Quantifying Spatio-temporal Variations Of Nonpoint Source Phosphorus Export Coefficients For Different Land-use Types At The Watershed Scale

Water eutrophication due to excessive nutrient (e.g., phosphorus) input has been one of major limitation for water resource sustainable utilization in many regions across the world. Compared to point source pollution, nonpoint source (NPS) pollution progressively becomes the dominative cause of excessive phosphorus (P) in many watersheds. Quantitative knowledge of P loss density for various land-use types within a watershed is critical for developing efficient NPS P pollution control strategies. This thesis aimed at quantitative identification of spatio-temporal variations of total P (TP) export coefficient (EC) of each of four land-use types (i.e., the farmland, the residential land, the forest and the barren land). Fist, a global meta-analysis was conducted to evaluate the variability of TP ECs of various land-use types (total of 791 observations with 287 for the farmland, 180 for the residential land,183 for the forest and 141 the barren land) over time (1965-2010) and space as well as to determine their statistical distribution characteristics. Second, adopting LOADST model and Bayesian method, individual TP ECs of four land-use types were estimated as for the Yongan watershed during the period 1980-2010. Finally, we were able to introduce the ratio of soil topography index and distance away from the river (TI/D)as a new parameter for exploring potentially spatial distributions of estimated TP ECs for each land-use type across the watershed with help of ArcGIS. The main studied results acquired are as follows:(1) The global meta-analysis showed that the TP ECs for four land-use types synchronously conformed to the lognormal distributions. Among the four land-use types, the order of TP ECs followed significantly: the residential land > the agricultural land > barren land > forest. For the agricultural land and barren lands, the TP ECs estimated by the catchment monitoring method were higher than those estimated by the watershed modeling approach and the plot experiment. For the forest,the TP ECs estimated by the plot experiment were higher than those estimated by the catchment monitoring method and the watershed modeling approach. However, there was no significant difference in estimated ECs for the residential land among three approaches.(2) Compared to Europe and North America, Asia usually presented the higher TP ECs of each land-use type, which might be associated with the higher population density and P fertilizer application rate in Asia. Although observed global TP ECs for farmland, residential land and barren land showed no significant changing trends in past 50 years, TP ECs for the forest presented a weakly increasing trend in general.(3) Global TP ECs for the same land-use type showed highly spatial and temporal variability, which related to the variations in precipitation, runoff depth and P fertilizer application rate, as well as to the differences in slope, distance to water bodies, soil types and so on. Observed TP ECs for each of land-use types same showed a power-law fuction with annual precipitation or runoff depth. For the agricultural land,observed TP ECs also increase linearly with the increase of P fertilizer application.(4) The LOADEST mode was able to accurately predicted long-term changes of riverine P export dynamics in Henxi (upstream) and Baizhiao (downstream)monitoring sites of the Yongan River in 1980-2010. Riverine TP export load in Baizhiao and Henxi increased 25.5 and 18.5 times during 1980-2010, respectively.Annual variation of riverine P export load was mainly influenced by precipitation as well as other human activities. There was an 11-year lag times between P fertilizer application and riverine P export based on the cross-correlation analysis.(5) By considering relevant influencing factors, the model that related annual riverine TP export load with point source TP load, atmospheric TP deposition and individual TP ECs of four land-use types with their area was developed and calibrated by the Bayesian statistical method, resulting in high precisions. This Bayesian inversion provides effective means for estimating long-term changes in TP ECs of each land-use type within the watershed.(6) Point source pollution and non-point sources from farmland and residential lands were the main sources for riverine P export in both the Baizhiao and Hengxi sites.Among the four land-use types, the order of estimated TP ECs followed significantly:the residential land > the agricultural land > barren land > forest. Estimated TP ECs were 18-244 times higher in residential and agricultural lands than those estimated in the barren lands and forest.(7) By adopting the TI/D index, estimated TP ECs of each land-use type presented 2-4 orders of magnitude in spatial heterogeneity during the same time period. In terms of critical source areas, we found that 76-80% of total NPS TP pollution load was derived from the regions of <300 m distance away from the river.(8) In the 2000s, scenarios predictions indicated that 40% decrease in domestic wastewater P discharge and moving 10% of farmland area that located in the region of< 300m away from the river to the barren land that located in the region of > 1000m away from the river would reduce 5.4% and 6.7% of riverine P export loads,respectively.(9) This study highlights that both source reduction measures by improving domestic wastewater treatment and reducing excessive P fertilizer application rate and process intercept measures by adopting conservation tillage practices and extending wetland areas are necessary to efficiently control TP pollution in the Yongan watershed.