For this paper the relationship between passive microwaves and top soil moisture has been investigated. It is known that the combination of Remote Sensing and hydrology is very promising, since the physical property that effects the Remotely Sensed measurement is directly dependent on the soil moisture (Jackson et al. 1989). It is important to know soil moisture conditions, since this is a key factor influencing the potential evaporation, and evapotransporation, which are important terms in ecosystem dynamics (van de Griend et al., 1993), and consequently atmospherical studies. For atmospherical studies, large data sets covering large areas are required. A major advantage of the use of passive microwave techniques and remote sensing over conventional techniques is that over large areas the soil moisture can be frequently measured in a relatively short time (Jackson et al. 1989), under any weather and terrain conditions. In this study a linear relationship has been established between top soil moisture and passive microwave measurements, allowing for the creation of top soil moisture maps. With this relationship, passive microwaves, measured at any time, can be transformed into top soil moisture maps, without the necessity of field measurements.
The geology of this area, consists of Continental Terminal origin, which are the oldest surface sediments and are linked to the deposits of Northern Nigeria. These deposits are of Eocene age and are deposited under lacustrine and terrestrial conditions, sheets of plinthite developed over extensive areas. Three major sheets of plinthite have been formed (Epema et al., 1993). The vegetation within the HAPEX-Sahel study area consists of vegetation types like arable crops (millet), fallow savannah, sparse dry land forest known as 'tiger bush' and 'degraded bush' (Kabat et al. 1993).
- the energy driving forces, like the solar energy, and resulting radiation,
- the moisture availability, and/or situation of soil/vegetation interaction
- the capacity of the atmosphere to absorb energy, which is dependent on surface and air temperature, vapor pressure gradients and surface winds.
Using the energy balance approach, the evapotranspiration (or LE) can be calculated as a residual term, or as an explicit function of properties and variables like thermal inertia, surface radiation temperature, vegetation index and soil moisture.(E. van Proosdij, and R.Jochem, 1994).
A microwave radiometer (as used on the NASA aircraft), measures the thermal emission from the surface in the microwave range (1mm up to 1 meter). According to the theory, it is appropriate to account for the Emissivity (Wang et al. 1981). As Tsoil measurements were not available for this study, the direct applications of TB has been investigated, by using regression analyses to determine the relationship between the field soil moisture data and TB measurements. The major advantage of only using the TB is that there is no necessity to do measurements on the ground and TB is a variable which can be measured under all-weather circumstances. Factors influencing the emitted radiation are surface roughness and vegetation effects. In this study no corrections have been applied, since data required for a correction of the surface roughness and vegetation were not available. A problem that might arise in the future could be the general applicability of an empirical approach.
This study involves data from the 'West-Central Super site' (WC-site), which has a grid of about 15X15 kilometers. Within the WC-site several plots and transects were selected, taking criteria in consideration like vegetation cover and soil type.
Plots A and A" can be described as 'fallow bush/grassland', which consists mainly of shrubs and a ground layer of grasses and annual herbs. The spacing of the bushes is about five meters in the denser areas. Plot C is the 'tiger bush' site, this site consists of bushes which can reach six to eight meters in height, and has bare spacings of fifty meters in between. Plot D is called 'degraded bush', and has a cover of shrubs which cover the site for less than five percent. The plot is for more than sixty percent made up of bare soil. The name for the last plot is 'old millet', it is a field where the millet did not emerge in 1992, which is clearly visible on the satellite images, as it has a white color (Goutorbe et al., August 1992).
The measured transects vary in length. The transects follow the contours of the old river valley. At both the plots and the transects the soil moisture has been measured. The measurements were obtained in two ways; using a TDR (Time Domain Reflectometer), and by gravimetric sampling.
The TDR measurements were obtained by positioning two parallel rods five centimeters (corresponding to the skin depth of the PBMR-measurement) into the ground under an angle of 45ø. The main module gives the volumetric moisture content (é) on a digital display.
The measurement of top soil moisture through gravimetric sampling uses a ring which has a volume of exactly 100 cc. It was filled with a sample of the top five centimeter of the soil, then ring and sample were weighed together. Then the soil sample was dried and weighed separately. The volumetric soil moisture (é) can be determined by subtracting the 'dry'-sample weight from the 'wet'-sample weight.
The transects were measured using the TDR instrument, measuring in two parallel lines which were about 20 meters apart in a more or less south to north direction. At about every twenty five meters a measurement was taken.
The plots were measured by using the TDR, as well as the gravimetric method. The lay out of the measuring grids varied per plot. Sometimes the side of the measurement grid was 25 meters, and sometimes a 100 meters, yielding a total of measured points varying per plot from 24 up till 49 (Soet et al., 1993).
A data set of Brightness Temperature (TB) measurements has been gathered through the use of a (PBMR) Push Broom Microwave Radiometer, which had been mounted to a NASA C-130 aircraft. The PBMR-measuring device operates at L-band range, which has a frequency of 1.41 Ghz, and 21-cm wavelength. The PBMR has four horizontally polarized beams. The device takes measurements of a ray of 8 to 10 pixels wide, with each pixel size of 40X40 meters. Each row of pixels is situated under a 90 degree angle to the flight direction, forming a scan line of variable length. Data of the different scan-lines are obtained sequentially with the movement of the airplane. Each of the beams has a full width at half maximum power of about 16ø which yields a footprint on the ground 0.3*H for each beam and a total swath of about 1.2*H, where H is the aircraft flight altitude (Smugge, 1993).
The available PBMR data and soil moisture data were acquired on five different days in 1992, namely on 8/25, 8/26, 9/2, 9/4 and 9/12. The PBMR data were acquired with several overpasses over the study site at an altitude of approximately 300 meters in an East to West direction. With this technique the difference in position of the planned scan-lines and the actually measured scan-lines could be as high as 80 meters. This deviation occurred since the pilots depended on visual navigation, instead of on instruments. A problem which accompanies the low flight precision was that the PBMR-images not always completely cover the soil mois- ture field measuring sites. The measured range of the TB measurements was between 200 and 290 K. PBMR-flights and the field soil moisture measurements were mostly planned just after a rainstorm, which is the most dynamic time during which the soil was drying out. The field crew needed about six hours to measure the top soil moisture at the field sites (usually from 10.00 AM. till 16.00 PM.), whereas the scanner flights needed two hours to finish (from 12.00 PM. till 14.00 PM.).
After soil moisture calculations had been performed, it was established that the mass based approach yielded similar results to the direct volumetric water content approach. It was decided to discard the mass based approach, and only use the direct volumetric water content approach.
The exact coordinates of the locations of the field soil moisture measurements were not available in the right coordinate system. They were available in the longitude and latitude system, whereas the ERDAS-system used in this research only uses UTM-coordinates. As a proper coordinate transfer program was not available, the transfer calculation had to be executed using a formula derived from the long/lat relations and the UTM from a map.
An interesting standard deviation was found in the measurements performed on the fourth of September, these measurements did not follow the dry down pattern at all. But since the cause of this deviation could not be determined, these measurements were not excluded from the regression analyses. After the first regression analysis were executed, it became clear that the two top soil moisture measurement sets (plots and transects) had very diverse characters regarding the accuracy of the measurements. The spatial set up of the two were totally different (linear versus two dimensional), which was the basis for using to the two sets separately in further regression analyses. As the remaining time to do research was becoming short at this point, the cause of the deviations in the regression analyses was not fully investigated.
1 -> é = ( measured TB / -3.3008 ) + ( 194.9201 / 3.3008 )
2 -> é = ( measured TB / -5.1813 ) + ( 204.8885 / 5.1813 )
3 -> é = ( measured TB / -3.2650 ) + ( 191.6492 / 3.2650 )
The first formula has been based on the data of the transects only, the second formula has been based on the plots only and the third formula has been based on a combination of the data of all the transect and all the plots.
In some transects, due to topographical features, large variations in TB did occur. Whenever this was the case, in order to improve the regression results, some transects have been divided onto smaller stretches. The rule of thumb used was; if the values of measurements of the TB within the same transect differed more than 4 K, and these differences could be explained by a relation with a change in the topographical features, the transect was divided in two, or even three stretches. The split up of the transects was performed by combining the values of the part of the transect that had similar value patterns. It was found that if the values of the separate areas were averaged out, the regression results were greatly improved. It was decided to base spatial variations on the TB data, since the variations in the measurements of the soil moisture data were, generally speaking, too inconsistent to determine any spatial trends. The measured soil moisture values were spatially linked to the TB measure- ments through regression, so they always have the same spatial pattern as their accompanying averaged TB measurements do.
The first formula is based on the regression performed on the data of the transects only. The second formula is based on the regression performed on the plots only. The plots have been split up in four areas of equal size, because as has been previously found for the transects, there is a considerable amount of variation in the TB values due to topographical features. This holds especially true when the data were viewed in a north to south direction, since the old river valley ran more or less east to west. For the plots the same regression method was used for the transects. The third formula is based on a combination of the data of plots and the data of the transects. The same two data sets have been used for the third formula, as has been used for the first two formula's.
The regression results obtained from the three regression analyses seem to indicate that the regression performed only with the data of the plots yield results with the highest correlation between measured soil moisture and Tb. A combination of the plot data with the transect data has no satisfactory effect, it actually lowers the R squared. The use of just the transect data does not give good results, probably due to the lack of incorporation of two dimensional variability (transects only measure in one direction).
The three regression formulas, have been used to calculate three sets of soil moisture maps for the five PBMR-images available. Between these soil moisture maps some comparisons have been made. The calculation for the soil moisture at the transects has been compared with the actual measured soil moisture in the field.
The TDR measurements were linked to Tb measurements taken from a NASA plane, which flew overhead at an altitude of about 300 meters. In some cases the plane overflight track had a deviation from its intended position on the ground of 80 meters. The Tb measurement grid was 40X40 meters, so the plane could have been in a position as much as 2 pixels from the intended position on the ground. Soil moisture measurements were taken during the entire day, whereas the PBMR flight-Tb measurements only lasted for less then an hour. Two other important factors that could influence the Tb measurements are the vegetation density present at the location, and the soil roughness. A correction should be applied for these factors, however during this research no data were available to perform such a correction.
It was decided to investigate what differences in calculated soil moisture would result in the use of specific data combinations. For simplicity, only three (previously described) combinations were made. In the calculations of the soil moisture the three formulas tend to under estimate the field soil moisture condition, (see average errors). The wetter an area is, the more the formulas under estimate the soil moisture conditions, (see difference in calculated soil moisture results for plot five and plot six after a rainfall event). The formulas work better for the drier regions. This could be because in general on wetter areas there was more vegetation present, which can cause more disturbance in measurements. Vegetation tends to adsorb passive microwaves emitted by the soil surface, and instead transmits its own passive microwaves. Since no measured data were available on vegetation, the disturbance by vegetation could not be corrected.
Under dry conditions (transect 5), the formula based on only the plots predicts soil moisture the most accurate. However under medium wet conditions the other two formulas are more suited for the prediction of the soil moisture. The results from this study should be appreciated in the right perspective, since this has just been a preliminary study.
Three formulas have been created, based on two sets of field measurements. Overall the formula based on the transects has the highest accuracy in predicting soil moisture, this is logical since the formula based on the transects is used to predict soil moisture of the transects. This was the only method available to check the prediction accuracy of the formulas since no other soil moisture or Tb data were available. Therefore, for future research it would be better to verify the formulas accuracy, by administering them on an entirely new area, with field soil moisture data which have not been used in the derivation of the formulas.
More investigation is necessary on the effects of terrain variability, soil classes, vegetation effects, time coincidence of the measurements in the field and the measurements from the air.
The three created formulas tend to underestimate the soil moisture, however dependant on the amount of accuracy required in the prediction of the values, the formulas are able to predict soil moistures quite accurately. Keeping in mind that the results of this study will eventually be used to create climate models with pixel scales of a 100 km2, the errors in estimation are within a reasonable limit. Also one has to keep in mind that the results can be improved by correction for factors earlier mentioned if the necessary data are available.
This research fits in with the global research efforts to observe and try to understand the mechanisms of the climate and processes that take place above the earth. This research is just a small link within a huge apparatus. By carrying out this research on this particular part of the climatical process, we hope one day to achieve understanding of how all these processes work.
- Cultuurtechnische Vereniging, 1988. Cultuurtechnisch Vademecum. Cultuurtechnische Vereniging, Rijkskantorengebouw Westraven, Utrecht, The Netherlands.
- G.F.Epema et al., 1993. Monitoring land surface characteristics in Niger using complemen- tary Spot and ERS-1 SAR data, Wageningen Agricultural University, The Netherlands.
- Goutorbe et al., August 1992. Hydrological and Atmospheric Pilot Experiment in the Sahel, Niger 1992, Experimental plan.
- Goutorbe J.P., T.Lebel, A.Tinga, P.Bessemoulin, J.Brouwer, A.J.Dolman, E.T.Engman, J.H.C.Gash, M.Hoepffner, P.Kabat, Y.H.Kerr, B.Monteny, S.Prince, F.Said, P.Sellers, Jackson T.J., and T.Smugge, 1989: Passive Microwave Remote Sensing System for Sol Moisture : Some Supporting Research, IEEE Transactions on Geoscience and Remote Sensing, Vol. 27. NO. 2. March 1989.
- J.S.Wallace, 1993. HAPEX-SAHEL: A large scale study of land-atmosphere interactions in the semi-arid tropics, HAPEX-Sahel general paper, version: 28 September 1993.
- Griend van de A.A., M.Owe, and A.T.C. Chang, 1993. Botswana water and surface energy balance research program, Institute of Earth Sciences, Dept. of Hydrology, Vrije Universiteit, Amsterdam, The Netherlands.
- Journal of Hydrology, March 1 1993, vol.143 - Nos. 1-2, recognized by the European Geographical Society as one of its official journals.
- Jackson et al. 1991. Vegetation effects of the Microwave emission of soils, USDA-ARS Hydrology Lab, Beltsville, Maryland USA, published by Elsevier Science Publishing, New York.
- Kuhnel et al., 1991. Partial analyses applied to scale problems in surface moisture fluxes, Center for Water Resources Research Department of Civil engineering, University College, Dublin, Ireland, published by Kluwer Academic Publishers, Dordrecht, the Netherlands.
- P.J.van Oevelen et al., May 1993. Soil moisture and surface roughness measurements during HAPEX-Sahel 1992, ground data collection report, vakgroep waterhuishouding, landbouwuniversiteit Wageningen, the Netherlands.
- M.Owe, A.A. van de Griend, 1993. Surface moisture monitoring by passive microwave techniques in semi-arid environments, NASA/Goddart Space Flight Centre, Greenbelt, MD 20771 USA
- S.D.Prince, P.Kabat, and others, 1994. Geographical, biological and remote sensing aspects of the Hydrology Atmospheric Pilot Experiment in the Sahel (HAPEX-Sahel), Department of Geography, room 1113 Lefrak Hall, University of Maryland, College Park, MD 20742, USA.
- Proosdij E.M.H. van, and R.Jochem, 1994. The connection between energy equation and water balance, IAH-Larenstein, Velp. the Netherlands.
- T.Smugge, J.R.Wang, and G. Asrar, Oct. 1988. Results from the Push Broom Microwave Radiometer Flights over the Konza Prairie in 1985, USDA, Hydrology Lab, Beltsville, MD 20705, USA.
- T.Smugge, T.J.Jackson, W.P.Kustas, R.Roberts, and R.Parry, 1993. PBMR observations of surface soil moisture in MONSOON 90, USDA, Hydrology Lab, Beltsville, MD 20705, USA.
- M.Soet et al., December 1993. HAPEX-Sahel, Basic description of methods and data sets, vakgroep waterhuishouding, Wageningen, the Netherlands.
- R.Wang et al., October 1980. An empirical model for the complex dielectric permittivity of soils as a function of water content, Laboratory for atmospheric sciences, NASA/Goddard Space Flight Centre, Greenbelt, MD 20771, USA.