Citation Context

... was used to fit the logistic regression, following a backward stepwise procedure. The resulting stepwise model was compared to the ordinary model with a relative operation characteristic (ROC) test (=-=Hanley and McNeil 1982-=-), which checks the equality of the ROC area of each modality. The best-performing model was selected to generate the LD risk map. Model validation The statistical measures ROC and percentage of corre...

Citation Context

...t variable y being 1 given the independent factors x1…xn, ß0 is an intercept of the model, ßi (with 1<0 i<018) are estimated model parameters, which can be interpreted by analyzing odds of the model (=-=Rothman et al. 2008-=-). To avoid multicollinearity among model predictors, variance inflation factors (VIF) were calculated, and correlated factors were removed when VIF exceeded the threshold value of 5 (Belsley et al. 1...

Citation Context

...e probability exceeded a commonly accepted threshold value of 0.5, the cell was marked as degraded land (Manel et al. 1999). In addition, the goodness of fit was evaluated with chi-square statistics (=-=Moore and McCabe 1998-=-). Environ Monit Assess (2013) 185:4775–4790 4781 Results Linear trend analysis The resulting ΣNDVI-based trend map highlights areas that experienced constant vegetation losses during the monitoring d...

Citation Context

...p of LD. Semi-structured interviews were held with 119 farmers from seven districts of Khorezm in September 2009. For the survey, a combination of purposive and snowball sampling methods was applied (=-=Kumar 1999-=-), i.e., initially randomly sampled respondents were asked to nominate another farmer in the neighborhood who had been farming for at least several years, possibly on degraded land. Linear trend analy...

Citation Context

...02) and composed of the best observations during 16-day periods with regard to overall pixel quality (aerosol content, low view angle, and absence of clouds/cloud shadows) and observational coverage (=-=Justice et al. 2002-=-). The data were pre-processed by (1) identifying and removing low-quality pixels based on the data quality flags specified in MOD13Q1, (2) filling data gaps with linear interpolation, and (3) smoothi...

Citation Context

...sampling with a 3×3 cell window (750 m×750 m). Sampling was applied to minimize the impact of spatial dependency between observations, which might cause unreliable estimation of the model parameters (=-=Irwin and Geoghegan 2001-=-). The sample was equally divided into calibration and validation datasets. The former was used to fit the logistic regression, following a backward stepwise procedure. The resulting stepwise model wa...

Citation Context

...ving low-quality pixels based on the data quality flags specified in MOD13Q1, (2) filling data gaps with linear interpolation, and (3) smoothing images with an adaptive Savitsky–Golay filter (Fig. 2; =-=Jönsson and Eklundh 2002-=-). During the smoothing procedure, the data quality flags were applied to weigh the data; higher weights were assigned to higher-quality pixels, while lower-quality data had a minor influence on the c...

Citation Context

...lities. In the case of the PCP, the modeled degradation was assigned to the pixels, i.e., if the probability exceeded a commonly accepted threshold value of 0.5, the cell was marked as degraded land (=-=Manel et al. 1999-=-). In addition, the goodness of fit was evaluated with chi-square statistics (Moore and McCabe 1998). Environ Monit Assess (2013) 185:4775–4790 4781 Results Linear trend analysis The resulting ΣNDVI-b...

Citation Context

...performing model was selected to generate the LD risk map. Model validation The statistical measures ROC and percentage of correct predictions (PCP) were calculated to evaluate the model performance (=-=Christensen 1997-=-). The ROC value ranges from 0 to 1, where 1 indicates a perfect fit, 0.5 indicates a random fit, while values between 0.5 and 1 show some association between dependent and independent variables (Pont...

Citation Context

...l. 2011) and agricultural ecosystems similar to the presented case (Tottrup and Rasmussen 2004; Fuller 1998). Degradation in drylands manifests itself in the reduced productive potential of the land (=-=Reynolds et al. 2007-=-). In arid and semi-arid areas, the sum of the NDVI over a growing season (∑NDVI) is strongly correlated with the vegetation production (Nicholson et al. 1998), revealing that a decreasing linear tren...

Citation Context

... reduced productive potential of the land (Reynolds et al. 2007). In arid and semi-arid areas, the sum of the NDVI over a growing season (∑NDVI) is strongly correlated with the vegetation production (=-=Nicholson et al. 1998-=-), revealing that a decreasing linear trend is a good indicator of the vegetation loss for an early warning of LD (Wessels et al. 2004; Budde et al. 2004). Fig. 1 Location of study area in the Khorezm...

Citation Context

...1997). The ROC value ranges from 0 to 1, where 1 indicates a perfect fit, 0.5 indicates a random fit, while values between 0.5 and 1 show some association between dependent and independent variables (=-=Pontius and Schneider 2001-=-). The PCP is defined as the percentage of correctly predicted pixels to the total number of pixels in the map. For validation, the final model was applied to the validation dataset, and the probabili...

Citation Context

...n is that the medium-scale data, for example from the AQUA/TERRA-Moderate Resolution Imaging Spectroradiometer (MODIS), until recently did not cover sufficiently long periods to allow trend analyses (=-=Fensholt and Proud 2012-=-; Prince et al. 2009). The comparatively higherscale images from the LANDSAT program, recorded since 1972, are not always in place for all geographical areas, e.g., Central Asia, on the frequent and r...

Citation Context

...n be captured by trend analyses of multi-year satellite images (Lambin and Linderman 2006; Udelhoven 2011). Trend analyses were routinely employed for LD assessment, using coarse-scale imagery (e.g., =-=Wessels et al. 2007-=-). Only a limited number of studies used satellite time series of medium and high spatial resolution, which are likely to be more appropriate for monitoring of fragmented landscapes of drylands (Sonne...

Citation Context

...ries has been used to effectively describe a vegetation trend in natural environments (Sonnenschein et al. 2011) and agricultural ecosystems similar to the presented case (Tottrup and Rasmussen 2004; =-=Fuller 1998-=-). Degradation in drylands manifests itself in the reduced productive potential of the land (Reynolds et al. 2007). In arid and semi-arid areas, the sum of the NDVI over a growing season (∑NDVI) is st...

Citation Context

... the main research focus was on LD. The quality of the calculated trends can be evaluated by assessing the impact of the errors associated with a particular image and its position in the time series (=-=Hostert et al. 2003-=-). The effect is greater at the beginning and end of the time series, while errors in the middle of the series have less influence on the direction of the trend (Röder et al. 2008). To test an impact ...

Citation Context

...st of these studies focused on natural and semi-natural dryland environments (e.g., Röder et al. 2008). Much less attention was paid to the remotesensing monitoring of LD in agricultural areas (e.g., =-=Tottrup and Rasmussen 2004-=-), where LD trends can be masked by land management practices, such as irrigation and fertilizer application. The above-mentioned LULC analyses are suitable for describing spatial changes in land-use ...

Citation Context

...Zhang et al. 2008; Biro et al. 2010). This approach implies, among others, image classification (Gao and Liu 2008), spectral mixture modeling (Tromp and Epema 1998), and principal component analysis (=-=Chikhaoui et al. 2005-=-), followed by spatial comparison of the derived maps to quantify changes in degradation classes (Li et al. 2007). For example, Collado et al. (2002) applied spectral mixture modeling to bi-temporal L...

Citation Context

...owed by summer crops (Khalikov and Tillaev 2006). On the area of land that is not assigned to cotton, and following the winter wheat, farmers grow maize, sorghum, watermelons, melons, and vegetables (=-=Conrad et al. 2007-=-). The irrigation water is supplied from the Amu Darya River via a dense network of 16,000 km of irrigation canals. Seasonal salt leaching is practiced for coping with soil salinization and is support...

Citation Context

...ons (low-lying location, relief flatness), the shallow water tables result from losses from the irrigation network; the situation is exacerbated by the sub-optimal performance of the drainage system (=-=Ibrakhimov et al. 2007-=-). Data sources The data used in the study include (1) raster data: MODIS images (MOD13Q1, https://lpdaac.usgs.gov/) and (2) vector data: LULCmaps for 2001–2009 derived from 250 m MODIS data, infrastr...

Citation Context

...e data, for example from the AQUA/TERRA-Moderate Resolution Imaging Spectroradiometer (MODIS), until recently did not cover sufficiently long periods to allow trend analyses (Fensholt and Proud 2012; =-=Prince et al. 2009-=-). The comparatively higherscale images from the LANDSAT program, recorded since 1972, are not always in place for all geographical areas, e.g., Central Asia, on the frequent and repeatable basis requ...

Citation Context

...correlated with the vegetation production (Nicholson et al. 1998), revealing that a decreasing linear trend is a good indicator of the vegetation loss for an early warning of LD (Wessels et al. 2004; =-=Budde et al. 2004-=-). Fig. 1 Location of study area in the Khorezm region, Uzbekistan. The region’s border coincides with the extent of irrigated land 4778 Environ Monit Assess (2013) 185:4775–4790 The time series of ND...

Citation Context

...ually conducted at a small scale (Gao and Liu 2010). Mapping of LD can be achieved by studying spatio-temporal dynamics of land-use and land-cover (LULC) changes (e.g., Kessler and Stroosnijder 2006; =-=Lu et al. 2007-=-; Zhang et al. 2008; Biro et al. 2010). This approach implies, among others, image classification (Gao and Liu 2008), spectral mixture modeling (Tromp and Epema 1998), and principal component analysis...

Citation Context

... 2007). Only a limited number of studies used satellite time series of medium and high spatial resolution, which are likely to be more appropriate for monitoring of fragmented landscapes of drylands (=-=Sonnenschein et al. 2011-=-). The reason is that the medium-scale data, for example from the AQUA/TERRA-Moderate Resolution Imaging Spectroradiometer (MODIS), until recently did not cover sufficiently long periods to allow tren...

Citation Context

...of quantifying gradual degradation processes within one land-use class (Röder et al. 2008). Such processes can be captured by trend analyses of multi-year satellite images (Lambin and Linderman 2006; =-=Udelhoven 2011-=-). Trend analyses were routinely employed for LD assessment, using coarse-scale imagery (e.g., Wessels et al. 2007). Only a limited number of studies used satellite time series of medium and high spat...

Citation Context

...garded as binary, where values 1 and 0 were used to denote its presence and absence, respectively. Consequently, the probability of LD occurring was computed with a logistic regression model (Eq. 1) (=-=Hosmer and Lemeshow 2000-=-): PðyÞ 1=1þ exp b0þ Pn i1 bixið Þ ð1Þ where P(y) is probability of the dependent variable y being 1 given the independent factors x1…xn, ß0 is an intercept of the model, ßi (with 1<0 i<018) are e...

Citation Context

... 220100 Urgench, Uzbekistan ASB, the expansion of the agricultural areas from about 4.5 Mha in 1960 to almost 7.9 Mha by 1999 was made possible through the construction of massive irrigation systems (=-=Saiko and Zonn 2000-=-). The land appropriation of irrigated agriculture, mostly for cultivation of the water-consuming cash crop cotton, took its toll on the natural land and water resources in the region. The result is t...

Citation Context

...namics of land-use and land-cover (LULC) changes (e.g., Kessler and Stroosnijder 2006; Lu et al. 2007; Zhang et al. 2008; Biro et al. 2010). This approach implies, among others, image classification (=-=Gao and Liu 2008-=-), spectral mixture modeling (Tromp and Epema 1998), and principal component analysis (Chikhaoui et al. 2005), followed by spatial comparison of the derived maps to quantify changes in degradation cla...

Citation Context

...chniques to detect LD trends, remote-sensing provides a cost-effective evaluation over large areas, whereas in-situ process studies are resource demanding and thus usually conducted at a small scale (=-=Gao and Liu 2010-=-). Mapping of LD can be achieved by studying spatio-temporal dynamics of land-use and land-cover (LULC) changes (e.g., Kessler and Stroosnijder 2006; Lu et al. 2007; Zhang et al. 2008; Biro et al. 201...

Citation Context

...ity only becomes obvious in the long run, as annual leaching practices counterbalance the salinization process. The importance of water management for the distribution of soil salinity at farm scale (=-=Akramkhanov and Vlek 2011-=-) was also indicated by the results of the farmer survey, thus confirming the significance of this factor at the local level. Most of the surveyed farmers attributed declined yields not only to soil s...

Citation Context

... and smoothed 16-day, 250 m MODIS NDVI time series of one pixel Environ Monit Assess (2013) 185:4775–4790 4779 experts and a review of literature for the Khorezm region (e.g., Ibrakhimov et al. 2007; =-=Akramkhanov et al. 2011-=-). The identified factors ranged from soil, groundwater, and relief characteristics to land ownership and management. The main factors for which the data were available served as inputs to the logisti...

Citation Context

...which were assumed to represent the most important factors influencing the spatial distribution of LD. Incorporation of more variables was subject to data constraints, a common issue for LULC models (=-=Dubovyk et al. 2011-=-). Aiming to provide a regional overview, the derived risk map renders a visual representation of areas under risk that could be prioritized in more detailed analyses and the attention of decision mak...

Citation Context

...e natural sandy deserts Karakum and Kyzylkum in the south and east, and belongs to the Central Asian semi-desert zone with an extreme, continental climate. The annual precipitation, averaging 100 mm (=-=Tischbein et al. 2012-=-), falls mostly outside the crop-growing season (April–October) and is greatly exceeded by annual evaporation (Conrad et al. 2012). Thus, crop production entirely depends on irrigation. The irrigated ...

Citation Context

...n entirely depends on irrigation. The irrigated cropland extends over 270,000 ha with an average field size of 2.59 ha. The same crop often grows on the adjacent fields, exceeding the area of 250 ha (=-=Conrad et al. 2011-=-). Most of the arable land is occupied by cotton (60–70 %) and winter wheat (20–30 %) (Shi et al. 2007), cultivated under the state procurement system, where production goals are set for these strateg...

Citation Context

... land degradation in the Khorezm region of Uzbekistan. The percentages indicate the frequency of the farmers’ replies Environ Monit Assess (2013) 185:4775–4790 4785 winter wheat in the cropland area (=-=Djanibekov et al. 2010-=-; Shi et al. 2007). Furthermore, summing the NDVI over the whole growing season, thus integrating vegetation peaks in the fields with different land uses, reduced the possibility of misinterpretations...

Citation Context

... were abandoned. Recent studies in the region show that the alternative use of these areas for afforestation with native tree species could increase the productive and economic potential of the land (=-=Djanibekov et al. 2012-=-; Khamzina et al. 2012). The results of the accuracy assessment confirm the validity of the elaborated approach, suggesting its applicability to regional LD monitoring. The robustness of the calculate...

Citation Context

... resource demanding and thus usually conducted at a small scale (Gao and Liu 2010). Mapping of LD can be achieved by studying spatio-temporal dynamics of land-use and land-cover (LULC) changes (e.g., =-=Kessler and Stroosnijder 2006-=-; Lu et al. 2007; Zhang et al. 2008; Biro et al. 2010). This approach implies, among others, image classification (Gao and Liu 2008), spectral mixture modeling (Tromp and Epema 1998), and principal co...

Citation Context

...ctral mixture modeling (Tromp and Epema 1998), and principal component analysis (Chikhaoui et al. 2005), followed by spatial comparison of the derived maps to quantify changes in degradation classes (=-=Li et al. 2007-=-). For example, Collado et al. (2002) applied spectral mixture modeling to bi-temporal Landsat images to delineate areas affected by desertification in Argentina. Chen and Rao (2008) utilized 3-year L...

Citation Context

...ars between 2000 and 2010 and averaged over the eleven years. Logistic regression Coupled with GIS, logistic regression is an appropriate tool for explanatory analysis of the factors of LULC changes (=-=Menz et al. 2010-=-). We applied this model to quantify the contribution of the LD factors and to identify areas at risk of LD. Spatial distribution of LD was explained as a function of these factors (Table 2). The natu...

Citation Context

...icator, introduced in the Soviet Union (Karmanov 1980) and still relevant in a number of ASB states, to assess the land suitability for cropping, using cotton as the reference crop in the assessment (=-=Ramazonov and Yusupbekov 2003-=-). It is an aggregate of several parameters, including field characteristics and soilinherent properties, e.g., soil texture, organic matter content. Values range from 0 to 100 points with values <40 ...

Citation Context

... land and water resources in the region. The result is the Aral Sea shrinkage due to irrigation water withdrawal from the tributary rivers, rising groundwater tables and subsequent soil salinization (=-=Spoor and Krutov 2003-=-). The land degradation (LD) due to salinity has plagued about 75 % of the irrigated area of the ASB (van Dijk et al. 1999), leading to reduced productivity of the arable land and, eventually, its wit...

Citation Context

...es (e.g., Kessler and Stroosnijder 2006; Lu et al. 2007; Zhang et al. 2008; Biro et al. 2010). This approach implies, among others, image classification (Gao and Liu 2008), spectral mixture modeling (=-=Tromp and Epema 1998-=-), and principal component analysis (Chikhaoui et al. 2005), followed by spatial comparison of the derived maps to quantify changes in degradation classes (Li et al. 2007). For example, Collado et al....

Citation Context

...at a small scale (Gao and Liu 2010). Mapping of LD can be achieved by studying spatio-temporal dynamics of land-use and land-cover (LULC) changes (e.g., Kessler and Stroosnijder 2006; Lu et al. 2007; =-=Zhang et al. 2008-=-; Biro et al. 2010). This approach implies, among others, image classification (Gao and Liu 2008), spectral mixture modeling (Tromp and Epema 1998), and principal component analysis (Chikhaoui et al. ...

Citation Context

...udy, the LD trend was analyzed based on a decrease in the vegetation cover which, in land-use systems, may also occur due to changes in land 4784 Environ Monit Assess (2013) 185:4775–4790 management (=-=Bai and Dent 2009-=-). In Uzbekistan, the land-use decisions largely remained unchanged during the study period given the area-based, state production targets for cotton and the prevalence of cotton and Fig. 5 Risk map o...

Citation Context

...ao and Liu 2010). Mapping of LD can be achieved by studying spatio-temporal dynamics of land-use and land-cover (LULC) changes (e.g., Kessler and Stroosnijder 2006; Lu et al. 2007; Zhang et al. 2008; =-=Biro et al. 2010-=-). This approach implies, among others, image classification (Gao and Liu 2008), spectral mixture modeling (Tromp and Epema 1998), and principal component analysis (Chikhaoui et al. 2005), followed by...

Citation Context

...e, continental climate. The annual precipitation, averaging 100 mm (Tischbein et al. 2012), falls mostly outside the crop-growing season (April–October) and is greatly exceeded by annual evaporation (=-=Conrad et al. 2012-=-). Thus, crop production entirely depends on irrigation. The irrigated cropland extends over 270,000 ha with an average field size of 2.59 ha. The same crop often grows on the adjacent fields, exceedi...

Citation Context

...studies in the region show that the alternative use of these areas for afforestation with native tree species could increase the productive and economic potential of the land (Djanibekov et al. 2012; =-=Khamzina et al. 2012-=-). The results of the accuracy assessment confirm the validity of the elaborated approach, suggesting its applicability to regional LD monitoring. The robustness of the calculated LD trend was compara...

Citation Context

... soil salinization and is supported by the 8,000 km network of drainage collectors. Irrigation water supply to crop fields is determined according to the standard guidelines originating in the 1960s (=-=Rakhimbaev et al. 1992-=-). Gravity irrigation prevails in the flat topography of the region, where slopes do not exceed 10 %. Water pumping from irrigation canals is practiced in the elevated areas on the southwestern border...

Citation Context

...ze of 2.59 ha. The same crop often grows on the adjacent fields, exceeding the area of 250 ha (Conrad et al. 2011). Most of the arable land is occupied by cotton (60–70 %) and winter wheat (20–30 %) (=-=Shi et al. 2007-=-), cultivated under the state procurement system, where production goals are set for these strategic crops. Cotton has always been produced in Uzbekistan as a means of gaining export earnings, whereas...

Citation Context

...S$ 12 million (World Bank 2002). The international concern about LD in Central Asia led to the launch of the Subregional Action Programme for the Central Asian Countries on Combating Desertification (=-=UNCCD 2003-=-). Despite the recognized severity of the problem, information on the long-term changes in the state of the land has yet to be collected to develop a common assessment and monitoring system in the reg...

Citation Context

...hman et al. 2008). To avoid multicollinearity among model predictors, variance inflation factors (VIF) were calculated, and correlated factors were removed when VIF exceeded the threshold value of 5 (=-=Belsley et al. 1980-=-). The sample size for logistic regression of 8,112 observations resulted from the systematic unbalanced random sampling with a 3×3 cell window (750 m×750 m). Sampling was applied to minimize the impa...

Citation Context

... in Uzbekistan as a means of gaining export earnings, whereas wheat was introduced in the 1990s for national wheat self-sufficiency. Cotton can be rotated with winter wheat, followed by summer crops (=-=Khalikov and Tillaev 2006-=-). On the area of land that is not assigned to cotton, and following the winter wheat, farmers grow maize, sorghum, watermelons, melons, and vegetables (Conrad et al. 2007). The irrigation water is su...

Citation Context

..., but they are not capable of quantifying gradual degradation processes within one land-use class (Röder et al. 2008). Such processes can be captured by trend analyses of multi-year satellite images (=-=Lambin and Linderman 2006-=-; Udelhoven 2011). Trend analyses were routinely employed for LD assessment, using coarse-scale imagery (e.g., Wessels et al. 2007). Only a limited number of studies used satellite time series of medi...