『Abstract
　As a limiting nutrient in aquatic systems, phosphorus (P) plays an important role in controlling freshwater and coastal primary productivity and ecosystem dynamics, increasing frequency and severity of harmful and nuisance algae blooms and hypoxia, as well as contributing to loss of biodiversity. Although dissolved inorganic P (DIP) often constitutes a relatively small fraction of the total P pool in aquatic systems, its
bioavailability makes it an important determinant of ecosystem function. Here we describe, apply, evaluate, and interpret an enhanced version of the Global NEWS-DIP model: NEWS-DIP-Half Degree (NEWS-DIP-HD). Improvements to NEWS-DIP-HD over the original NEWS DIP model include: 1) the preservation of spatial resolution of input datasets at the 0.5 degree level, and 2) explicit downstream routing of water and DIP from half degree cell to half degree cell using a global flow-direction representation. NEWS-DIP explains 78% and 62% of the variability in per-basin DIP export (DIP load) for USGS and global stations, respectively, similar to the original NEWS-DIP model and somewhat more than other global models of DIP loading and export. NEWS-DIP-HD output suggests that hot spots for DIP loading tend to occur in urban centers, with the highest per-area rate of DIP loading predicted for the half-degree grid-cell containing Tokyo (6,366 kg P km^-2 y^-1). Furthermore, cities with populations >100,000 accounted for 35% of global surface water DIP loading while covering less than 2% of global land surface area. NEWS-DIP-HD also indicates that humans supply more DIP to surface waters than natural weathering over the majority (53%) of the Earth’s land surface, with a
much larger area dominated by DIP point sources than non-point sources (52% versus 1% of the global land surface, respectively). NEWS-DIP-HD also suggests that while humans had increased DIP input to surface waters more than 4-fold globally by the year 2000, human activities such as dam construction and consumptive water use have somewhat moderated the effect of humans on P transport by preventing (conservatively)
0.35 Tg P y^-1 (〜20% of P inputs to surface waters) from reaching coastal zones globally.』

1. Introduction
2. methods
　2.1. NEWS-DIP-HD description
　2.2. Model validation data
　2.3. Model input data
　2.4. Post-processing of model output
　2.5. Model evaluation (Model uncertainty, sensitivity, and efficiency)
3. Results and discussion
　3.1. Model performance
　3.2. Model output
　　3.2.1. Spatial distribution of DIP export and sources
　　　Export
　　　Sources
　　3.2.2. Global and regional analyses
　3.3. Sources of uncertainty and future directions
　　3.3.1. Model efficiency and sensitivity
　　3.3.2. Future directions
4. Acknowledgments
5. References cited
Table legends
Figure legends

Table Legends

Table 1. Input datasets for NEWS-DIP-HD
Table 2. Metrics of model performance for NEWS-DIP-HD and other models, validated with dataset described in section 2.2 unless otherwise noted.. R² is model efficiency as defined in section 3.1, and r² is the coefficient of determination. Errors are computed as the difference between the predicted and measured values of stream phosphorus yield (kg km^-2 yr^-1) expressed as a percentage of the measured export (section 3.1).
Table 3. Model efficiencies for comparison of log-transformed measured and model predicted DIP yield, using NEWS-DIP-HD with various components removed. All available measurement data were used for this analysis.
Table 4: Results of a sensitivity analysis indicating mean change in predicted DIP yield as a function of increasing input datasets and model parameters by +10%.

Figure Legends

Figure 1. Spatial distribution of basins used to validate the NEWS-DIP-0.5 model (n=206 overall). Xs represent sampling stations within the Amazon River Basin (Devol et al., 1995). Plus signs represent sampling stations monitored by the Global Environment Monitoring System (GEMS) Water program not used in Harrison et al., 2005 (n= 33; http://www.gemswater.org/index.html). Hollow circles represent sampling stations within the Mississippi, Sacramento, and San Joaquin River Basins (n=54; Data source: Alexander et al. (1996)). Black diamonds represent stations used in calibration and validation of the original NEWS-DIP model (n = 118; Harrison et al., 2005). See Appendix B for data, model output, and station names.
Figure 2. Measured versus modeled DIP load (Top; kg P basin^-1 yr^-1) and yield (Bottom; kg P km^-2 y^-1) for Mississippi River Basin stations (hollow circles), global coastal stations included in Harrison et al. (2005; black diamonds), global stations from the United Nations Global Environmental Monitoring System (GEMS; plus signs), and data from the Amazon Basin ([Devol et al., 1995]; exes). See Appendix B for data, model output, and basin names. The 1:1 line is also shown. Symbols are the same as in Figure 1.
Figure 3. A) NEWS-DIP-predicted DIP yield by half-degree grid cell (kg P km^-2 y^-1) and B) NEWS-DIP-HD-predicted DIP yield by half-degree grid cell (kg P km^-2 y^-1). White areas are either endoreic (A) or have a predicted DIP loading to surface waters equal to zero (B).
Figure 4. NEWS-DIP-HD-estimated DIP retention (kg P km^-2 y^-1) globally by half degree. Estimates without information regarding reservoir locations and consumptive water use were assumed to retain no DIP, making this quite a conservative estimate of DIP retention within watersheds globally.
Figure 5. Dominant source of DIP by half-degree grid cell. “Dominant source” is defined as the modeled source that NEWS-DIP-HD predicts contributes the largest single fraction of DIP to the coast.
Figure 6. River export of DIP (Tg P ^-1) from continents and to ocean basins. Relative influence of various P sources calculated according to NEWS-DIP-HD.