Monitoring the dynamics of global surface water 1999-2017

Land cover and land use (LCLU) is changing in our world faster than ever, yet few types of land cover can change or fluctuate as quickly as surface water, and few things are as critical to life while also having so much power for devastation. Water affects and is affected by all forms of LCLU. Forests can create rain systems and filter water before it reaches the stream network. Agriculture is dependent on water and is responsible for 70% of surface water use. Irrigation has drained lakes such as the Aral Sea and Lake Urmia, and reservoirs are built to supply it as well as urban areas. Cropland also is a source of pollutants and excess sediment in the waterways. In order to understand all the relationships of LCLU change, we must understand the changing surface water. The authors have created water dynamics layers highlighting changes through an automated process mining the entire 1999-2017 Landsat 5, 7, and 8 archive. A classification tree model was developed to identify land, water, cloud, shadow, haze, and snow in each scene. This was applied to all 3 million scenes, totaling 1.98 petabytes. The land and water observations were used to create monthly percent water layers, which are defined as the number of water observations divided by the number of clear observations, and seasonally-weighted annual water percent layers. An inter-annual dynamics model was developed to visualize the changes happening through the years and accurately characterize them into eight classes. A probability-based sample assessment of the time-series was conducted to validate the static and change classes 1999-2015 and to estimate the areas of each. Clear observations from 13,000 scenes were used for the assessment. There are 2.85 (± 0.08) million km² of permanent water from 1999-2015. 1.97 million km² changed between water and land at least once during the period. Maps of mean water percent for each month are combined to show the seasonality of surface water. The maps of each month and of the seasonal and inter-annual dynamics can be used within a comprehensive dynamic LCLU monitoring system to assess the impacts of LCLU change, habitat suitability, natural resource levels, and ecosystem services. Figure: Percent forest cover from Hansen et al. 2013, Global Man-made Impervious Surface from Brown de Colstoun et al. 2017.