Next Article in Journal
Expression and Functional Analysis of the PaPIP1-2 Gene during Dormancy and Germination Periods of Kernel-Using Apricot (Prunus armeniaca L.)
Previous Article in Journal
YOLOv5-ACS: Improved Model for Apple Detection and Positioning in Apple Forests in Complex Scenes
Previous Article in Special Issue
Hyperspectral Analysis and Regression Modeling of SPAD Measurements in Leaves of Three Mangrove Species
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Forests
Volume 14
Issue 12
10.3390/f14122305
forests-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Monitoring the Condition of Wetlands in the Syr Darya Floodplain—How Healthy Are the Tugai Forests in Kazakhstan?

by
Christian Schulz
* and
Birgit Kleinschmit
Geoinformation in Environmental Planning Lab, Technische Universität Berlin, Straße des 17. Juni 145, 10623 Berlin, Germany
*
Author to whom correspondence should be addressed.
Forests 2023, 14(12), 2305; https://doi.org/10.3390/f14122305
Submission received: 22 September 2023 / Revised: 13 November 2023 / Accepted: 21 November 2023 / Published: 24 November 2023
(This article belongs to the Special Issue Restoration and Monitoring of Forested Wetlands and Salt Marshes)
Browse Figures
Versions Notes

Abstract

:
Tugai wetlands, including the forests of Populus euphratica Oliv. and P. pruinosa Schrenk, are major biodiversity hotspots within semi-arid and arid ecozones. However, for over a century, Central Asian river systems have been severely affected by dam regulation, water withdrawals for large-scale irrigated agriculture, and deforestation. To support sustainable use and protection of this threatened forest type, we provide information on the distribution and degradation status of Tugai wetlands in the Syr Darya floodplain using Normalized Difference Vegetation Index (NDVI) time series from Landsat 7 and Moderate Resolution Imaging Spectroradiometer (MODIS). An accuracy assessment confirmed the validity of the MODIS-based wetland map, with an overall accuracy of 78.6%. This was considerably better than the Landsat product, mainly due to the greater temporal frequency of the MODIS time series. We further calculated trends and breakpoints between 2001 and 2016 using the BFAST algorithm. We found negative trends for nearly a third of the wetlands. Breakpoint detection showed major stress events in the years 2001, 2009, and 2016. Our study revealed the temporal and spatial distribution and vitality of an endangered forest ecosystem that has rarely been studied thus far. Climate change may accelerate the destabilization of the Tugai forests at the Syr Darya floodplain.

1. Introduction

Tugai forests, often referred to as desert riparian forests [1] or Central Asian riparian floodplain forests [2], are biodiversity hotspots in the arid ecozones of Eurasia [3]. The term Tugai stands for a composition of wetlands dominated by Euphrates poplar (Populus euphratica Oliv.) or Desert poplar (Populus pruinosa Schrenk), mixed with shrub and reed communities consisting of Tamarix, Elaeagnus, Salix, and Phragmites [3,4]. As a component of the UNESCO and RAMSAR sites in Central Asia, they are of great importance for international wetland conservation [5]. They are highly relevant for the local populations’ well-being as they provide a wide range of ecosystem services, such as the provision of food, fiber, and fodder, as well as fulfilling aesthetic, educational, and recreational functions [3,6]. In addition, the Tugai forests help to combat desertification by regulating climate and flooding and by reducing wind erosion.
Historically, this forest type was mainly distributed in Central Asia (30.8%) and in China (60.9%) [4]. Today, Tugai forests are predominantly found in five global terrestrial ecoregions [7] (Table A1): (a) Central Asian Riparian Woodlands (Amu Darya, Syr Darya, and Ili River); (b) Tarim Basin Deciduous Forests and Steppe (Tarim River); (c) Gobi Lakes Valley Desert Steppe (Ekhiin-Gol Oasis); (d) Alashan Plateau Semi-Desert (Heihe Oasis); and (e) Junggar Basin Semi-Desert (Boro-Tala River). The health status of the remaining Tugai forests is highly critical. Recent studies have addressed overgrazing [8], the loss of genetic diversity [7,9,10], the loss of generative reproduction [11], insect infestation [2], salinization [4], and droughts [12]. Progressive degradation caused by climate change, land use change, and habitat fragmentation may lead to an ecosystem tipping point, where changes increase rapidly and may be irreversible [13,14,15].
Although many of these threats are known, protection and restoration efforts are still insufficient. Thus, large-scale management strategies and controls are of great importance. However, monitoring the remaining Tugai forests through site visits is challenging due to their dispersed geographical distribution. Hence, spatial and temporal information obtained from optical satellite imagery could substantially support monitoring tasks.
With free access to long-term archives from different satellite missions, time-series-based analysis has become an essential tool in the remote sensing of forests [16,17,18]. For instance, Hansen et al. [19] produced annual woodland distribution maps from Landsat data that are used for change detection. Land cover products from other sources also cover different woodland classes [20,21,22]. However, these global datasets tend to underestimate forests in drylands [23]. Also, local studies dealing with land cover changes, vegetation dynamics, and forest distribution in Central Asia [24,25,26,27,28,29,30,31,32] omit the Tugai wetlands.
Because of the limitations of pre-existing maps derived from satellite-based image time series, novel methods for mapping and monitoring the Tugai forests in Central Asia are urgently needed. Vegetation changes can be captured on many spatial–temporal scales and resolutions using remote sensing products [33,34,35]. Numerous tools and algorithms have been developed, e.g., phenological indices [36], anomaly identification [37], and the calculation of long-term vegetation trends, shifts, and breaks [38]. In the highly degraded Tugai forests at the Tarim Basin, the Breaks For Additive Season and Trend (BFAST) algorithm [38] has been successfully applied to map forest disturbances and breakpoints in vegetation signal trends [39]. By using a dense Landsat time series spanning 14 years, the authors were able to unveil important relationships between water availability and the forest’s vitality [39,40]. However, such time series approaches have not yet been applied in our study region.
Our study aims to monitor the health status of the Tugai forests in the Syr Darya floodplain with optical time series data. The main objectives of our study were: (1) to compare the suitability of Landsat 7 (L7) and Moderate Resolution Imaging Spectroradiometer (MODIS) data for mapping wetlands in Central Asian drylands and (2) to capture the degradation status of the remaining Tugai forests at the Syr Darya floodplain.

2. Materials and Methods

2.1. Study Area

With a catchment area of 783,000 km2 and a length of 3018 km, the Syr Darya is the second largest river in Central Asia and supplies large parts of the Aral Sea Basin with water [41]. Our study area (Figure 1) comprised the lower and middle reaches of the Syr Darya from the Chardara Reservoir (236 m above sea level) to the river delta close to the North Aral Sea (45 m above sea level). Overall, our study area covered 53,326 km2 and spanned a river length of 1350 km. We selected the study site for its known distribution of Populus euphratica and P. pruinosa, which mainly exist in the riparian zones of the Syr Darya flatlands [42]. According to the MODIS-based land cover classification by Klein et al. [43], the major land cover classes at the study area in 2009 were sparse vegetation (28.2%), irrigated (22.7%), grassland (18.7%), and bare area with salt flats (14.9%). The broadleaved class, which mostly consists of riparian forests, was estimated to cover only 4.8% of the study area. The arid climatic conditions, characterized by hot summers and cold winters, correspond to the cold arid desert (BWk) and cold arid steppe climates (BSk), according to the Köppen–Geiger Climate Classification [44]. Most run-off comes from melting water in the upper mountain regions. As the average temperature in Central Asia is predicted to rise faster than the global average [39], increasing amounts of melting water from the mountain glaciers can be expected. Nonetheless, the climate in the study area is expected to be generally drier and hotter, which leads to desertification and threatens the local population.
With the building of extensive canal networks and river dams during the Soviet era (Chardara Reservoir, Aydar Reservoir, and Kayrakkum Reservoir), large-scale irrigation farming was introduced in the Syr Darya catchment region [45]. The regulation of water by river dams since the 1960s has fundamentally changed the natural run-off of the Syr Darya, resulting in reduced water quantities during the vegetation season. Additionally, after the independence of the bordering countries, energy production from the dams during the winter season resulted in abrupt water release, leading to overflow events in certain years [45,46]. As a consequence, the new Koksaray Reservoir in Kazakhstan was built in 2008 for water storage and to prevent the riparian zones from large-scale damage caused by further flooding events. With the introduction of irrigated agriculture and river water control by dams, large amounts of water became unavailable for natural wetlands, which has led to long-term changes and degradation since the Soviet era. Today, irrigated agriculture for cotton production is still the major use of land in the region [47]. The agricultural zones (i.e., cultivated wetlands) comprise complex patterns of active and abandoned irrigation cropland, reeds, small lakes, and shrublands. More pristine wetlands with different types of degradation can be found closer to the main stream of the river. Figure 2 shows different vegetation types found in the study area.

2.2. Data

2.2.1. Satellite Imagery

Since area-wide and high resolution satellite imagery in our study area is missing from the 1980s and 1990s, our study time frame starts with the year 2000, for which data from the multispectral MODIS and L7 missions are available. The MODIS mission, operated by the National Aeronautics and Space Administration (NASA), was carried out by two satellites: Terra (1999) and Aqua (2002). These satellites incorporate 36 spectral bands, including visible and infra-red light spectra [48]. It was designed to collect continuous global data for long- and short-term changes in the Earth system. It provides imagery with spatial resolutions of 250 m (bands 1–2), 500 m (bands 3–7), and 1000 m (bands 8–36) [48]. The L7 mission is part of the Landsat Earth observation program, which has been run by NASA since 1972, and is operated by a single satellite with the Enhanced Thematic Mapper Plus (ETM+) instrument, incorporating seven spectral bands in the visible and infra-red ranges [49]. The spatial resolution of L7 imagery reaches 30 m (bands 1–5, 7) and 60 m (band 6) [49].
We chose to use NDVI time series for our study as it includes spectral information of near-infrared and red light spectra, which are sensitive to chlorophyll at vegetated surfaces [50]. Hence, biophysical parameters can be used to describe vegetation stress and phenological changes on the spatial and temporal levels. For time series application in vegetation monitoring, Mayr et al. [51] discovered that studies using NDVI products from different multispectral sensors were frequently found to be robust. Using NDVI data from the MODIS and L7 missions allows us to obtain complementary information about the wetland distribution and to evaluate the suitability of time-series-based methods across different spatial and temporal resolutions.
Google Earth Engine (GEE) [52] provides equally preprocessed 8-day NDVI time series products from both sensors. We used the cloud computing platform and its JavaScript-based application programming interface (API) for downloading time series vectors for the study area between 2001 and 2016 (Table 1). We post-processed the tabular data with R 3.4.1 [53]. Our methodological workflow is outlined in Figure 3. First, we created a 250 m regular cell grid for the study area, resulting in a list of 840,000 points, and downloaded the NDVI time series vectors of MODIS and L7 from GEE. Second, we excluded observations outside the vegetation period from March to October. This step reduced the percentage of observations impacted by clouds and snow, which mostly appear in the study area during winter. Consequently, the mean observation numbers per grid cell were 192 for the L7 data and 444 for the MODIS data (Figure 4). These time series data were used for wetland mapping and mask creation. Third, for trend and breakpoint detection using the BFAST algorithm, gap free time series data was mandatory [38]. Thus, the remaining gaps in the time series, e.g., caused by cloudy observations, were filled by linear interpolation. The final datasets included equidistant and gap-free 8-day NDVI time series vectors with 30 time steps per year.

2.2.2. Validation Data

For the derivation of ground truth data, we searched for validation sampling sites at the lower and middle reaches of the Syr Darya in a field campaign in 2017. We decided on three 1500 km2 large regions, which are displayed in the study area map (Figure 1): (a) The sampling region Kyzylorda is characterized by large-scale irrigation farming. We found few, small-scaled, and strongly degraded Populus euphratica forest stands which were highly impacted by fuel wood extraction and grazing and exhibited limited rejuvenation. Other vegetation subclasses, like Elaeagnus shrubland and Phragmites reeds, were more common. (b) The sampling region Zhanakorgan was dominated by large-scale irrigation farming. It comprised larger Populus pruinosa forest stands and few Elaeagnus shrublands. Although there were signs of partial rejuvenation in the forests, the impact of grazing and wood extraction was still high. (c) The sampling region Syr Darya–Turkestan Nature Park also comprised large-scale irrigation farming, but was less degraded. Additionally, it featured large Populus pruinosa forest stands close to the river course. Although the conservation status of the forests forbids wood extraction and grazing, the forest stands showed limited rejuvenation.
After deciding on the sampling sites, we derived validation points through visual interpretations of very high spatial resolution Digital Globe images via Google Earth (Table 2). WorldView imagery from the years 2015 to 2016 was used, which was provided by Maxar technologies. The initial validation point dataset consisted of 2916 samples belonging to the main classes of wetland and non-wetland. To attain more validation points for the smaller wetland subclasses of forest1 and forest2, we collected additional 492 validation points outside of the sampling regions.

2.3. Methods

2.3.1. Linear-Regression-Based Wetland Mask Creation

To create the wetland mask, we used linear regression analysis between a phenological reference profile [54] (NDVI_reference) and the NDVI time series vectors of each spatial cell (NDVI_observed) (Figure 5). NDVI_reference is the mean time series vector of ten healthy and natural Tugai forest stands at the Syr Darya–Turkestan Nature Park. NDVI_observed are the original time series vectors of each cell in the 250 m grid.
We chose Pearson’s correlation coefficient r [55] as the regression parameter for finding statistical relationships, where the value 1 indicates a strong positive linear correlation, 0 indicates no linear correlation, and −1 indicates a strong negative linear correlation. Because linear regression requires normally distributed input data, we conducted the Shapiro–Wilk test [56] to test the hypothesis of normal distribution within the data. The test is particularly suitable for data with a small sample size, as in our case with a maximum of 480 observations per time series vector. For all the individual time series vectors in the MODIS and L7 datasets, we observed a p value larger than 0.05, which means the hypothesis of normal distribution was not rejected. Next, the linear regression analysis was conducted for all 840,000 time series vectors from the MODIS and L7 datasets.
After the linear regression analysis, a threshold for Pearson’s r was chosen to create a wetland mask aiming to delineate most wetlands. We then evaluated the MODIS and L7 masks created in this step against the validation dataset. The wetland mask with the highest overall accuracy (OA) was used as a spatial boundary for further trend and breakpoint analysis (Figure 3).

2.3.2. Trend and Breakpoint Detection

To calculate long-term vegetation changes (i.e., trends) and abrupt changes (i.e., breakpoints) in the NDVI time series, it is necessary to decompose them into the trend, seasonal, and noise component using the R package bfast (version 1.6.1) [38]. From the trend component, we calculated the linear regression slope values to find positive and negative trends of the NDVI time series within the wetland mask. These long-term trends were interpreted as gains and losses of vegetation activity within the observed time frame of 16 years. If a regression slope was less than −0.0001, we defined it as a decreasing trend. In the opposite case, a regression slope larger than +0.0001 was considered a positive trend. Regression slope values between these values were interpreted as stable vegetation activity. From the trend component, we also calculated the breakpoints within the NDVI time series. They showed major years with abrupt changes in vegetation. For easier interpretability, we set the number of breakpoints to a maximum of two for the entire time series. We then visualized the breakpoints by their detection year.

3. Results

3.1. Wetland Masks and Accuracy Assessment

Figure 6 shows the MODIS and L7 maps derived from the linear regression analysis. The correlation coefficient r ranges between −0.57 and +0.91 in the MODIS product and between −0.68 and +0.86 in the L7 product. In both maps, high values (r > +0.6) were found close to the main course and tributaries of the Syr Darya. Middle values (r +0.5 to +0.6) were found at agricultural belts and the river delta. Low values (r < +0.5) were found away from the river course. The correlation coefficients among the validation samples from the wetland class (i.e., riparian forests, shrubs, and reeds) reached a mean value of +0.61 in the MODIS and +0.55 in the L7 products. The non-wetland classes (i.e., farmlands, drylands, and settlements) had mean values of +0.15 for MODIS and +0.23 for L7. Nevertheless, the subclass cropland1 (i.e., actively irrigated farmland) often reached r values larger than +0.5 in both products. Thus, achieving full separation between irrigated farmlands and natural wetlands using the chosen method was not possible. As a trade-off, we then chose the correlation coefficient value of r = +0.5 as the threshold for wetland masking to ensure that major parts of the natural wetlands were included and agricultural lands were partially excluded. Hence, for MODIS, we quantified 13,864 km2 of wetlands or 26.9% of the study area. For L7, we quantified 5216 km2 of wetlands or 11.0% of the study area.
The confusion matrices in the Appendix (see Table A2 and Table A3) show the classification accuracies of the MODIS and L7 wetland masks, considering the validation dataset. Although the L7 map achieved higher producer’s accuracies (PA) for the non-wetland classes (cropland1 (84.9%) and cropland2 (80.4%)), the MODIS classification performed better for the wetland classes (forest1 (75.1%) and forest2 (95.5%)). For the full validation dataset, the MODIS classification yielded an OA of 78.6%, whereas the L7 classification produced an OA of 63.0%. Because of the higher OAs and PAs for the wetland subclasses, the subsequent analysis was applied to the MODIS wetland mask.

3.2. Trend and Breakpoint Detection

Figure 7 and the detail maps (Figure A3) show the results of trend detection for the cells within the MODIS wetland mask. The regression slope values in the study area ranged between −0.00053 and +0.00042, with a mean value of −0.00000 for the whole wetland area. Larger patches of decreasing trends were observed in most wetlands close to the main course of the Syr Darya. Few spots with positive trends were found at the river delta and at a tributary north-east of Kyzylorda. Mixed values were found for the wetlands at the large cotton belt in the southern study area. We quantified an area measuring 4296 km2 with a decreasing trend, constituting 29.7% of the entire wetland area. Positive trends were quantified for an area of 1242 km2 or 8.5% of the wetlands. A total of 61.8% of the wetlands were classified as stable.
Figure 8 shows the year of the first and second breakpoints within the trend components. According to the BFAST algorithm, almost all the wetland cells underwent at least one breakpoint within the study time frame. Most of the breakpoints were observed for the years 2001, 2002, and 2009 (first breakpoint) and in the years 2013, 2014, and 2016 (second breakpoint). The detail maps (Figure A4) further reveal spatial relationships between the observed vegetation trends and breakpoints. The spots with positive trends at the river delta and at the tributary northern to Kyzylorda match with the breakpoint events around the year 2009. The zones with positive trends at the southern cotton belt are related to a breakpoint event in 2016. On the other hand, the areas with negative trend slopes are more closely associated with breakpoints in 2001 and 2002.

4. Discussion

4.1. Wetland Detection and Masking

Uncertainties relating to the large-scale monitoring of wetlands with remote sensing products in Central Asia are driven by large data gaps in the Landsat program [57] and the lack of ground truth information regarding the distribution of Tugai forests [58]. We completed a wetland mask for estimating the recent distribution of Tugai wetlands at the middle and lower reaches of the Syr Darya using MODIS and L7 NDVI time series and reference data from field observations.
Despite large differences in the observation numbers of the input time series, we were able to distinguish between wetlands and non-wetlands in the research area with an OA of 78.6%. MODIS data, with a much higher observation number, led to more precise classification results compared to L7. The main class (wetlands) was better separated using the MODIS product, with 81.5% PA, compared to the L7 product, with 41.8% PA. However, we found wetland classification uncertainties and misclassifications in both mapping products. The uncertainties are mainly explained by the phenological similarities between the irrigated croplands (e.g., cotton belts) and the Tugai wetlands (e.g., forests, reeds, and shrubs). Furthermore, misclassification in the MODIS product can be explained by mixed pixel effects due to the coarse spatial resolution of 250 m. For instance, we observed that cells in dry Tugai forest stands have a much weaker NDVI signals than stands closer to the river, which leads to lower PAs for the wetland subclasses. In contrast, the Tugai wetlands within the irrigated cotton belts are better distinguished using the L7 product due to its higher spatial precision.
Given the expanding archive of the Sentinel and Landsat missions and advancements in data fusion approaches, such as Landsat-Sentinel-2 or Landsat-MODIS [59,60], a higher quality of mapping products is expected, which could support local forest and conservation authorities in their monitoring duties. However, large-scale ground truth data is also needed to improve monitoring of the Tugai forests.

4.2. Trend and Breakpoint Detection

We found that the high temporal density of the MODIS time series data substantially benefitted trend and breakpoint detection in our study area. Many statistical tools and algorithms for vegetation disturbance modeling require temporally dense time series with preferably high spatial resolution [61]. For the wetlands, we found a decreasing trend in NDVI for up to one-third of the area, which implies ongoing degradation of the riparian zones at the Syr Darya. We found that flooding events have a positive influence on the Tugai wetlands. For example, in the winter of 2004, massive overflow occurred at the Syr Darya middle reaches [46]. This event led to a pronounced positive trend and breakpoint at that location over the following years. Positive vegetation trends were also detected in some areas of the Syr Darya river delta. This pattern may be the result of successful water discharges during the study period, which has been a common political goal following the drying up of the Aral Sea in the 1960s.
In many parts of Central Asia, water overuse occurs due to economic reasons [62]. We assume that ongoing water scarcity, intensive agricultural land use, and the extraction of fuel wood will remain the main drivers for degradation of the Tugai wetlands. Climate change will accelerate water scarcity for the riparian forests, especially during summer droughts. The temperature diagrams from the climate stations in Kyzylorda and Kazalinsk (Figure A1) at the Syr Darya provide clear evidence of rapid temperature increases over the past four decades [63,64]. By 2050, an increase in the region’s annual mean temperature by 2.0 to 3.6 °C is expected [65]. Although the annual amount of precipitation may not significantly change, a shift in the seasonal rainfall distribution is projected, with a significant decrease in precipitation in summer and autumn [66].

5. Conclusions

Due to their scattered distribution, the Tugai forests in Kazakhstan have remained understudied, particularly using remote sensing approaches. Hence, the past and recent health status of these ecosystems has remained largely unknown until now. Based on the example of the Syr Darya floodplain, we demonstrated a novel approach for precise and cost-efficient wetland mapping and monitoring in Central Asian drylands. Using MODIS NDVI time series data, we could detect trends and abrupt changes in vegetation vitality at the regional level. These metrics facilitate the monitoring of this highly threatened biodiversity hotspot.
Our work aimed to achieve two main objectives: (1) Comparing the suitability of L7 and MODIS data for mapping wetlands in the Central Asian drylands. Both satellite missions show advantages regarding their different temporal and spatial resolutions. For long-term vegetation monitoring in Central Asia, the MODIS data is more applicable due to its higher repetition rate. However, we found linear regression analysis, based on a phenological reference profile and the observed NDVI time series, to be a very useful tool for wetland detection with both data sources. (2) Capturing the degradation status of the remaining Tugai forests at the Syr Darya floodplain. Considering the wetland mask derived from MODIS data, we found a decreasing trend for 29.7% of the wetland area. The breakpoint detection revealed two main events which occurred in the years 2001/2002 and 2016. These events contributed to ongoing degradation of the remaining Tugai forests at the Syr Darya. The adverse health status of the riparian ecosystem is exacerbated by activities such as water extraction for irrigation farming, fuel wood extraction, overgrazing, and the lack of natural flooding. Without further protective measures, restoration of the ecosystem to stable conditions is becoming increasingly unattainable.

Author Contributions

Conceptualization, C.S. and B.K.; methodology, C.S. and B.K.; software, C.S.; validation, C.S.; formal analysis, C.S.; investigation, C.S. and B.K.; resources, B.K.; writing—original draft preparation, C.S.; writing—review and editing, C.S. and B.K.; visualization, C.S.; supervision, B.K.; project administration, B.K.; funding acquisition, B.K. All authors have read and agreed to the published version of the manuscript.

Funding

This article contributes to the ForeCeA project (Riparian forests in arid regions of Central Asia under pressure by use and limited water resources: definition of tipping points to ensure sustainable management). This research was funded by the German Ministry of Research and Education, grant number 01LC1707C. We acknowledge the support received from the Open Access Publication Fund of TU Berlin.

Data Availability Statement

The maps and validation data have been made available online at http://dx.doi.org/10.17632/sxyjnhvbm5.1 (accessed on 8 January 2021).

Acknowledgments

We thank Frank Thomas (University of Trier), Sabir Nurtazin (Al-Farabi Kazakh National University), and Margulan Iklasov (Al-Farabi Kazakh National University) for their contributions to organization and the joint field campaigns. Ruslan Salmurzauli (Al-Farabi Kazakh National University), Igor Klein (Deutsches Zentrum für Luft- und Raumfahrt e.V.), He Yin (Kent State University), Florian Betz (Katholische Universität Eichstätt-Ingolstadt), and Niels Thevs (World Agroforestry Centre Bishkek) kindly shared data with us. We also thank Stenka Vulova and Qiuandra Taylor for revising our English. The authors express their gratitude to the four anonymous reviewers for their constructive comments that enhanced the quality of the manuscript.

Conflicts of Interest

The authors declare no conflict of interest.

Appendix A

Table A1. List of regions and conservation sites of Tugai forest stands in Central Asia, defined by the distribution of the tree species Populus euphratica Oliv. and P. pruinosa Schrenk.
Table A1. List of regions and conservation sites of Tugai forest stands in Central Asia, defined by the distribution of the tree species Populus euphratica Oliv. and P. pruinosa Schrenk.
NameCountryRiver SystemSource(s)
Syr Darya–Turkestan State Regional Park (P. pruinosa)KZSyr Darya Middle Reaches[*]
Syr Darya near Zhanakorgan (P. pruinosa, P. euphratica)KZSyr Darya Middle Reaches[*]
Amudarya State Nature Reserve (P. pruinosa, P. euphratica)TMAmu Darya[12]
Lower Amu Darya State Biosphere Reserve (Badai-Tugai Nature Reserve) (P. euphratica, P. pruinosa)UZLower Amu Darya[5,67]
Northern Amu Darya Delta (P. euphratica, P. pruinosa)UZAmu Darya Delta[68]
Ili River Zheltorangy Genetic Reserve (P. pruinosa)KZIli River Delta[*], [3]
Ili River near Topar (P. euphratica, P. pruinosa)KZIli River Delta[*], [3]
Ili River near West Shonzy (P. pruinosa, P. euphratica)KZIli River Lower Reaches[*]
Ili River near South Koktal (P. pruinosa, P. euphratica)KZIli River Lower Reaches[*]
Tarim Huyanglin Nature Reserve (Tarim Euphrates Poplar National Nature Reserve) (P. euphratica)CNTarim River Middle and Lower Reaches[1,2,7,11,39,42,69,70]
Yingbazar (P. euphratica)CNTarim River Upper Reaches[10,71]
Argan (P. euphratica)CNTarim Lower Reaches[10,72]
Ebinur Lake (P. euphratica)CNBoro-Tala River[8,10]
Ekhiin-Gol Oasis (Great Gobi A) (P. euphratica)MNGobi Lakes Valley[73]
Ejin Diversifolius Poplar Forest National Nature Reserve (Heihe Oasis/East Junyan Lake/Ejina Oasis) (P. euphratica)CNHeihe River Basin/Ejina River[74,75,76]
Lower Chu River (Populus sp.)KZLower Chu Reaches[77]
Inner Mongolia, Gansu, Qinghai, Ningxia (Populus sp.)CN-[4]
* ForeCeA field campaigns in 2017 and 2018.
Forests 14 02305 g0a1
Figure A1. Yearly average temperature up to the year 2000 for two climate stations at the Syr Darya, left: Kyzylorda [63] and right: Kazalinsk [64].
Figure A1. Yearly average temperature up to the year 2000 for two climate stations at the Syr Darya, left: Kyzylorda [63] and right: Kazalinsk [64].
Forests 14 02305 g0a1
Forests 14 02305 g0a2
Figure A2. Pearson’s correlation coefficient r of the observed 8-day NDVI time series (2001–2016) and the simulated average NDVI time series of the Tugai wetlands at the Syr Darya delta (left), Kyzylorda (middle), and Shardara (right). Top row: MODIS. Bottom row: Landsat 7.
Figure A2. Pearson’s correlation coefficient r of the observed 8-day NDVI time series (2001–2016) and the simulated average NDVI time series of the Tugai wetlands at the Syr Darya delta (left), Kyzylorda (middle), and Shardara (right). Top row: MODIS. Bottom row: Landsat 7.
Forests 14 02305 g0a2
Forests 14 02305 g0a3
Figure A3. Linear regression slope values of the trend component of the 8-day NDVI time series (2001–2016) for the wetlands at the Syr Darya delta (left), Kyzylorda (middle), and Shardara (right).
Figure A3. Linear regression slope values of the trend component of the 8-day NDVI time series (2001–2016) for the wetlands at the Syr Darya delta (left), Kyzylorda (middle), and Shardara (right).
Forests 14 02305 g0a3
Forests 14 02305 g0a4
Figure A4. Year of breakpoints within the trend component of MODIS 8-day NDVI time series (2001–2016) at the Syr Darya delta (left), Kyzylorda (middle), and Shardara (right). Top row: first breakpoint. Bottom row: second breakpoint.
Figure A4. Year of breakpoints within the trend component of MODIS 8-day NDVI time series (2001–2016) at the Syr Darya delta (left), Kyzylorda (middle), and Shardara (right). Top row: first breakpoint. Bottom row: second breakpoint.
Forests 14 02305 g0a4
Table A2. Confusion matrix of the MODIS wetland classification (rows) and the validation dataset (columns). PA = Producer’s Accuracy for the subclasses. PA2 = Producer’s Accuracy for the main classes. OA = Overall Accuracy.
Table A2. Confusion matrix of the MODIS wetland classification (rows) and the validation dataset (columns). PA = Producer’s Accuracy for the subclasses. PA2 = Producer’s Accuracy for the main classes. OA = Overall Accuracy.
wetlandnon-wetland
forest1forest2forest1forest2forest1forest2forest1forest2
wetland58516837144023052687
non-wetland19481015223350439395
PA75.1%95.5%78.6%89.4%50.3%49.0%98.7%82.9%
PA281.5%74.9%
OA78.6%
Table A3. Confusion matrix of the Landsat 7 wetland classification (rows) and the validation dataset (columns). PA = Producer’s Accuracy for the subclasses. PA2 = Producer’s Accuracy for the main classes. OA = Overall Accuracy.
Table A3. Confusion matrix of the Landsat 7 wetland classification (rows) and the validation dataset (columns). PA = Producer’s Accuracy for the subclasses. PA2 = Producer’s Accuracy for the main classes. OA = Overall Accuracy.
wetlandnon-wetland
forest1forest2waterwetlandcropland1cropland2drylandsettlement
wetland429104112587020252
non-wetland3507246123439382443430
PA55.1%59.1%0.2%52.4%84.9%80.4%99.6%89.2%
PA241.8%90.3%
OA63.0%

References

  1. Ding, J.; Zhao, W. Comparing Chinese and International Studies of Riparian Forests: A Bibliometric Survey (1981–2014). Acta Ecol. Sin. 2016, 36, 377–385. [Google Scholar] [CrossRef]
  2. Schäfer, P.; Saleh, M.; Yu, R.; Zhang, X.; Thomas, F.M. Decrease in Growth Increment of Populus euphratica upon Defoliation by Lepidopteran Larvae in a Central-Asian Floodplain Forest. J. Arid. Environ. 2017, 146, 99–102. [Google Scholar] [CrossRef]
  3. Immentai, A.; Thevs, N.; Schmidt, S.; Nurtazin, S.; Salmurzauli, R. Vegetation, fauna, and biodiversity of the Ile Delta and southern Lake Balkhash—A review. J. Great Lakes Res. 2015, 41, 688–696. [Google Scholar] [CrossRef]
  4. Thevs, N. Tugay Vegetation in the Middle Reaches of the Tarim River—Vegetation Types and Their Ecology. Arch. Naturschutz Landschaftsforschung 2005, 44, 63–84. [Google Scholar]
  5. Tesch, N.; Thevs, N. Wetland Distribution Trends in Central Asia. Cent. Asian J. Water Res. 2020, 6, 39–65. [Google Scholar] [CrossRef]
  6. Millennium Ecosystem Assessment. Ecosystems and Human Well-Being: Synthesis; World Resources Institute: Washington, DC, USA, 2005. [Google Scholar]
  7. Olson, D.M.; Dinerstein, E.; Wikramanayake, E.D.; Burgess, N.D.; Powell, G.V.N.; Underwood, E.C.; D’amico, J.A.; Itoua, I.; Strand, H.E.; Morrison, J.C.; et al. Terrestrial Ecoregions of the World: A New Map of Life on Earth. BioScience 2001, 51, 933. [Google Scholar] [CrossRef]
  8. Säumel, I.; Ziche, D.; Yu, R.; Kowarik, I.; Overdieck, D. Grazing as a Driver for Populus euphratica Woodland Degradation in the Semi-Arid Aibi Hu Region, Northwestern China. J. Arid. Environ. 2011, 75, 265–269. [Google Scholar] [CrossRef]
  9. Wang, J.; Li, Z.; Guo, Q.; Ren, G.; Wu, Y. Genetic Variation within and between Populations of a Desert Poplar (Populus euphratica) Revealed by SSR Markers. Ann. For. Sci. 2011, 68, 1143. [Google Scholar] [CrossRef]
  10. Kramp, K.; Schmitt, T.; Lang, P.; Jeschke, M.; Schäfer, P.; Kulanek, D.; Zhang, X.; Yu, R.; Thomas, F.M. Clones or No Clones: Genetic Structure of Riparian Populus euphratica Forests in Central Asia. J. Arid. Land 2018, 10, 750–766. [Google Scholar] [CrossRef]
  11. Lang, P.; Ahlborn, J.; Schäfer, P.; Wommelsdorf, T.; Jeschke, M.; Zhang, X.; Thomas, F.M. Growth and Water Use of Populus euphratica Trees and Stands with Different Water Supply along the Tarim River, NW China. For. Ecol. Manag. 2016, 380, 139–148. [Google Scholar] [CrossRef]
  12. Thevs, N.; Buras, A.; Zerbe, S.; Kuhnel, E.; Abdusalih, N.; Ovezberdiyeva, A. Structure and Wood Biomass of Near-Natural Floodplain Forests along the Central Asian Rivers Tarim and Amu Darya. Forestry 2012, 85, 193–202. [Google Scholar] [CrossRef]
  13. Scheffer, M.; Bascompte, J.; Brock, W.A.; Brovkin, V.; Carpenter, S.R.; Dakos, V.; Held, H.; Van Nes, E.H.; Rietkerk, M.; Sugihara, G. Early-Warning Signals for Critical Transitions. Nature 2009, 461, 53–59. [Google Scholar] [CrossRef] [PubMed]
  14. Xu, C.; Van Nes, E.H.; Holmgren, M.; Kéfi, S.; Scheffer, M. Local Facilitation May Cause Tipping Points on a Landscape Level Preceded by Early-Warning Indicators. Am. Nat. 2015, 186, E81–E90. [Google Scholar] [CrossRef] [PubMed]
  15. Brook, B.W.; Ellis, E.C.; Perring, M.P.; Mackay, A.W.; Blomqvist, L. Does the Terrestrial Biosphere Have Planetary Tipping Points? Trends Ecol. Evol. 2013, 28, 396–401. [Google Scholar] [CrossRef] [PubMed]
  16. Ban, Y.; Yousif, O. Change Detection Techniques: A Review. In Multitemporal Remote Sensing; Ban, Y., Ed.; Springer International Publishing: Cham, Switzerland, 2016; Volume 20, pp. 19–43. ISBN 9783319470351. [Google Scholar]
  17. Zhu, Z. Change Detection Using Landsat Time Series: A Review of Frequencies, Preprocessing, Algorithms, and Applications. ISPRS J. Photogramm. Remote Sens. 2017, 130, 370–384. [Google Scholar] [CrossRef]
  18. Kuenzer, C.; Dech, S.; Wagner, W. Remote Sensing Time Series Revealing Land Surface Dynamics: Status Quo and the Pathway Ahead; Remote Sensing Time Series; Kuenzer, C., Dech, S., Wagner, W., Eds.; Springer International Publishing: Cham, Switzerland, 2015; Volume 22, pp. 1–24. ISBN 9783319159669. [Google Scholar]
  19. Hansen, M.C.; Potapov, P.V.; Moore, R.; Hancher, M.; Turubanova, S.A.; Tyukavina, A.; Thau, D.; Stehman, S.V.; Goetz, S.J.; Loveland, T.R.; et al. High-Resolution Global Maps of 21st-Century Forest Cover Change. Science 2013, 342, 850–853. [Google Scholar] [CrossRef] [PubMed]
  20. Friedl, M.A.; Sulla-Menashe, D.; Tan, B.; Schneider, A.; Ramankutty, N.; Sibley, A.; Huang, X. MODIS Collection 5 Global Land Cover: Algorithm Refinements and Characterization of New Datasets. Remote Sens. Environ. 2010, 114, 168–182. [Google Scholar] [CrossRef]
  21. Jun, C.; Ban, Y.; Li, S. Open Access to Earth Land-Cover Map. Nature 2014, 514, 434. [Google Scholar] [CrossRef]
  22. European Space Agency ESA CCI Global Land Cover Maps v2.0.7. Available online: https://www.esa-landcover-cci.org/ (accessed on 1 December 2020).
  23. Bastin, J.-F.; Berrahmouni, N.; Grainger, A.; Maniatis, D.; Mollicone, D.; Moore, R.; Patriarca, C.; Picard, N.; Sparrow, B.; Abraham, E.M.; et al. The Extent of Forest in Dryland Biomes. Science 2017, 356, 635–638. [Google Scholar] [CrossRef]
  24. Conrad, C.; Rudloff, M.; Abdullaev, I.; Thiel, M.; Löw, F.; Lamers, J.P.A. Measuring Rural Settlement Expansion in Uzbekistan Using Remote Sensing to Support Spatial Planning. Appl. Geogr. 2015, 62, 29–43. [Google Scholar] [CrossRef]
  25. Bohovic, R.; Dobrovolny, P.; Klein, D. The Spatial and Temporal Dynamics of Remotely-Sensed Vegetation Phenology in Central Asia in the 1982–2011 Period. Eur. J. Remote Sens. 2016, 49, 279–299. [Google Scholar] [CrossRef]
  26. Eisfelder, C.; Klein, I.; Niklaus, M.; Kuenzer, C. Net Primary Productivity in Kazakhstan, Its Spatio-Temporal Patterns and Relation to Meteorological Variables. J. Arid. Environ. 2014, 103, 17–30. [Google Scholar] [CrossRef]
  27. Eckert, S.; Hüsler, F.; Liniger, H.; Hodel, E. Trend Analysis of MODIS NDVI Time Series for Detecting Land Degradation and Regeneration in Mongolia. J. Arid. Environ. 2015, 113, 16–28. [Google Scholar] [CrossRef]
  28. Eisfelder, C.; Klein, I.; Bekkuliyeva, A.; Kuenzer, C.; Buchroithner, M.F.; Dech, S. Above-Ground Biomass Estimation Based on NPP Time-Series—A Novel Approach for Biomass Estimation in Semi-Arid Kazakhstan. Ecol. Ind. 2017, 72, 13–22. [Google Scholar] [CrossRef]
  29. Jiang, L.; Jiapaer, G.; Bao, A.; Guo, H.; Ndayisaba, F. Vegetation Dynamics and Responses to Climate Change and Human Activities in Central Asia. Sci. Total Environ. 2017, 599–600, 967–980. [Google Scholar] [CrossRef]
  30. Xu, H.; Wang, X.; Yang, T. Trend Shifts in Satellite-Derived Vegetation Growth in Central Eurasia, 1982–2013. Sci. Total Environ. 2017, 579, 1658–1674. [Google Scholar] [CrossRef] [PubMed]
  31. Yin, H.; Khamzina, A.; Pflugmacher, D.; Martius, C. Forest Cover Mapping in Post-Soviet Central Asia Using Multi-Resolution Remote Sensing Imagery. Sci. Rep. 2017, 7, 1375. [Google Scholar] [CrossRef]
  32. Klein, I.; Gessner, U.; Künzer, C. Generation of Up to Date Land Cover Maps for Central Asia. In Novel Measurement and Assessment Tools for Monitoring and Management of Land and Water Resources in Agricultural Landscapes of Central Asia; Mueller, L., Saparov, A., Lischeid, G., Eds.; Springer International Publishing: Cham, Switzerland, 2014; pp. 329–346. ISBN 9783319010168. [Google Scholar]
  33. Hermosilla, T.; Wulder, M.A.; White, J.C.; Coops, N.C.; Hobart, G.W.; Campbell, L.B. Mass Data Processing of Time Series Landsat Imagery: Pixels to Data Products for Forest Monitoring. Int. J. Digit. Earth 2016, 9, 1035–1054. [Google Scholar] [CrossRef]
  34. Cai, Z.; Jönsson, P.; Jin, H.; Eklundh, L. Performance of Smoothing Methods for Reconstructing NDVI Time-Series and Estimating Vegetation Phenology from MODIS Data. Remote Sens. 2017, 9, 1271. [Google Scholar] [CrossRef]
  35. Oeser, J.; Pflugmacher, D.; Senf, C.; Heurich, M.; Hostert, P. Using Intra-Annual Landsat Time Series for Attributing Forest Disturbance Agents in Central Europe. Forests 2017, 8, 251. [Google Scholar] [CrossRef]
  36. Jönsson, P.; Eklundh, L. TIMESAT—A Program for Analyzing Time-Series of Satellite Sensor Data. Comput. Geosci. 2004, 30, 833–845. [Google Scholar] [CrossRef]
  37. Čampulová, M.; Čampula, R.; Holešovský, J. An R package for identification of outliers in environmental time series data. Environ. Model. Softw. 2022, 155, 105435. [Google Scholar] [CrossRef]
  38. Verbesselt, J.; Hyndman, R.; Newnham, G.; Culvenor, D. Detecting Trend and Seasonal Changes in Satellite Image Time Series. Remote Sens. Environ. 2010, 114, 106–115. [Google Scholar] [CrossRef]
  39. Gärtner, P.; Förster, M.; Kurban, A.; Kleinschmit, B. Object Based Change Detection of Central Asian Tugai Vegetation with Very High Spatial Resolution Satellite Imagery. Int. J. Appl. Earth Obs. Geoinf. 2014, 31, 110–121. [Google Scholar] [CrossRef]
  40. Gärtner, P. Monitoring Dynamics of Semi-Arid Forests with Multi-Sensor Time Series. Ph.D. Thesis, Technische Universität Berlin, Berlin, Germany, 2017. [Google Scholar]
  41. Yu, Y.; Pi, Y.; Yu, X.; Ta, Z.; Sun, L.; Disse, M.; Zeng, F.; Li, Y.; Chen, X.; Yu, R. Climate Change, Water Resources and Sustainable Development in the Arid and Semi-Arid Lands of Central Asia in the Past 30 Years. J. Arid Land 2019, 11, 1–14. [Google Scholar] [CrossRef]
  42. Thevs, N.; Zerbe, S.; Schnittler, M.; Abdusalih, N.; Succow, M. Structure, Reproduction and Flood-Induced Dynamics of Riparian Tugai Forests at the Tarim River in Xinjiang, NW China. Forestry 2008, 81, 45–57. [Google Scholar] [CrossRef]
  43. Klein, I.; Gessner, U.; Kuenzer, C. Regional Land Cover Mapping and Change Detection in Central Asia Using MODIS Time-Series. Appl. Geogr. 2012, 35, 219–234. [Google Scholar] [CrossRef]
  44. Rubel, F.; Kottek, M. Observed and Projected Climate Shifts 1901-2100 Depicted by World Maps of the Köppen-Geiger Climate Classification. Meteorol. Z. 2010, 19, 135–141. [Google Scholar] [CrossRef]
  45. Wegerich, K.; Van Rooijen, D.; Soliev, I.; Mukhamedova, N. Water Security in the Syr Darya Basin. Water 2015, 7, 4657–4684. [Google Scholar] [CrossRef]
  46. NASA Syr Darya River Overflows. Available online: https://earthobservatory.nasa.gov/images/12739/syr-darya-river-overflows (accessed on 1 December 2020).
  47. Conrad, C.; Schönbrodt-Stitt, S.; Löw, F.; Sorokin, D.; Paeth, H. Cropping Intensity in the Aral Sea Basin and Its Dependency from the Runoff Formation 2000–2012. Remote Sens. 2016, 8, 630. [Google Scholar] [CrossRef]
  48. Xiong, X.; Barnes, W. An overview of MODIS radiometric calibration and characterization. Adv. Atmos. Sci. 2006, 23, 69–79. [Google Scholar] [CrossRef]
  49. Goward, S.N.; Masek, J.G.; Williams, D.L.; Irons, J.R.; Thompson, R.J. The Landsat 7 mission: Terrestrial research and applications for the 21st century. Remote Sens. Environ. 2001, 78, 3–12. [Google Scholar] [CrossRef]
  50. Rouse, J.W.; Haas, R.H.; Schell, J.A.; Deering, D.W. Monitoring vegetation systems in the great plains with ERTS. In Third Earth Resources Technology Satellite-1 Symposium: Section A-B. Technical Presentations, 1st ed.; Freden, S.C., Mercanti, E.P., Becker, M.A., Eds.; NASA: Washington, DC, USA, 1974; pp. 309–318. [Google Scholar]
  51. Mayr, S.; Kuenzer, C.; Gessner, U.; Klein, I.; Rutzinger, M. Validation of Earth Observation Time-Series: A Review for Large-Area and Temporally Dense Land Surface Products. Remote Sens. 2019, 11, 2616. [Google Scholar] [CrossRef]
  52. Gorelick, N.; Hancher, M.; Dixon, M.; Ilyushchenko, S.; Thau, D.; Moore, R. Google Earth Engine: Planetary-Scale Geospatial Analysis for Everyone. Remote Sens. Environ. 2017, 202, 18–27. [Google Scholar] [CrossRef]
  53. R Core Team. R: A Language and Environment for Statistical Computing, v 3.4.1; R Foundation for Statistical Computing: Vienna, Austria, 2021. [Google Scholar]
  54. Verger, A.; Baret, F.; Weiss, M.; Kandasamy, S.; Vermote, E. The CACAO Method for Smoothing, Gap Filling, and Characterizing Seasonal Anomalies in Satellite Time Series. IEEE Trans. Geosci. Remote Sens. 2013, 51, 1963–1972. [Google Scholar] [CrossRef]
  55. Pearson, K. Notes on Regression and Inheritance in the Case of Two Parents. Proc. R. Soc. Lond. 1895, 58, 240–242. [Google Scholar] [CrossRef]
  56. Shapiro, S.S.; Wilk, M.B. An Analysis of Variance Test for Normality (Complete Samples). Biometrika 1965, 52, 591. [Google Scholar] [CrossRef]
  57. Yu, Y.; Chen, X.; Disse, M.; Cyffka, B.; Lei, J.; Zhang, H.; Brieden, A.; Welp, M.; Abuduwaili, J.; Li, Y.; et al. Climate Change in Central Asia: Sino-German Cooperative Research Findings. Sci. Bull. 2020, 65, 689–692. [Google Scholar] [CrossRef]
  58. Liang, J.; Gamarra, J.G.P. The Importance of Sharing Global Forest Data in a World of Crises. Sci. Data 2020, 7, 424. [Google Scholar] [CrossRef]
  59. Frantz, D.; Stellmes, M.; Röder, A.; Udelhoven, T.; Mader, S.; Hill, J. Improving the Spatial Resolution of Land Surface Phenology by Fusing Medium- and Coarse-Resolution Inputs. IEEE Trans. Geosci. Remote Sens. 2016, 54, 4153–4164. [Google Scholar] [CrossRef]
  60. Claverie, M.; Ju, J.; Masek, J.G.; Dungan, J.L.; Vermote, E.F.; Roger, J.-C.; Skakun, S.V.; Justice, C. The Harmonized Landsat and Sentinel-2 Surface Reflectance Data Set. Remote Sens. Environ. 2018, 219, 145–161. [Google Scholar] [CrossRef]
  61. Wulder, M.A.; Loveland, T.R.; Roy, D.P.; Crawford, C.J.; Masek, J.G.; Woodcock, C.E.; Allen, R.G.; Anderson, M.C.; Belward, A.S.; Cohen, W.B.; et al. Current Status of Landsat Program, Science, and Applications. Remote Sens. Environ. 2019, 225, 127–147. [Google Scholar] [CrossRef]
  62. Weinthal, E. Beyond the State: Transnational Actors, NGOs, and Environmental Protection in Central Asia. In The Transformation of Central Asia: States and Societies from Soviet Rule to Independence; Luong, P., Ed.; Cornell University Press: Ithaca, NY, USA, 2004; pp. 246–270. [Google Scholar]
  63. NEESPI Meteorological Station Data of Kyzylorda from 1891 to 2006. Station Code: #38062. Available online: https://neespi.sr.unh.edu/cgi-bin/station/meteo_station.pl?id=38062 (accessed on 3 June 2019).
  64. NEESPI Meteorological Station Data of Kazalinsk from 1881 to 2006. Station Code: #34849. Available online: https://neespi.sr.unh.edu/cgi-bin/station/meteo_station.pl?id=34849 (accessed on 3 June 2019).
  65. Djanibekov, U.; Dzhakypbekova, K.; Chamberlain, J.; Weyerhaeuser, H.; Zomer, R.J.; Villamor, G.; Xu, J. Agroforestry for Landscape Restoration and Livelihood Development in Central Asia; ICRAF Working Paper; World Agroforestry Centre East and Central Asia: Kunming, China, 2015. [Google Scholar]
  66. Lioubimtseva, E.; Henebry, G.M. Climate and Environmental Change in Arid Central Asia: Impacts, Vulnerability, and Adaptations. J. Arid. Environ. 2009, 73, 963–977. [Google Scholar] [CrossRef]
  67. Kupke, L. Riparian Forest Structure and Regeneration in the Lower Amu Darya Biosphere Reserve (Uzbekistan) with Regard to Observed Degradation Factors. Master’s Thesis, Ernst Moritz Arndt University, Greifswald, Germany, 2017. [Google Scholar]
  68. Rüger, N.; Schlüter, M.; Matthies, M. A Fuzzy Habitat Suitability Index for Populus euphratica in the Northern Amudarya Delta (Uzbekistan). Ecol. Model. 2005, 184, 313–328. [Google Scholar] [CrossRef]
  69. Gärtner, P.; Förster, M.; Kleinschmit, B. The Benefit of Synthetically Generated RapidEye and Landsat 8 Data Fusion Time Series for Riparian Forest Disturbance Monitoring. Remote Sens. Environ. 2016, 177, 237–247. [Google Scholar] [CrossRef]
  70. Keyimu, M.; Halik, Ü.; Kurban, A. Estimation of Water Consumption of Riparian Forest in the Lower Reaches of Tarim River, Northwest China. Environ. Earth Sci. 2017, 76, 547. [Google Scholar] [CrossRef]
  71. Zeng, Y.; Zhao, C.; Kundzewicz, Z.W.; Lv, G. Distribution Pattern of Tugai Forests Species Diversity and Their Relationship to Environmental Factors in an Arid Area of China. PLoS ONE 2020, 15, e0232907. [Google Scholar] [CrossRef]
  72. Aishan, T.; Halik, Ü.; Betz, F.; Tiyip, T.; Ding, J.; Nuermaimaiti, Y. Stand Structure and Height-Diameter Relationship of a Degraded Populus euphratica Forest in the Lower Reaches of the Tarim River, Northwest China. J. Arid Land 2015, 7, 544–554. [Google Scholar] [CrossRef]
  73. Tugay. Available online: https://en.wikipedia.org/w/index.php?title=Tugay&oldid=1175204409 (accessed on 3 June 2019).
  74. Ott, K.; Kerschbaumer, L.; Köbbing, J.F.; Thevs, N. Bringing Sustainability Down to Earth: Heihe River as a Paradigm Case of Sustainable Water Allocation. J. Agric. Environ. Ethics 2016, 29, 835–856. [Google Scholar] [CrossRef]
  75. Ding, J.; Zhao, W.; Daryanto, S.; Wang, L.; Fan, H.; Feng, Q.; Wang, Y. The Spatial Distribution and Temporal Variation of Desert Riparian Forests and Their Influencing Factors in the Downstream Heihe River Basin, China. Hydrol. Earth Syst. Sci. 2017, 21, 2405–2419. [Google Scholar] [CrossRef]
  76. Guo, Y.; Li, X.; Zhao, Z.; Wei, H. Modeling the Distribution of Populus euphratica in the Heihe River Basin, an Inland River Basin in an Arid Region of China. Sci. China Earth Sci. 2018, 61, 1669–1684. [Google Scholar] [CrossRef]
  77. Egger, G.; Rood, S.B.; Becker, I.; Betz, F.; Chepinoga, V.; Deil, U.; Lashchinskiy, N.; Magnússon, B.; Roth, A.; Stewart, G.; et al. Riparian Vegetation of Gravel-bed Rivers—A Global Review. In Encyclopedia of Inland Waters, 2nd ed.; Elsevier: Amsterdam, The Netherlands, 2022; Volume 3, pp. 182–213. [Google Scholar] [CrossRef]
Forests 14 02305 g001
Figure 1. Study area and field sampling sites in the lower and middle reaches of the Syr Darya. Base map: land use and land cover from Klein et al. [43]. Overview map: main rivers of the Aral Sea Basin in Central Asia.
Figure 1. Study area and field sampling sites in the lower and middle reaches of the Syr Darya. Base map: land use and land cover from Klein et al. [43]. Overview map: main rivers of the Aral Sea Basin in Central Asia.
Forests 14 02305 g001
Forests 14 02305 g002
Figure 2. Vegetation types within Tugai wetlands at the Syr Darya: Elaeagnus shrubs (top left), Phragmites reeds mixed with Tamarix shrubs (top right), Populus euphratica reforestation site (bottom left), and Populus pruinosa forest at Syr Darya–Turkestan State Regional Park (bottom right). These photographs were taken during a field survey in September 2017.
Figure 2. Vegetation types within Tugai wetlands at the Syr Darya: Elaeagnus shrubs (top left), Phragmites reeds mixed with Tamarix shrubs (top right), Populus euphratica reforestation site (bottom left), and Populus pruinosa forest at Syr Darya–Turkestan State Regional Park (bottom right). These photographs were taken during a field survey in September 2017.
Forests 14 02305 g002
Forests 14 02305 g003
Figure 3. Methodological workflow diagram.
Figure 3. Methodological workflow diagram.
Forests 14 02305 g003
Forests 14 02305 g004
Figure 4. Number of observations per 250 m grid cell from 2001 to 2016 for MODIS (MCD43A4_NDVI) (top) and Landsat (LE7_L1T_8DAY_NDVI) time series (bottom) after vegetation season, filtering from March to October.
Figure 4. Number of observations per 250 m grid cell from 2001 to 2016 for MODIS (MCD43A4_NDVI) (top) and Landsat (LE7_L1T_8DAY_NDVI) time series (bottom) after vegetation season, filtering from March to October.
Forests 14 02305 g004
Forests 14 02305 g005
Figure 5. Example of the linear-regression-analysis-based wetland detection for a single NDVI time series vector (2001–2016) of a wetland cell. r is the Pearson’s correlation coefficient between the observed time series vector (blue) and the reference time series vector (green). The red line is the regression line of both vectors.
Figure 5. Example of the linear-regression-analysis-based wetland detection for a single NDVI time series vector (2001–2016) of a wetland cell. r is the Pearson’s correlation coefficient between the observed time series vector (blue) and the reference time series vector (green). The red line is the regression line of both vectors.
Forests 14 02305 g005
Forests 14 02305 g006
Figure 6. Pearson’s correlation coefficient r of the observed 8-day NDVI time series (2001–2016) and the simulated average NDVI time series of the Tugai wetlands at the Syr Darya. Top: MODIS (MCD43A4_NDVI). Bottom: Landsat 7 (LE7_L1T_8DAY_NDVI).
Figure 6. Pearson’s correlation coefficient r of the observed 8-day NDVI time series (2001–2016) and the simulated average NDVI time series of the Tugai wetlands at the Syr Darya. Top: MODIS (MCD43A4_NDVI). Bottom: Landsat 7 (LE7_L1T_8DAY_NDVI).
Forests 14 02305 g006
Forests 14 02305 g007
Figure 7. Linear regression slope values of the trend component of the 8-day MODIS NDVI time series (2001–2016).
Figure 7. Linear regression slope values of the trend component of the 8-day MODIS NDVI time series (2001–2016).
Forests 14 02305 g007
Forests 14 02305 g008
Figure 8. Year of breakpoints within the trend component of the 8-day MODIS NDVI time series (2001–2016). Top: first breakpoint. Bottom: second breakpoint.
Figure 8. Year of breakpoints within the trend component of the 8-day MODIS NDVI time series (2001–2016). Top: first breakpoint. Bottom: second breakpoint.
Forests 14 02305 g008
Table 1. Image collections from the Google Earth Engine used in this study. Original Source: United States Geological Survey.
Table 1. Image collections from the Google Earth Engine used in this study. Original Source: United States Geological Survey.
Image CollectionFull NameSpatial ExtentSpatial ResolutionTemporal ExtentTemporal Resolution
MCD43A4_NDVI MODIS Combined 16-Day NDVI global250 m2000–20238-day
LE7_L1T_8DAY_NDVILandsat 7 8-Day NDVI Compositeglobal30 m1999–20238-day
Table 2. Validation point dataset (n = number of samples per subclass).
Table 2. Validation point dataset (n = number of samples per subclass).
Main ClassSubclassDescriptionn
wetlandforest1Riparian forests with Populus euphratica or P. pruinosa779
wetlandforest2Riparian forests and shrubs with Elaeagnus sp.176
wetlandwetlandRiparian shrubs, reeds, and grassland492
wetlandwaterRivers, lakes, and ponds469
non-wetlandcropland1Actively irrigated farmland463
non-wetlandcropland2Abandoned irrigated farmland102
non-wetlanddrylandOpen soil, steppe, and dry shrubs445
non-wetlandsettlementCities, built-up482
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Schulz, C.; Kleinschmit, B. Monitoring the Condition of Wetlands in the Syr Darya Floodplain—How Healthy Are the Tugai Forests in Kazakhstan? Forests 2023, 14, 2305. https://doi.org/10.3390/f14122305

AMA Style

Schulz C, Kleinschmit B. Monitoring the Condition of Wetlands in the Syr Darya Floodplain—How Healthy Are the Tugai Forests in Kazakhstan? Forests. 2023; 14(12):2305. https://doi.org/10.3390/f14122305

Chicago/Turabian Style

Schulz, Christian, and Birgit Kleinschmit. 2023. "Monitoring the Condition of Wetlands in the Syr Darya Floodplain—How Healthy Are the Tugai Forests in Kazakhstan?" Forests 14, no. 12: 2305. https://doi.org/10.3390/f14122305

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop