The Boundaries and Remote Sensing Classification Datasets on Large Wetlands of International
Importance in 2001 and 2013

Zhang, H. Y.¹  Niu, Z. G.^1*  Xu, P. P.¹  Chen, Y. F.¹  Hu, S. J.¹  Gong, N.²

1. State Key Laboratory of Remote Sensing Science, Institute of Remote Sensing and Digital Earth,
Chinese Academy of Sciences, Beijing 100101, China??

2. Department of Resources and Environment, Shandong Agricultural University, Tai’an 271018, China

Abstract: Wetland is one of the ecosystems with the highest biodiversity and productivity in the world. In this paper, we developed a dataset of 100 large Wetlands of International Importance (the “Ramsar Sites”), which was developed by the data integration and mining from the World Database on Protected Areas (WDPA) and remote sensing images, expecially the high resolution imagery provided on the Google Earth and the Tianditu platforms. Each site covers at least an area of 200,000 hm². Based on the Moderate Resolution Imaging Spectro radiometer (MODIS) 16-day composite products MOD13Q1 of these sites in 2001 and 2013, we applied the Savitzky-Golay filtering and time series reconstruction algorithms to Normalized Difference Vegetation Index (NDVI) data. The method of the Support Vector Machine (SVM) classifier was used to classify and map wetlands of these Ramsar sites. The time series data based wetland classification results were compared with high resolution remote sensing imagery and the relevant literatures. This method is more advantageous than single temporal imagery based wetland monitoring technique. The dataset consists of three components which are saved as .shp and .kmz file formats. The total data size is 102 MB in compressed file format.

Keywords: wetlands; boundary; remote sensing classification; wetland mapping; Ramsar sites

1 Introduction

Wetland is one of the ecosystems with the highest biodiversity and productivity in the world^[1]. In recent years, large scale wetland monitoring and mapping emerged as a hot research topic and has been attracting great attention of many scholars because of its ecological significance. To understand the spatio-temporal change characteristics of wetland ecosystem is the key component of global change research^[2]. The fast development of remote sensing technique provides an objective means to global wetlands monitoring research. The remotely sensed time series data play an irreplaceable role in the research of wetland changes in the growth cycle or in a calendar year of wetland vegetation^[3]. It is the main data source of the remotely sensed monitoring on large wetlands of international importance at the global scale of this paper. Along with our related scientific reports published, we make this dataset available for public use.

2 Metadata of Dataset

The metadata of the boundaries and remote sensing classification datasets on large wetlands of international importance in 2001 and 2013 is summarized in Table 1^[4]. It includes the dataset full name, short name, authors, year of the dataset, temporal resolution, spatial resolution, data format, data size, data files, data publisher, and data sharing policy, etc.

Table 1  Summary of the RamsarSites_Top100_WetlandCover_2001/2013 metadata.

3 Methods

3.1 The Objective and Indicators of Wetland Monitoring by Remote Sensing

Based on some principles such as wetlands with an area larger than 200,000 hm², with established geographic extent and map information, 100 wetlands of international importance all over the world distributed at different ecological-climatic zones and continents were selected from the database of Ramsar sites (https://rsis.ramsar.org/) (Figure 1).

Indices for monitoring large wetlands of international importance mainly include the types and the area of land cover (Table 2). The conversion between wetlands and non-wetlands, the conversion in different wetland types and the changes on the area of different types, were used as important indicators to eva-

luate the ecological and environmental conditions of the wetlands. In general, if wetlands are converted into non-wetlands, or natural wetlands are converted into artificial wetlands, they usually lead to a deteriorated environmental condition and pose some threats to wetland biodiversity conservation. On the contrary, it often causes healthy development on the wetland environments.

3.2 Methodology of Wetland Boundaries and Classification Dataset Development

Most of the wetland boundaries were delineated by reference to the scanned paper maps published on the Ramsar website (https://rsis.ramsar.org/. Retrieved on July 23, 2014)^[6]. Another small part of them were derived from World Database on Protected Areas (https://www.iucn.org/; http://www.protectedplanet.net/. Retrieved on July 23, 2014) and the other products of literature and document. The maps were georeferenced in ArcGIS. Because there are many differences of wetland management and monitoring techniques in each country, some inconsistencies of the boundaries often occur in such occasion. Therefore we

made modification and verification by using high resolution remote sensing images on Google Earth platform, based on the physiographic forms of wetlands such as walking along rivers, roads, and the ridge lines. The MODIS product MOD13Q1 (16-day composite, 250 m)

Table 2  Land cover classification system adopted in remote sensing based mapping

on wetlands of international importance

in 2001 and 2013 downloaded from NASA (https://ladsweb.nascom.nasa.gov/data/) was used as the main data source. There are 23 periods of the images in each year. In this study, the images on the public service platform Tianditu (http://www.tianditu.com/map/index.html) powered by the National Geomatics Center of China (NGCC) and other geographic data were also used as auxiliary data. The Normalized Difference Vegetation Index (NDVI) can reflect the growth status of different vegetation types. By comparing the NDVI differences including the maximum, the minimum, and the range of annual NDVI, the emergence date of the positive NDVI value, and the growing season length, etc. between different vegetation types in the growth cycle or within a calendar year, and analyzing plants growth pattern^[7], the accuracy of interpretation on different types of land cover types could be effectively improved. We applied Savitzky-Golay filtering to reconstruct NDVI time series dataset^[8]. Based on the clustering results of time series data and the analysis on the features of different land cover types in high resolution remote sensing images, we collected 10-20 training samples for each type and used SVM classifier to delineate and map different types of wetlands. The minimum of the map unit is about 56.25 hm² for 3×3 pixels. The main methods and the technical flow are described in Figure 2.

4 Data Products

The dataset consists of three components, namely Top100_boundary kmz (1,470.14 KB), Ramsar2001.rar (52,245.46 KB) and Ramsar2013.rar (50,858.42 KB). They represent the boundaries of 100 large wetlands of international importance in the world, and the wetland classif­ication results in 2001 and 2013, respectively (http://www.geodoi.ac.cn/WebEn/doi.aspx?

Id=243).

4.1 Statistics on Large Wetlands of International Importance

The total area of 100 large wetlands of international importance is 138,392,121.05 hm². Okavango Delta System (Ramsar Site No.879), located in Botswana, ranks first. The statistical results are listed in Table 3.

Table 3  Statistics on large wetlands of international importance

4.2 Remote Sensing Classification Results on Large Wetland of International Importance in 2001 and 2013

Based on the remote sensing classification results on 100 large wetlands of international importance in 2001 and 2013, we calculate the subtotal area of different land cover types. The statistical results are shown in Table 4.

In 2001, the largest three wetland classes are forested/shrub-dominated wetlands (deciduous), seasonal marshes, and permanent marshes, respectively. Between 2001 and 2013, the total area of wetland reduced from 58,134,960 hm² to 57,734,032 hm², the big three losers were water, forested/shrub-dominated wetlands (deciduous), and tundra wetlands, respectively.

Table 4  Area of different land cover types in large wetlands of international importance

5 Data Validation

Validation on the world’s large wetlands classification results accuracy was applied by

using high resolution remotely sensed images. We randomly selected 10 wetlands of international importance in different continents as the testing regions except for the Antarctic. About 2,386 validation samples randomly collected were interpreted based on high resolution images of China’s satellites GF-1, ZY-3. Accuracy of all land cover types was assessed. In general, the overall accuracy for all types is 88% and 89% in 2001 and 2013 respectively. And the Kappa coefficients are 0.86 and 0.87 respectively.

The dataset contains the boundaries of the world’s 100 large wetlands of international importance and wetland classification results on them in 2001 and 2013 based on 250 m spatial resolution, 16-day composite product MOD13Q1. It provides important reference data for research on global wetlands and climate change.

Author Contributions

Six coauthors were engaged in the design and development of the boundaries and remote sensing classification datasets on large wetlands of international importance in 2001 and 2013. Zhang, H. Y. and Niu, Z. G. participated in all the technological processes of the design and methodological development of boundaries extraction, data collection and processing, data analysis and validation, etc. Zhang, H. Y., Xu, P. P., Chen, Y. F., Hu, S. J. and Gong, N. took part in the boundaries extraction and wetlands classification. And the author list was arranged in their workload order. Zhang, H. Y. wrote this paper, and Niu, Z. G. was in charge of the paper proofreading, editing and revision.

Acknowledgements

We appreciate the National Geomatics Center of China, Google Earth, and the National Aeronautics and Space Administration (NASA), US for providing online public services of Tianditu, Google Earth and MODIS data.

References

[1]       Costanza, R., d’Arge, R., de Groot, R., et al. The value of the world’s ecosystem services and natural capital [J]. Nature, 1997, 387: 253-260.

[2]       Kloiber, S. M., Macleod, R. D., Smith, A. J., et al. A semi-automated, multi-source data fusion update of a wetland inventory for East-Central Minnesota, USA [J]. Wetlands, 2015, 35: 335-348.

[3]       Dong, J., Xiao, X., Kou, W., et al. Tracking the dynamics of paddy rice planting area in 1986-2010 through time series Landsat images and phenology-based algorithms [J]. Remote Sensing of Environment, 2015, 160: 99-113.

[4]       Zhang, H. Y., Niu, Z. G., Gong, N., et al. The wetland cover datasets on the large wetlands of international importance in 2001 and 2013 by remote sensing data integration [DB/OL]. Global Change Research Data Publishing & Repository, 2015. DOI: 10.3974/geodb.2015.02.09.V1.

[5]       GCdataPR Editorial Office. GCdataPR Data Sharing Policy [OL]. DOI: 10.3974/dp.policy.2014.05 (Updated 2017).

[6]       Ramsar Convention Secretariat. The Ramsar convention manual: A guide to the convention on wetlands (Ramsar, Iran, 1971), 4th edition [Z]. 2006.

[7]       Jin, Z., Xu, B. A novel compound smoother?DRMMEH to reconstruct MODIS NDVI time series [J]. IEEE Geoscience and Remote Sensing Letters, 2013, 10: 942-946.

[8]       Chen, J., Jönsson, P., Tamura,M., et al. A simple method for reconstructing a high-quality NDVI time-series data set based on the Savitzky-Golay filter [J]. Remote Sensing of Environment, 2004, 91: 332-344.