The experience of land cover change detection by satellite data

Abstract

Sigificant dependence from climate and anthropogenic influences characterize ecological systems of Kazakhstan. As result of the geographical location of the republic and ecological situation vegetative degradation sites exist throughout the territory of Kazakhstan. The major process of desertification takes place in the arid and semi-arid areas. To allocate spots of stable degradation of vegetation, the transition zone was first identified. Productivity of vegetation in transfer zone is slightly dependent on climate conditions. Multi-year digital maps of vegetation index were generated with NOAA satellite images. According to the result, the territory of the republic was zoned by means of vegetation productivity criterion. All the arable lands in Kazakhstan are in the risky agriculture zone. Estimation of the productivity of agricultural lands is highly important in the context of risky agriculture, where natural factors, such as wind and water erosion, can significantly change land quality in a relatively short time period. We used an integrated vegetation index to indicate land degradation measures to assess the inter-annual features in the response of vegetation to variations in climate conditions from low-resolution satellite data for all of Kazakhstan. This analysis allowed a better understanding of the spatial and temporal variations of land degradation in the country.

Keywords

remote sensing / NOAA / land cover changes / vegetation indexes

Introduction

Remote sensing of the Earth is the most effective method of monitoring the underlying surface state, especially when it comes to the study of vast inaccessible areas, such as the arid and semi-arid regions of Kazakhstan. The Republic of Kazakhstan occupies a large area of 2.7 M km². The majority of this area is located in arid and semi-arid zones and is used for pastures. Currently, there is a desertification problem in these zones, which is connected to climate changes and anthropogenic impact. Space monitoring of the Kazakhstan Republic has been implemented since 2000 in order to register vegetation condition changes and detect desertification zones. To create an effective desertification area identification method, it is necessary to distinguish seasonal vegetation changes, which are caused by climate condition variations, from sustainable vegetation degradation over a long period. There is a need to map the spatial and temporal changes of vegetation cover within Kazakhstan. Long-term remote sensing data are the means for monitoring and mapping changes in vegetation. A set of vegetation indices derived from NOAA/AVHRR imagery were used to describe the state of vegetation cover and changes of Kazakhstan.

Analysis methods

A special method incorporating the integral index of vegetation was developed in order to have an objective assessment of vegetation dynamics. This method includes following procedures (Spivak et al., 2006; Spivak et al., 2008):

1) Processing of NOAA satellite images and referencing of images in Geographic Lat/Lon Projection for the WGS-84 spheroid;

2) Calculating daily values of normalized differential vegetation index (NDVI) using two reflective channels of the AVHRR/NOAA radiometer;

3) Constructing decade (ten days) composite values of NDVI. The calculation of decade (note: “decade” = ten days, in this paper) composites is based on maximal value compositing of NDVI for each a period of ten days. The long-term NDVI series are used for recording and analysis of environmental changes, vegetation states, and traditional land use (Townshend and Justice, 1986; Tucker and.Sellers, 1986; Ichii et al., 2002). The experiments carried out under a USAID research grant showed that the NOAA NDVI decadal maxima values of the entire territory of Kazakhstan well reflect the dynamics of vegetation during the growing season, and can be used to quantify the amount of green biomass (Kogan et al., 1998; Kogan et al., 2003). NDVI values describe the combined effects of natural (long-term) and weather (short-term) influences on vegetation productivity.

4) Calculating integral vegetation index (IVI), an indicator of green biomass, by summing NDVI composites for each vegetative growing season using the following equation:

(1)IVI={\displaystyle \sum _{i=\mathrm{10}}^{\mathrm{27}}NDV{I}_i},

where index “i” is the number of decades counted out from the beginning of the year. Analysis of long-term changes in productivity of vegetation is most effectively performed by using the integrated vegetation index (IVI), which characterizes the total amount of green biomass accumulated during the growing season in each pixel.

5) Calculating the vegetation conditions index (VCI). The term vegetation condition index, introduced by Kogan (1995), provides a numerical approximation of the impact of weather conditions on the productivity of vegetative communities. The value of VCI can be used as a coefficient for the impact of seasonal weather conditions on the amount of terrestrial biomass.

6) Constructing the integral vegetation conditions index (IVCI) for determining inter-annual variations of climate using Eq. (2):

(2)IVCI={\displaystyle \frac{IV{I}_i-IV{I}_{\min }}{IV{I}_{\max }-IV{I}_{\min }}}

where IVI_i is the value of the current decadal composite NDVI in a given pixel; IVI_max represents the maximum long-term value of decadal NDVI composite for the base period in the pixel; IVI_min represents the minimum long-term value of decadal NDVI composite for the base period in the pixel. These vegetation indicators were then mapped for Kazakhstan from 2000 to 2009, using the NOAA/AVHRR data at 1km spatial.

Results and analysis

The distribution of IVCI for the entire period is shown in Fig. 1. It is evident from preceding satellite data that 2002 was the most favorable year for vegetation growth during the considered period, while 2000, 2001, 2003 and 2005 could be marked as seasons with moderate weather conditions. These distributions show an increase in weather stress impacts on vegetation cover in the Republic’s territory since 2004, with maximal deterioration in 2006.

Territory zoning was completed based on digital maps of IVI for the period 2000-2008, and five zones of different productivity of vegetation were detected: Zone A - the high productivity zone; Zone B -the temperate productivity zone; Zone C - the medium productivity zone; Zone D -the low productivity zone; Zone E -the very low productivity zone (desert).

The arrangement of detected zones is represented in Fig. 2. It is shown that the latitudinal sequence of the location of zones with various productivities is connected with the typical vegetative cover and features of the geographical location of the Republic of Kazakhstan.

The changing dynamic of land degradation zones is represented in Fig. 3. The quantitative sizes of zones with high and low productivities essentially depend on seasonal weather conditions. An analysis of inter-seasonal dynamics of different zones’ squares ranked the vegetation seasons according to vegetation productivity. The most favorable year for vegetation productivity was 2002, while the worst was 2006. A significant territory increase in the Zone A, and a corresponding contraction of the Zone E are observed in years with favorable weather conditions. In those years when the vegetative cover is exposed to weather stresses, the spatial desert zone expands while the high productivity zone decreases. Interstitial zones experience less change.

The steady growth of the area of zones with low IVI value (deserts and semi-deserts) is notable. Conducting the zoning of Kazakhstan territory according to vegetative productivity allowed us a) to identify areas with sustainable vegetation degradation caused by variations in weather conditions and b) to assess the productivity of abandoned land in the major grain-producing regions of Northern Kazakhstan.

Discussion

Detection of sites with low levels of vegetative productivity

Desertification processes lead to the degradation of vegetation cover, expressed by a reduction of total biomass and a vegetation type change. Remote sensing data can be used for early detection of foci of desertification, as well as changes in vegetation over large, well-recorded areas from space.

To analyze long duration changes of vegetation productivity occurring in every zone, it is necessary to exclude seasonal weather impacts. For this purpose, a method of detection of “transit zone” with minimal weather impact was implemented. The locations of “transit zones” are shown in Fig. 4. The selected area is one of the most probable locations of seasonal and stable sites of deterioration of vegetation. It is important to note that the area is less than 13% of the total area of Kazakhstan. Areas with the increased anthropogenic influence, such as Semipalatinsk nuclear site, are not taken into account.

Transition zones marked by satellite data correspond to the dry-steppe and steppe agro-climatic zones of the republic defined by ground data (The National Atlas of the Republic of Kazakhstan, 2006). The ongoing degradation of vegetation cover over a long period is one of the most sensitive indicators defined by satellite climatic and anthropogenic influences. The transitional zone is one of the most probable locations of vegetation state degradation induced by natural rather than anthropogenic factors. Application of the method of designating territory of the Republic into zones of varying vegetative productivity according to annual maps of IVI reveals seasonal centers with low levels of vegetation. Next, the localization and dynamics of these zones were studied.

Identification of sites of seasonal vegetation degradation for the entire considered time was accomplished using maps of vegetation indices for moderate years with similar weather conditions, as shown in Fig. 5.

It should be noted that these sites are located in the lower part of the selected transition zone, mainly in the four areas with approximate longitude coordinates: 52°- 54° Е, 62°-64°30′ Е, 65°- 66° Е, 71° -73° Е. Areas with low vegetation productivity increased, thereby decreasing total biomass in the transition zone (Fig. 6.). There is a 5-fold increase in the squares of areas with low vegetation productivity in 2000-2009.

Estimation of agricultural land productivity using satellite vegetation indexes

Non-irrigated spring grain crops are the foundation of agricultural production in Kazakhstan. The Ministry of Agriculture plans to subsidize the increase of sown areas in near future. Estimation of the productivity of agricultural lands is of great importance in the context of risky agriculture, where natural factors (wind erosion, water erosion, etc.) can significantly change the quality of the land in a relatively short time period. Therefore, there is an emerging problem of scientific substantiation when selecting previously abandoned lands for spring crop use. Highly efficient methods for assessing the productivity of land that do not require long-term ground surveys may be developed based on satellite data. Accordingly, the development of methods of analyzing satellite data to evaluate the fertility of agricultural lands is of particular relevance.

Preliminary classification of agricultural land includes the construction of masks of the main types of farmland (the current crop rotation, deposits, natural grass). Masks were created based upon the separated high-resolution scenes (IRS/LISS, 23 m) and on monitoring data with medium resolution (MODIS, 250 m) in the period 2005-2008. Then, the originally obtained masks were adjusted to the resolution of normalized differential vegetation index (NDVI) maps constructed from NOAA satellite data. More details of this technique are described in Spivak et al. (2009). Land productivity estimations were carried out based on the archive of each decade’s quantities of NOAA/AVHRR/NDVI with resolution of 1 km, which have been formed since 2000. This information makes it possible to determine places with relatively high yield and areas of degraded vegetation, which may be connected to many factors- for example, wind erosion of topsoil. Fig. 7 provides an example of masks of land created for various purposes in the Kustanai province in northern Kazakhstan.

Zoning on the above-described 5-point scale was conducted within the distinguished land masks for different purposes. It is notable that there are areas with low levels of productivity in fields with free crop rotation, whereas there are areas with high productivity of vegetation in the abandoned fields.

Next, maps of high and low productivity of arable and abandoned land vegetation are created using multi-year series of remote sensing, in which sites are ranked by productivity. An example of such maps is shown in Fig. 8, where sections belonging to different classes of vegetation productivity are displayed. The received estimates can be used to justify the expansion of cultivated areas and to prioritize the abandoned lands input in the rotation.

Conclusions

The technology of forming long-term series of vegetation indices is complex. It is based on remote sensing data from low resolution AVHRR/NOAA images. Satellite data provides a general integral evaluation of the productivity of large areas of land without the consideration of local landscaping features of individual sites. Zoning the territory of the Republic of Kazakhstan in terms of productivity of vegetation was achieved through the use of long-term distributions of vegetation indices. The location of the zone with the highest probability of identifying sites of stable vegetation degradation was detected. The transition zone selected by using the satellite data coincides with the dry steppe zone of Kazakhstan.

The productivity of un-irrigated agricultural lands depends significantly on seasonal weather conditions. Therefore, in order to achieve an objective estimation of land potential it is necessary to use long-term series of remote sensing data. By using these, it is possible to define zones with consistently high yield as well as zones of stable degradation of vegetative cover, which are associated with desertification. Additionally, high-resolution satellite imagery must be used in order to improve the accuracy of individual fields’ and sites’ estimates.

References

[1] Ichii K, Kawabata A, Yamaguchi Y (2002). Global correlation analysis for NDVI and climatic variables and NDVI trends 1982–1990. Int J Remote Sens, 23(18): 3873–3878

[2] Kogan F (1995). Droughts of the Late 1980s in the United States as derived from NOAA Polar Orbiting Satellite Data. Bul. Amer. Meteor. Soc., 76(5): 655–668

[3] Gitelson A A, Kogan F, Zakarin E, Spivak L, Lebed L (1998). Using AVHRR data for quantitative estimation of vegetation conditions: calibration and validation. Adv Space Research, 22(5): 673–676

[4] Kogan F, Gitelson A, Zakarin E, Spivak L, Lebed L (2003). AVHRR-based spectral vegetation index for quantitative assessment of vegetation state and productivity: calibration and validation. Photogramm Eng Remote Sensing, 69(8): 899–906

[5] Spivak L, Terekhov A, Vitkovskaya I, Batyrbayeva M (2009). Use of long-term satellite data of different resolutions for complex estimation of the state of vegetative cover of Kazakhstan territory. Modern Problems of Remote Sensing of Earth Institute of Space Research of the Academy of Russian Sciences, 52–59

[6] Spivak L, Vitkovskaya I, Batyrbayeva M (2008). Analysis of inter-seasonal variations of productivity of vegetative cover of Kazakhstan using temporal remote sensing rows. News of National Academy of Sciences of Kazakhstan. Phys Math Ser, (4): 29–32

[7] Spivak L, Batyrbayeva M, Vitkovskaya I, Mamedov B, Nurberdiev M, Orlowsky L (2006). Analysis of inter seasonal dynamic of vegetation for Turkmenistan territory. Problems of Desert, (4): 25–29

[8] The National Atlas of the Republic of Kazakhstan (2006). Almaty I, Natural Conditions and Resources.

[9] Townshend J R G, Justice C (1986). Analysis of the dynamics of African vegetation using the normalized difference vegetation index. Int J Remote Sens, 7(11): 1435–1447

[10] Tucker C J, Sellers P J (1986). Satellite remote sensing of primary production. Int J Remote Sens, 7(11): 1395–1416

RIGHTS & PERMISSIONS

Higher Education Press and Springer-Verlag Berlin Heidelberg