avoided.
(vi) In the biomass clippings of only 1/16 m 2 an edge-effect may have been introduced, especially in the higher vegetation types. This can be compared to the difficulties faced when establishing a rooted or shoot frequency.
(vii) In all cases the area for which the ratio was measured (at ground surface at least 1400 cm 2 ) was considerably larger than the area clipped (only 625 cm 2 ).
It may be noted that some of the above sources of error will probably be unavoidable in a future operational system.

CONCLUSIONS
1. The use of spectral reflectance measurements for aerial (green) biomass estimation seems feasible in a variety of structurally and floristically different vegetation types. The technique still requires an extensive testing and calibration in each type separately, whereby the phenology should be given careful consideration.

2.
Measurements in the shade or of plots with large amounts of shade are of little value. This can be minimized by measurement at clear sky in the same direction as the incoming solar radiation.
3. The presence of brightly coloured parts (flowers, coloured leaves, etc.) can be a source of great errors. Whenever possible such parts should be excluded from the measurements.

577
ESTIMATING AERIAL BIOMASS 4. Measurement at clear sky from above, or in general under an angle of observation of 50-55°g ives least interference with vegetation structural differences (literature data, see Beck 1979).
5. Measurements should be taken as far as possible at about the same time of the day and under similar weather conditions. 6. With non-optimal equipment and vertical measurement the best linear positive correlation between the band ratio and the aerial biomass was found for vegetation with the following characteristics (i) many vertical elements and little horizontal overlap, (ii) cover less than 80%, (iii) not too high a vegetation, (iv) no