Drought and Vegetataion Monitoring in the Arid and Semi-Arid Regions of the Mongolia using Remote Sensing and Ground Data
NDVI and the LST maximum value composites (MVC; Holben, 1986) were produced from all the images available within the monthly periods for the four study years. The MVC minimizes the effects of atmosphere, scan angle, and cloud contamination (Lambin and Ehrlich, 1996).
In this study we were tested NDVI-LST space based drought indicators which are developed by Karnieli (1999). Also we were developed drought monitoring indicators based on normalization method in the NDVI-LST space of a local area.
Figure 1. Dryness zone classification of the Mongolia (based Kharin et. al.,2000 and
Shiirevdanba, 1998). Thirteen plots position are used for drought estimation.
4. Analysis and Results
The warm seasonally (from April to September) temporal analysis of the arid and the semi-arid areas in the NDVI-LST space were performed on the NDVI and LST data sets for the selected four years. In the summer of 1992 and in the summer and fall of 1995 were severe droughts within the whole territory of the Mongolia. The average summer air temperatures were 19oC and 18oC in the 1992 and 1995, respectively. The average total rainfalls were 52 and 76 mm in the arid area and 100 and 114 mm in the semi-arid area in the summer of 1992 and 1995, respectively. In 1996, the worst year in Mongolian history, altogether 417 forests and steppe fires (over 10.5 million hectare) occurred due to dryness - as effects of sever drought which occurred in the fall of 1995. The other two years were relatively wet years-1993 and 1994 with rainfalls 100 and 75 mm in the arid and 155 and 142 mm in the semi-arid zone, respectively. From the warm months temporal variations of the NDVI and LST on the some plots of the arid and semi-arid zones presented in figure 2 and it might be noticed that the a lower NDVI and a higher LST values occurs on the dry years and a higher NDVI and a lower LST values occur on the wet years, respectively, on the both arid and semi-arid areas.
In order to discover any relations in the warm seasonal NDVI, LST and total rainfall the correlation coefficients were performed in the different dryness zones (Table 1). As can be seen from table 1, the correlations between of those parameters vary in throughout the dryness zones. On this table, two directions of warm seasonal variations of the NDVI, LST and total rainfall values: (i) for arid zone, characterized by a both negative relations of the NDVI with LST and rainfall, respectively and (ii) for other dryness zones, a relations of NDVI-LST and NDVI-rainfall are positive and increases from semi-arid zone (low latitude) to humid zone (high latitude). But only the correlation of LST and Rainfall were almost equal and positive in whole territory, during the vegetation growing period. An interesting conclusion from table 1 and figure 2 is that, in the whole territory, most of the temporal variability during the warm period in ecological characteristics depends a local condition (i.e., land cover type, landform, local climate, latitude).
Figure 2. Warm months temporal variations of the NDVI and LST on the sample
plots correspond to arid (A) and the semi-arid (B) zones.
Table 1. Correlation coefficients between the NDVI, LST and the total Rainfall on the different dryness zones in the warm (April to September) seasons.
| Conditions
|
Dryness zone
|
| Arid
|
Semi-Arid
|
Sub-Humid
|
Humid
|
| Wet years (1993, 1994)
|
NDVI and LST
|
-0.10
|
0.35
|
0.36
|
0.50
|
| NDVI and Rainfall
|
-0.20
|
0.36
|
0.54
|
0.76
|
| LST and Rainfall
|
0.75
|
0.61
|
0.51
|
0.59
|
| Drought years (1992, 1995)
|
NDVI and LST
|
-0.35
|
0.14
|
0.44
|
0.56
|
| NDVI and Rainfall
|
-0.27
|
0.17
|
0.57
|
0.73
|
| LST and Rainfall
|
0.69
|
0.69
|
0.66
|
0.74
|
Most suitable method for estimate local a condition in any case is establishing a "normal" value. We may considered that the NDVI and LST of 1993 are used as a "normal" years and calculated as:
NDVIdifferenced =[NDVI1993 - NDVIi]/NDVI1993
LSTdifferenced =[LST1993 - LSTi]/LST1993
where NDVI
differenced and LST
differenced are subtracted values from a year of normal precipitation -1993. NDVI
i and LST
i are values of any particular year. In order to characterize the drought years, warm seasonal time trajectories of the "differenced" two products were applied for other drought occurred and wet years and some selected points from arid and semi-arid area illustrated in the Figure 3. Due to moisture stress on the vegetation, NDVI/LST values recorded in the dry years should be lower/higher than those values recorded in a "normal" year; therefore drought occurred areas will have high/low NDVI
differenced / LST
differenced values. Bottom part of Figure 3 shows that the mean values and the line connected between the maximum LST
differenced with minimum NDVI
differenced and minimum LST
differenced with maximum NDVI
differenced values. It might be seen that the drought occurred years, 1992 and 1995; lines and mean values are very similar terms of position and angle in the both arid and semi-arid zones. The mean value and the line - connected maximum and minimum values of the calculated NDVI
differenced and LST
differenced were used as indicators for detecting and monitor drought events. The lines of the drought years'
