Trend analysis of the decadal variations of water bodies and land use/land cover through MODIS imagery: an in-depth study from Gilgit-Baltistan, Pakistan

Water is a primary element for human life on Earth. Fresh water, which includes rivers, lakes, streams, and ponds, contributes less than one thousandth of a percent of the total water on Earth, but it is critical for the environment and human life. Change in land use and land cover (LULC) is a foremost concern in global environment change. Rapid changes in LULC lead to the degradation of ecosystems and have adverse effects on the environment. There is an urgent need to monitor changes in LULC and evaluate the effects of these changes in order to inform decision makers on how to support sustainable development. This study used Moderate Resolution Imaging Spectroradiometry images to detect and investigate changes in LULC patterns in Gilgit-Baltistan, Pakistan, between 2008 and 2017. Six types of LULC were used to explain the major changes of LULC in the study area. The results showed that there was a reduction of barren lands and an increase of urban areas. It also showed an inconsistent behavior of water bodies during the study. Snow area, which also increased, needs further investigation.


GRAPHICAL ABSTRACT INTRODUCTION
Freshwater resources are a primary necessity for human life on Earth; however, comprehensive evidence about variations in freshwater bodies and the capacity to store the water from these water bodies is surprisingly limited (Khandelwal et al. ). Despite the significance of research into long-term variations and trends of freshwater sources, the current global awareness is surprisingly inadequate (Döll et al. ).
Water flows on the Earth's surface, water storage on the land surface, and water storage under the Earth's surface have key roles in sustaining human life and maintaining the ecological system. They have been gradually transformed by human activities including urbanization, deforestation, greenhouse gase production changes in land use, and the construction of dams and barrages, many of which cause severe and rapid flooding and can affect atmospheric processes, sea level increment, and global biogeochemical cycles.
These transformations affect whole ecosystems across most regions of the world, and rapid development by humans is leading to water scarcity. Surface water observation is a functional necessity to study the processes of ecology and hydrology (Huang et al. ). It is important to characterize the natural and human component of freshwater systems for environmentally friendly land and water management systems and for a better ecosystem. In recent years, the continuous monitoring of water resources with respect to natural and man-made changes has been a concern all over the world. Study of spatiotemporal dynamic of surface water, configurational and compositional features of water areas and their link with different LULC types are vital for managing and protecting ecosystems of specific types because these type a are often highly vulnerable to human activities like tourism and urbanization. Zhichao Li and Yujie Feng recently presented a study about the monitoring of water body patterns in a Mediterranean lagoon using remote sensing techniques.
The spatio-temporal dynamics of water surface by water frequency index was studied on a yearly basis (Li et al. ).
Remotely sensed data using satellites has been widely adopted to determine the amount of LULC change across the world (Ehlers et  However, the spatial resolution of MODIS is not great for detecting small patches of water. Satellite images with high spatial resolution have been useful in such cases to obtain detailed geographical information about LULC. However, their resolution does not have great temporal resolution and they cannot perform a continuous observation at a comparatively large spatial scale. Fast and simple extraction of water bodies from remote sensing images are often done by multi-spectral thresholding methods. It is especially useful for large-scale and long time series studies. In general, to extract the water body by MODIS imagery, single band thresholds or index methods cannot efficiently handle the effect of cloud shadows and mountains, and the spectral relationship method will cause the reduction of spatial resolution of data (Huang et al. ).
Water patches are important for wildlife because they offer a food source, breeding areas, a safe refuge and they harbor many animal and plant species that would not survive in the adjacent landscape. Obviously, monitoring the surface water dynamics for water management is very important. However, the monitoring of water bodies dynamics is also important for ecosystem valuation (Ahamad et al. ) and biodiversity preservation over a long period of time (Li et al. ).
However, the spatial and temporal dynamics of surface water over a long period in this area of research has not given much attention to the researchers. In addition, studies have not yet been conducted on the linkage of surface hydrodynamics with the types of LULC. In this regard, this study focuses on a comprehensive overview of surface water dynamics in Gilgit-Baltistan, Pakistan, as well as the diagnosis of the effects of natural and human activities on frequency of water changes during the period 2008-2017, including the following: • Monitoring all frequency maps on a yearly basis using all MODIS (MCD12Q1) images available from 2008 to 2017.
• Studying the variation in spatial, temporal, compositional, and configurational patterns of water areas using MODIS data.
• Analyzing the connection between different LULC types and yearly surface water dynamics based on maps created by MCD12Q1.

METHODS AND DATA
The MODIS Land Cover Type MCD12Q1 uses six land cover legends to deliver an annual collection of science

Data-preprocessing
Covers a large area of ground, from which a subset of Gilgit-Baltistan images was extracted. MODIS land products are provided by the U.S. Geological Survey in hierarchical data format, and the projection of this data is a sinusoidal (SIN) projection. Conventional data-processing software cannot handle storing this format and projection, and therefore, for further use, each scene is re-projected as a more commonly used projection known as Universal Transverse Mercator (UTM, WGS84) (Huang et al. ). The study area consisted of two tiles from the MCD12Q1 images and so preprocessing of the data involved the seamless combination of the mosaic of tiles and the preparation of the subsets from the full scenes.

Study area
This paper covers the high elevation region of Gilgit-Baltistan in northeast Pakistan and includes the mountain ranges of Karakorum, Hindukush, and Himalayas (Bilal Elevation range of Gilgit-Baltistan is 950-8,538 m above sea level (ASL)-approximately 90% is situated above 3,000 m above ASL and 12% is above 5,500 m ASL (Bin & Yinsheng ). The Indus River Basin water catchment relies on Gilgit-Baltistan, and the, majority of Pakistan is dependent on the area for hydroelectrically and for irrigation. The high mountain region is dominated by winter rain and is the backbone for water in Pakistan (Raza et al. ). This region hosts the world's three longest glaciers outside the polar regions: the Biafo Glacier, the Bartoro Glacier, and the Batura Glacier. It also has the world's second highest mountain, K-2, and several high-altitude lakes ( Figure 1).

METHODOLOGY
Several change detection techniques have been used by researchers to identify differences in phenomenon or in Another advancement was the gap-fill technique to enhance the quality of data. The IGBP full name for the SDS is Land Cover Type 1, and its short name is LC_Type1. It is 8-bit unsigned data with valid range 1-17 and fill value is 255 (Table 1).

Water bodies and LULC detection
This annual composite data shows effectively and accurately the spatial and annual variations of surface water and LULC change during 2008-2017. The mosaicing and reprojection The canopy is greater than 2 m of evergreen conifer trees, the tree cover is greater than 60% area.
Forest with broadleaf trees 2 The canopy is greater than 2 m of evergreen with broad leaves and palmate trees, the tree cover greater than 60% area.
Forest with deciduous needle leaf trees 3 The canopy is over 2 m and the trees cover more than 60% of the deciduous needle leaf (larch) trees.

Deciduous broadleaf forests 4
The canopy is over 2 m and the trees cover more than 60% of the deciduous broad leaf trees.
Forest with mixed trees 5 Neither deciduous nor evergreen (40-60%) tree type, canopy more than 2 m and tree coverage of more than 60%.
Wetlands with permanent water 11 Always swamped lands with 30-60% water cover and vegetal cover greater than 10%. In MS Excel, the following formula was used to calculate the percentage of different LULC types: (3)  (6) and (7) were used to calculate slope estimate P t :

). Equations
where Y u represents the data values at time U and Y v represents data values at time V. If Y u holds total number of n values in the time series, we have as many as N ¼ n(n À 1)/2 slope estimates P t . Median of these N values of P t is Sen's estimator of slope, where N values of P t were ranked in increasing order, along with the Sen's estimator.
An upward trend can be detected by the positive value of P, while a negative value of P represents a downward trend in time series.

Reference data and validation
The different LULC can be visually distinguished using highresolution imagery provided by Google Earth. Google Earth imagery was therefore used to verify the sample points extracted from Landsat7 imagery with a 30-m resolution.
Google Earth imagery with high resolution is the optimum platform for verifying sample points (El-zeiny & Hala ) for any LULC extracted from Landsat7. MODIS Tera MOD09A1 v. 6, which provided surface reflectance data set with bands 1-7 with a temporal resolution of 8 days, was used in this study to verify surface water bodies.
MODIS Terra Snow Cover 8-Day with spatial resolution of 500 m (MOD10A2) was used to verify the snow cover area as a reference. Water bodies extraction was done using  ), and this should be taken into account in further studies.
Moderate spatial resolution data (MCD12Q1) was used to analyze different types of LULC in Gilgit Baltistan.
MODIS data is USED extensively to study LULC and related environmental impacts OVER a large area BECAUSE rapid and convenient processing is possible (Liu et al. b).
Using high-resolution imagery like Landsat is very time consuming for analyzing the impact of the LULC changes on surface water patches for Gilgit Baltistan. Higher spatial resolution satellite data also usually has lower temporal resolution. However, using moderate resolution data may compromise the accuracy compared to using higher resolution data like Landsat. Future studies should use a fusion of multi-sensor data (e.g. Landsat series and MODIS) to investigate the LULC changes in a specific city or a region