The registration of NOAA - AVHRR and ground truth data

The registration of NOAA - AVHRR and ground truth data

R. Oyun, M. Bayasgalan, Sh. Munkhtuya, M. Enkhbayar
National Remote Sensing Center, Mongolia

Abstract
Here is shown a possibility to regist NOAA-AVHRR and regular meteorolocial data and to extract spatial environment information with help of combined method for remotely sensed and ground truth data. In this case we analyse various natural resource data like soil, vegetation, landscape maps and surface physical parameters as spectral albedo, integral temperature over ground observation stations surrounding area and ground measurement data as vegetation height, biomass and surface temperature. On the base of this analyse we classified ground meteorological observation stations into 2 classes due to their data correspondence to spatially distributed satellite data.

1. Introduction
Dynamic monitoring of the natural environment using regular remote sensing data required also regular ground truth data, some of them might be provided by meteorological station network. From several years' application of NOAA-AVHRR data in Mongolia for surface parameter estimation, it is known, the important and necessary role of the correct and accurate registration of ground and satellite data, because firstly, they have different spatial and time scale, secondly, geometric transformation of AVHRR data to the geographic map rejection has some error, thirdly, most meteorological station is in or near to the populated area, therefore in some cases the measured data characterize more urban area than surrounding natural environment.

In this paper we present some result of analysis of Mongolian meteorological station surrounding environment, using landscape, relief, vegetation and soil type and different environment ground stations' data correlation to radiance temperature, vegetation index, derived from AVHRR data.

2. Data and methods
Soil, landscape, grassland and fodder resource map, regular meteorological and agrometeorological ground measurement data and AVHRR data had been used for this investigation.

Geographic maps, digitized and transformed into Lampert Gaussian projection and multitemporal satellite data are overlaid, Had been analysed correspondence of satellite data and geographic maps, satellite and ground measurement data, for example, vegetation index and radiance temperature with soil type, fodder, relief maps and ground observation data like vegetation height, productivity and land surface temperature.

3. Interpretation analyses
To classify meteorological station by registration possibility with satellite data was selected an area of 25*25 km around every meteostation.

We describe below some examples of analyses of ground desert, semi desert, steppe and mountain area.

Fig. 1 Location of analyzed meteostation
Station Bayanbulag
The most area of this station occupies block high mountains with golets, meadow and intermountain hollow of Central Khangai sub province. In the south-western part, small area belongs to dry steppe and desert steppe plateau of Southern Khangai. But soil and vegetation maps of this area show various types of soil and vegetation , by which we can include the station into area of heterogeneous environment classes. There is a mountain dark chestnut soil belt, elongated from south western to north eastern of the area, with minimum fodder productivity 400-500 kg/h and plain and hollow chestnut soil, covered by steppe and dry-steppe pasture, where is observed maximum vegetation index value 0.29-0.35 and minimum radiance temperature 20-28 0C. Also although there is a minimum fodder resource ( 300 - 400 kg/h ) correspondence. The area covered by a. Vegetation b. Fodder c. Landscape d. Soil pasture among mountain's forest has minimum vegetation index and maximum radiance temperature, naturally, At the same time ground station measured 12 cm height and 20 kg/h biomass of grass, that is much less than the value from fodder resource map. Reason for such big distinction does not understandable, it might be difference of biologic and agro meteorological measures. So, we found there is multifaced correspondence between geographic maps, ground and satellite data. In this case it is difficult to ground and satellite data. In this case it is difficult to directly regist ground and satellite data. It is difficult to directly regist ground and satellite data. It is interesting that there is good correspondence between vegetation index and radiance temperature classes ( Fig. 2e & 2f ). There is low temperature, where the value of vegetation index is higher.

Fig. 2 Geographic and satellite data maps surrounding Bayanbulag
Station saikhan
Fig. 3c shows the northern part of the area belongs to desert of northern Bulgan, but the southern part - to semidesert. On the map, shown in Fig. 2e the area under fodder resource of 200 - 300 kg/h is delineated by its low value of vegation inxex ( 0-0.02 ) and area with productivity of 200-400 kg/h appears with more value of vegetation index ( 0.04-0.09). The station has low value of vegetation index. Ground agro meteorological station measured quite high 4-10 cm height and low 60 kg/h biomass of grass, Measured biomass much less than fodder resource as previous station, There is a good correspondence between above mentioned parameters. Low temperature is observed over those area under dense vegetation and vise versa. Also it is found that ground measurement temperature and integral radiance temperature has a good correlation. To the high value of soil temperature ( 45-460C, derived from Channel 4 of AVHRR data and low value of radiance temperature ( 26-280C) reflects low soil temperature 370C. In other hand the temperature value depends on relief. In the northern part with 1200 meters of altitude the temperature is higher in comparison of southern part, where the altitude ranges 1400-1500 meters. Thus the station is registable with satellite data.

Fig 3. Geographics and satellite data maps Surrounding Saikhan
Station Erdenesant
The station belongs to the desert of southern Khan-Khongor (Fig. 4). The south eastern part of the station is meadow-steppe pasture with mountain chernozems soil and fodder resource of 900-1000 kg/h, where vegetation index value ranges 0.37-0.39 and 0.44-0.51, to the part with fodder resource of 700-800 kg/h corresponds 0.25-0.33 of vegetation index and the other parts of corresponds 0.25-0.33 of vegetation index and the other parts of the station is covered by mountain dark chestnut soil with productivity 400-500 kg/h, which characterized by vegetation index value of 0.21-0.24. Vegetation index map has a good correlation with soil and vegetation map, In the south western part of the station, where is sandy soil, the value of vegetation index is 0.12-0.16. But complicating with this, on fodder map this area has 700-800 kg/h of productivity, which is in doubt. At ground measured 6 cm height and 100 kg/h biomass of grass. Vegetation that is 4-5 times lower than fodder resource. Comparing satellite and ground temperature data the radiance temperature reflects real temperature difference at the surface. Temperature difference between two observation period was 6.80C and radiance temperature of corresponding period was almost same ( 8 0C) the specific feature of this station is that more detail classes delineated on satellite data than resource maps. we include the /station into Registable class.

Fig. 4 Geographic and satellite data maps surrounding Erdenesant
Station Khanbogd
From landscape map we can see north western part of the station belongs to dry-, sandy-desert of Northern Khanbogd and the other parts to the desert of Southern Khanbogd (Fig. 5). Comparison of fodder resource and vegetation index maps shows the highest value of fodder resource ( 200-300 kg/h ) corresponds to vegetation index value. Height of grass is 3-8 cm by ground measurement.

Northern part of the area has brown soil with aeolian deposits and semistabilized sand, where spectral albedo ranges 0.21-0.23, but southern part is covered by steppe-desert brown soil with coarse cobble, albedo of which is 0.16-0.20

At the north-eastern and south-western parts of the area, where is observed low vegetation index value, the temperature is more by 1-30C than other parts . Measured soil surface temperature at this station was 48.60C on 24 June 1991 but in July it decreased by 12 0C and radiance temperature during these periods came down 110C . The temperature field is more homogeneous. We classified the station as Registable.

Fig. 5 Geographic and satellite data maps surrounding Khanbogd surrounding Erdenesant
Station Darkhan
The station's surrounding area is northern dry-steppe and meadow (Fig.6) . There flows Kharaa river from south to north direction. The environment of this station is heterogeneous. In comparison of above analyzed stations the station has a lot of vegetation types with wide range of fodder resource ( 200 - 1300).

Fig. 6 Geographical and satellite data maps surrounding Darkhan surrounding Erdenesant
It is proved with vegetation index value, which varies within relatively large range of 0.26-0.56. The maximum value of fodder resource ( 1200-1300 kg/h ) is observed along the river valley and corresponding value of fodder resource ( 200-300 kh/h ) and vegetation index ( -.26-0.28 ) are mainly in the south western part from the center, where the altitude is no higher than less than 700 meters. The value of vegetation index, corresponding to the fodder resource of 200-300 kg/h for this station is more than other stations. Comparing to other stations, at the station darkhan where measured maximum value of vegetation height ( 17-31 cm ) and biomass of grass ( 310 - 480 kg.hectare ), which corresponds to high value of vegetation index ( 0.31-0.37 ).

It is detected heat source, which is caused due to comparatively large and more populated city influence. There are less classes are delineated on the satellite data than on the resource maps. Thus, we include the station into indirect registable class.

Station Mated
The station is located in dry-steppe of Zuunbulag, covered by steppe and dry-steppe pasture with rock and productivity of 300-400 kg/h (fig7). Beyond this part the area is covered by dry and dry-steppe pasture with productivity of 500-600 kg/h . There is 200 meters difference on relief map over this area.

Fig. 7 Geographic and satellite data maps surrounding Mated surrounding Erdenesant
The region has only unique dark-chestnut soil without charbonate. As well as the surrounding of this station is more homogenous by vegetation fodder resource, olandscape and soil maps, the behaviour correlated. It is proved as the vegetation index has less value 0.21-0.27 at the center, where is naturally observed high temperature 36-390C, than at the edge ( Vi=0, 29-0.35, Ti = 32-350C , than at the edge (Vi=0. 29-0.35, Ti = 32-350C). The result of our analyse allows to include this station into class as registable.

4. Results and disussion
Result of our investigation shows that all meteorological stations are divided into two major classes like the stations directly and indirectly registable by their registration possibility of remotely sensed and ground measurement data. Indirectly registable stations are located mainly in large urban area or in that area through which boundaries of different nature environment classes are crossed.

In case of directly registable stations there is some correlation between vegetation index derived from AVHRR and fodderlation between vegetation index derived from AVHRR and fodder resource obtained from ready map. For certain station the variability of vegetation index and fodder resource has very god correspondence. But the same value of productivity faced different vegetation index value. The example is shown in Table.1. Therefore the estimation of productivity using remotely sensed data is required not only vegetation index values but also there is important to assist the correlation between vegetation index and fodder resource for every vegetation types. Also we found that grass biomass and vegetation height at those stations where is observed wide range of productivity and vegetation index value. But, in future we should make more detail investigation for linking vegetation index and fodder resource map with height and biomass of vegetation measured at ground using high temporal, regular data.

Table 1. The value of vegetation index in correspondence of fodder resource

Radiance temperature and measured land surface temperature good correlated with each other but absolute value of them has some differences. Spatial distribution of temperature depends on vegetation and relief, therefore registration of satellite and ground measured temperature is more effectively on station situated in homogeneous flat area.

As well as we have not developed yet a method for data, to get more stable regression between satellite and ground truth data it is better to select such area, where is able to use the data of registable stations. Therefore to improve effective use of remotely sensed data regularly it is necessary to established new ground regular observation stations in those established new ground regular observation stations in those places, which are in the middle of certain environment class.