| GISdevelopment.net ---> AARS ---> ACRS 1992 ---> Agriculture |
Methodology for estimation of
soil moisture based on complex processing of ground observation and
satellite data
R. Oyun
National Remote Sensing Center,
Ulaanbaatar-11, Mongolia
National Remote Sensing Center,
Ulaanbaatar-11, Mongolia
Abstract
We developed methodology for integral estimation of soil moisture in Mongolia. There is used method of mathematic statistics, digital image processing and solved heat balance equation. The methodology based on geographic information, regular ground and NOAA satellite AVHRR data. the result of this investigation are soil moisture characteristics as well as, climatological value, productive moisture map, regression equation for calculation of digital map, evaporation, and moisture deficit.
1. Introduction
The most territory of Mongolia belongs to the region with insufficient and unstable soil moisture. Existing in Mongolia traditional methods for gathering of soil moisture information, based on contact measurement, using thermostat method, operating at about 50 ground stations now, do not allow to get integral information on spatial distribution of soil moisture over the territory of Mongolia.
Thus the development of operative methods for identification of soil moisture based on of integral approach, using ground observation and remotely sensed data, is a quite actual and urgent problem.
We have put a task to work out methods for determination of soil moisture characteristics and develop a system for operative processing of ground observation and satellite data. The system can implement mapping of soil moisture and it will be one of the components of an environment Monitoring System.
2. Data and method of study
In this work have been used ground measurement data of soil moisture, atmospheric characteristics, maps of vegetation, soil, forest, underground water level and relief, morphometric characteristics and multichannel AVHRR data from NOAA satellite.
This research work consists of 3 steps.
In the first step we analyzed ground measurement data and soil-landscape characteristics, using methods of correlation and regression analyses and principal components.
In the second step we analyzed satellite data. There has been used methods of multidimensional digital data processing, involving radiometric and geometric transformation, classification of AVHRR data and its registration with ground measurement data of soil moisture. Has been calculated spectral albedo, normalized vegetation index, spectral and integral radiant temperature.
A geographic registration of satellite data has been done with help of combined methods and it was transformed into Lambert-Gaussian polyconic projection.
To register ground and satellite data, the first, has been carried out estimation of soil moisture spatial variability, quantitative definition of necessary ground measurements, the second assessment of representative means characteristics around measured points by determination of homogenous spectral characteristic’s regions. The necessary condition of reliable registration is the existence of homogenous spatial regions of size not less than real geographic registration errors.
In the third step have been analyzed ground measurement and satellite data together with soil-landscape and morphometric characteristics. Has been work out method for integral estimation of soil moisture dynamic characteristics, as rate of water loss, evaporation, effective precipitation.
The method is based on heat balance equation of a simple system, where is not considered non-stationary of processes, factor of snowmelt, photosynthesis and horizontal advection. The calculation has been done by improved formula of Penman-Monteith (Bratsert, 1985).
Various resource maps, satellite information, operative synoptic and agro meteorological information, climatological characteristics and other parameters are used as input to this method.
In some cases the method is similar to a system for estimation of soil moisture of soil moisture characteristics, used in Meteorological Office of Great Britain (Thompson, 1981). The distinction is that there is considered influence of local relief variation and used satellite data. Also the surface types and morphometric characteristics have been prepared as digital map, which is result of NOAA-AVHRR multidimensional classification and realization of relief model. These maps are called key-maps as well as they characterize every pixel on image and they link the satellite data with corresponding physical model parameters.
Relief model realized over limited territory in basin of Orkhon-Selenga rivers and approximated on the base of partially bicubic function (Marchuk, 1980).
The data processing consists of 3 major parts :
- Precipitation of key-maps
- Calculation of radiation balance components.
- Calculation of soil moisture characteristics
3. Application analysis
3.1 Application of ground data
The analysis of ground measurement data of soil moisture shows that the number of ground observation stations is rare, its distribution over the territory is irregular and the range of observation period is comparatively short and heterogeneous.
At all stations the mean statistical value of soil moisture did not reach the lowest value of field water capacity ranging between hydroscopic water capacity and capillary breaking off.
By the observation results, obtained in certain period of time the water contents in soil are variable and there is daily dynamic (Fig. 1 and Fig. 2), that shows necessity of the regular, high temporal measurement.
Fig 1 Dynamic change of soil moisture (Station Darkhan, 1987)
Soil moisture in neighboring layers has a good correlation each to other, its correlation coefficient composes 77-94%, that allows to estimate soil moisture in the lower layers, using upper layers’ values.
Soil moisture at 20 cm depth has direct correlation with amount of precipitation and indirect correlation with air moisture deficit, wind speed , duration of sun light and diurnal amplitude of soil surface temperature. It is found that the correlation between soil moisture and diurnal amplitude of soil temperature is the highest and most stable one. But absolute value of the correlation coefficient is still not sufficient (0.30 – 0.60) to consider the temperature amplitude as a unique or main indicator of the soil moisture.
To obtain general nature of spatial distribution has been produced productive soil moisture map, based on ground measurement data (Fig. 3). But there was a problem to interpolate those spatial discrete values. Therefore we analyzed soil-landscape characteristics and relief variation with the help of existing maps. These maps are digitized and interpolated, depending on soil water contents. After statistical processing has been selected the vegetation as most integrated characteristics of surface, well conformed with location of discrete ground data, that gave possibility to delineate boundary of productive soil moisture spatial distribution (Fig. 4).
Climatological value of productive soil moisture
Map of productive soil moisture
3.2 Application of satellite data
For whole territory of Mongolia we attempted to estimate soil moisture in upper layers by spectral albedo, vegetation index and diurnal amplitude of the land surface temperature, derived from AVHRR.
Firstly, has been analyzed more than 1o image spectral albedo data, but we have not got significant results. The main reason for it, we concluded, that there is an intensive surface evaporation in arid regions.
Secondly, we have investigated vegetation index. As well know, vegetation index is successfully used for estimation of vegetation condition, but in arid region the determination factor of vegetation condition is just soil water content. The experience, obtained in last few years, shows that vegetation index reliably reflects the real condition of those regions, where was occurred continuous drought and dried-up pasture. In the steppe and arid-steppe zone of Mongolia was observed an increase of vegetation index value after continuous rainfall. But has not been detected its stable quantitative correlation with soil moisture on whole territory of Mongolia.
Thirdly, has been analyzed diurnal amplitude of surface temperature. The correlation coefficient between diurnal radiant temperature amplitude and soil moisture is not high, that was expected, because, they have different ground resolution and by ground measurement data this correlation was not high. But from the view of qualitative it has certain interest. Radiant temperature variation for water surface has minimum value, in case of forest area and wetland along river valley it was more but still low value. High value of temperature amplitude occurs in semi-arid region of arid-steppe and desert zone of Mongolia.
According to above mentioned facts, we concluded that the information of vegetation index and diurnal amplitude of radiant temperature could be used as basemap, on which regular data, measured at stations, are plotted and thus we can provide agro meteorologists by additional information on spatially interpolated of soil moisture characteristics.
Finally, during the investigation of some regions, for example, in watershed area of the rivers Orkhon-Selenga, after combined analyses of vegetation index (Vi) and radiant temperature amplitude (dT) with soil moisture at 10 cm layer (W) we have got following regression equation with confidence probability of more than 80%:
W=10Vi – 0.67dT + 38.04, r = 0.68, s=1.56, p=0.80, (1)
W=0.38dT-1.32dT +40.40, r=0.86, s=1.10, p=0.90, (2)
W=-7Vi-1.13dT+49.90, r=0.85, s=1.10, p=0.85, (3)
Where r – multivariate – correlation coefficient, s – standard error p – confidence probability
Among the analyzing parameters daily variation of radiant temperature, derived from channel 3 has highest correlation. (Table 1), with increasing of soil moisture the reflectance and diurnal temperature variation of the surface is decreased, but difference of channel 3 radiant temperatures in daytime and nighttime (dT) characterizes these properties and it is the most informative factor for estimation of soil moisture in surface layer.
| Elements | W | Vi | dT | dT |
| W Vi dT dT |
1 0.41 -0.62 -0.84 |
1 -0.25 -0.62 |
1 0.85 |
1 |
In Fig. 5 has been shown fragment of soil moisture map, calculated by formula (1).
Calculated value of soil moisture at 10cm depth, Orkhon-Selenge river basin, 15 June 1989
3.3 Application of satellite and ground data complex
Here is given some statistics of the different surface characteristics, those are the heat balance equation calculation result. Each characteristic classified and defined the factors caused the classes.
In the study area have been delineated 4 different classes by their value of daily net short-wave radiation balance. The main factory of different classes is relief aspect (table 2a).
There are 3 classes by the calculated value of daily mean surface temperature. The main factory of differences is vegetation cover and relief altitude (Table 2b).
| No | Classes | Number of elements | Radiation | Aspect | ||
| R | A | |||||
| 1. | Nearly flat sur.& the slopes with south asp. | 6234 | 506.6 | 40.3 | 3.2 | 5.5 |
| 2 | Slopes with SW & SE aspect | 5361 | 492.5 | 20.4 | 47.0 | 7.8 |
| 3 | Slopes with E & W asp. | 1367 | 478.1 | 13.6 | 113.1 | 3.6 |
| 4 | Slopes with NE & NW aspect | 4643 | 467.4 | 10.5 | 144.0 | 6.9 |
| 5 | Slopes with north asp. | 1375 | 465.8 | 10.0 | 173.7 | 3.4 |
| No | Classes | Number of elements | Temperature | Veg. index | Altitude | |||
| T | Vi | H | ||||||
| 1. | Forest | 323 | 11.7 | 0.6 | 49.6 | 12.7 | 1296 | 264 |
| 2 | Vegetation | 7114 | 15.2 | 1.3 | 31.3 | 11.5 | 1040 | 232 |
| 3 | Soil & sparse vegetation | 9060 | 19.6 | 1.1 | 10.9 | 6.0 | 856 | 122 |
By calculated value of evaporation there is clearly delineated four classes Fig. 6). Statistics structure of evaporated for a single class has Gaussian distribution. The amount of evaporation is tightly correlated with vegetation index and temperature. The boundary of every class identified (Table 3) and within a class is observed almost linear correlation between evaporation and temperature.
Calculated value of daily evapotranspiration Orkhon-Selenge river basin, 15ne 1989
The analysis of results has been done by means of comparison of calculated and measured data from meteorological stations, located within the study polygon. In this case the accuracy of moisture determination ranges 1 – 3% and the error of evaporation calculation is less than 1.5mm. The lack of spatially averaged value of ground measurement data makes difficulties of the estimation. Generally, the error of various characteristics calculation by this method is less than 15% of absolute value of it.
4. Conclusions
Analyzing the ground measurement data we have got climatological characteristics, general nature of spatial and temporal distribution for soil moisture and its correlation with meteorological and other characteristics of Earth surface.
Using AVHRR data there was defined there was defined regression equation for soil moisture calculation, developed technology for preparation of soil moisture map and proposed method on registration of satellite and ground measurement data in the area with complicated relief.
Geographic map, satellite and ground data complex gave the method for an integral estimation of soil moisture dynamic characteristics, as well as, daily variation of moisture, evapotranspiration, evaporation, soil moisture deficit and other surface characteristics such as radiant balance components, albedo, temperature and morphometric characteristics. Such complex approach as in terms of information provision also data processing method gives quite accurate results and has more potential. Effective realization of it needs more complex database and integrated geographic information, ground and image data processing system.
It needs to note that the results, taken by us are only primary attempt for spatial distribution of some surface characteristics and their relationships. It shows the real possibilities of meteorological satellite digital data application in case of Mongolia.
Reference
- Batjargal, Z., Oyun, R., 1989, The application of mathematic-statistic method for investigation of natural phenomena, Ulaanbaatar, Mongolia, 211p. (In Mongolian language).
- Bratsert, U.H. and others, 1985, Evapotranspiration onto atmosphere, Gidrometeoizdat, A.B., 1980. The methods of numerical mathematic. Science, Moscow, 536 p. (In Russian language).
- Marchuk, A.B., 1980. The methods of numerical mathematic. Science. Moscow, 536 p. (In Russian language)
- Thompson, N., Barie, I.A, Ayles, M., 1981. The Meteorological Office Rainfall and Evaporation Calculation System MORECS//Hydrological Memorandum No. 45, Meteorological Office, London, 69p.
Polygon Tsagaantolgoi 15 June 1991
| No | Surface | N | Potential evapotranspiration | Real evapotranspiration | Vegetation index | Temperature | Total radiation | Altitude | Primary soil moisture 10sm | |||||||
| E | E | Vi% | T | H | Wo | |||||||||||
| 1 | Barren land & sparse vegetation | 9297 | 3.47 | 0.21 | 2.75 | 0.20 | 10.2 | 5.2 | 19.4 | 1.8 | 480 | 17 | 850 | 128 | 11 | 3.3 |
| 2 | Vegetation | 4704 | 4.38 | 0.23 | 3.91 | 0.24 | 26.1 | 5.8 | 16.0 | 1.9 | 475 | 22 | 991 | 208 | 15 | 4.2 |
| 3 | Dense veg. bush & forest | 2191 | 7.36 | 0.23 | 6.33 | 0.22 | 43.7 | 2.5 | 14.4 | 1.1 | 478 | 27 | 1216 | 184 | 21 | 3.7 |
| 4 | Forest | 276 | 9.61 | 0.26 | 7.70 | 0.26 | 50.6 | 1.1 | 13.7 | 1.1 | 468 | 23 | 1206 | 152 | 22 | 2.8 |