Development and application of a simple hydrologic model simulation for a Brazilian headwater basin

Physically based hydrologic models for watersheds are important tools to support water resources management and predict hydrologic impacts produced by land-use change. Grande River Basin is located in southern Minas Gerais State, and the Grande River is the main tributary of Basin which has 2080 km² draining into the Camargos Hydropower Plant Reservoir (CEMIG — “Minas Gerais State Energy Company”). The objectives of this work were: 1) to create a semi-physically based hydrologic model in semi-distributed to sub-basins approach and based on GIS and Remote Sensing tools and, 2) to simulate the hydrologic responses of the Grande River Basin, thus creating an important tool for management and planning of water resources for region. The hydrologic model is based on the SCS Curve Number (SCS-CN) and MGB/IPH models, and structured into three hydrologic components: estimation of the flow components (quick runoff, hortonian and base flows), propagation into the respective soil reservoirs (surface, sub-surface and shallow saturated zone) and propagation into the channels. Precipitation and discharge data sets were obtained from the Brazilian National Water Agency (ANA). Reference evapotranspiration (ETo) data were obtained from the Brazilian National Meteorological Institute (INMET). In order to estimate actual evapotranspiration, crop coefficient, soil moisture and satellite image interpretation of actual land-use were applied. The long-term hydrologic data series were structured for period between 1990 and 2003. The calibration and validation process was carried out by evaluating the behavior of the Nash–Sutcliffe Coefficient (C_NS), obtained from three different combinations of calibration and validation years. This allowed us to evaluate the model performance to simulate years in which El Niño (EN) and La Niña (LN) events were registered (1997–1998 and 1999–2000, respectively). The combinations of calibration and validation years were: the first 7 years to calibrate and remaining 6 years to validate; the first 9 years to calibrate and remaining 4 years to validate; and first 11 years to calibrate and the last 2 years to validate. The statistical precision showed that the model was able to simulate the hydrologic impacts, including years of EN and LN events, with C_NS scores greater than 0.70 in both situations. The evaluation of the C_NS scores showed small variation in the coefficient as the years of validation decreased. In addition, the model was also able to simulate the hydrologic impacts of land-use change in the Grande River Basin, based on the C_NS scores of 0.80 for different combinations of validation periods. The hydrologic impacts in Grande River Basin produced from grassland area converted to eucalyptus under three specific scenarios were evaluated, which predicted annual runoff mean reductions of up to 17.8%, due to an increase in evapotranspiration rate for the eucalyptus plantation.     

Development and application of a simple hydrologic model simulation for a Brazilian headwater basin

Physically based hydrologic models for watersheds are important tools to support water resources management and predict hydrologic impacts produced by land-use change. Grande River Basin is located in southern Minas Gerais State, and the Grande River is the main tributary of Basin which has 2080 km² draining into the Camargos Hydropower Plant Reservoir (CEMIG — “Minas Gerais State Energy Company”). The objectives of this work were: 1) to create a semi-physically based hydrologic model in semi-distributed to sub-basins approach and based on GIS and Remote Sensing tools and, 2) to simulate the hydrologic responses of the Grande River Basin, thus creating an important tool for management and planning of water resources for region. The hydrologic model is based on the SCS Curve Number (SCS-CN) and MGB/IPH models, and structured into three hydrologic components: estimation of the flow components (quick runoff, hortonian and base flows), propagation into the respective soil reservoirs (surface, sub-surface and shallow saturated zone) and propagation into the channels. Precipitation and discharge data sets were obtained from the Brazilian National Water Agency (ANA). Reference evapotranspiration (ETo) data were obtained from the Brazilian National Meteorological Institute (INMET). In order to estimate actual evapotranspiration, crop coefficient, soil moisture and satellite image interpretation of actual land-use were applied. The long-term hydrologic data series were structured for period between 1990 and 2003. The calibration and validation process was carried out by evaluating the behavior of the Nash–Sutcliffe Coefficient (C_NS), obtained from three different combinations of calibration and validation years. This allowed us to evaluate the model performance to simulate years in which El Niño (EN) and La Niña (LN) events were registered (1997–1998 and 1999–2000, respectively). The combinations of calibration and validation years were: the first 7 years to calibrate and remaining 6 years to validate; the first 9 years to calibrate and remaining 4 years to validate; and first 11 years to calibrate and the last 2 years to validate. The statistical precision showed that the model was able to simulate the hydrologic impacts, including years of EN and LN events, with C_NS scores greater than 0.70 in both situations. The evaluation of the C_NS scores showed small variation in the coefficient as the years of validation decreased. In addition, the model was also able to simulate the hydrologic impacts of land-use change in the Grande River Basin, based on the C_NS scores of 0.80 for different combinations of validation periods. The hydrologic impacts in Grande River Basin produced from grassland area converted to eucalyptus under three specific scenarios were evaluated, which predicted annual runoff mean reductions of up to 17.8%, due to an increase in evapotranspiration rate for the eucalyptus plantation.     

Modeling hydrologic responses using multi-site and single-site rainfall generators in a semi-arid watershed

Hydrologic response in a watershed is driven by precipitation. Multi-site rainfall generators can be used to model watersheds using spatially varied rainfall inputs to better analyze how the rainfall variability affects runoff generation. This study adopted both a single-site rainfall generator (CLIGEN) and a multi-site rainfall generator to generate two rainfall data sequences, which were then used to drive the Soil and Water Assessment Tool (SWAT) for runoff simulation. The 148-km² Walnut Gulch Experimental Watershed and its two sub-watersheds were selected to evaluate the hydrologic response. Runoff calibration was done against measured runoff in the watershed. Statistics showed that the single-site and multi-site rainfall generators gave similar results regarding annual precipitation. However, the multi-site generator performed much better than the single-site generator in both mean summer flow and for the different return period flows. The runoff derived from the single-site generator was significantly over-estimated in all three watersheds. As for the multi-site generator, the derived runoff was satisfactorily predicted in the smaller watersheds but only overestimated in the largest watershed. This indicated that in small to medium sized watersheds, the spatial variability of rainfall could play an important role for hydrologic response because of the heterogeneity of convective rainfall in this semi-arid region, which makes the application of multi-site rainfall generator a better option than the single-site generator.     

Prediction of soil and water conservation structure impacts on runoff and erosion processes using SWAT model in the northern Ethiopian highlands

Purpose

Land degradation due to soil erosion is a serious threat to the highlands of Ethiopia. Various soil and water conservation (SWC) strategies have been in use to tackle soil erosion. However, the effectiveness of SWC measures on runoff dynamics and sediment load in terms of their medium- and short-term effects has not been sufficiently studied.

Materials and methods

A study was conducted in 2011 to 2015 in the Gumara-Maksegnit watershed to study the impacts of SWC structures on runoff and soil erosion processes using the soil and water analysis tool (SWAT) model. The study was conducted in two adjacent watersheds where in one of the watersheds, SWC structures were constructed (treated watershed (TW)) in 2011, while the other watershed was a reference watershed without SWC structures (untreated watershed (UW)). For both watersheds, separate SWAT and SWAT-CUP (SWAT calibration and uncertainty procedure) projects were set up for daily runoff and sediment yield. The SWAT-CUP program was applied to optimize the parameters of the SWAT using daily observed runoff and sediment yield data.

Results and discussion

The runoff simulations indicated that SWAT can reproduce the hydrological regime for both watersheds. The daily runoff calibration (2011–2013) results for the TW and UW showed good correlation between the predicted and the observed data (R ²?=?0.78 for the TW and R ²?=?0.77 for the UW). The validation (2014–2015) results also showed good correlation with R ² values of 0.72 and 0.70 for the TW and UW, respectively. However, sediment yield calibration and validation results showed modest correlation between the predicted and observed sediment yields with R ² values of 0.65 and 0.69 for the TW and UW for the calibration and R ² values of 0.55 and 0.65 for the TW and UW for the validation, respectively.

Conclusions

The model results indicated that SWC structures considerably reduced soil loss by as much as 25–38% in the TW. The study demonstrated that SWAT performed well for both watersheds and can be a potential instrument for upscaling and assessing the impact of SWC structures on sediment loads in the highlands of Ethiopia.

     

SWAT模型在粤北连江流域的应用研究

选择粤北连江流域为研究区域,以分布式水文模型SWAT作为模拟工具,对流域内的水文过程进行模拟。利用流域内高道、凤凰山和黄麖塘3个水文站2001—2010年的实测月平均径流量进行敏感性分析和参数率定。以2001—2005年作为校准期,2006—2010年作为验证期,以相对误差（Re）、决定系数（R²）以及Nash-Suttcliffe效率系数（Ens）作为模型适用性的评价指标。校准期3个水文站的月径流量模拟值的相对误差分别是2.72%,5.91%,1.63%,决定系数均大于0.9,Nash-Suttcliffe系数分别为0.97,0.89,0.70,而验证期相对误差分别是2.62%,5.36%,9.32%,决定系数均大于0.9,Nash-Suttcliffe系数分别为0.90,0.69,0.69。各项评价指标均符合精度要求,说明SWAT模型可以用于连江流域的径流模拟。     

SWAT模型在黄土丘陵区参数敏感度分析及率正研究

SWAT模型是基于水文过程的、具有很强物理机制的流域分布式水文模型。应用SWAT模型,结合1986—2000年罗玉沟流域实测水文资料对径流进行模拟,并进行了模型参数敏感度分析和率正。通过敏感度分析,辨析出影响该流域产流模拟结果精度的主要参数因子,如径流曲线数CN2等。依据上述结论调整参数值,对模型进行了率正和验证,得出率正期的相对误差为14.0%,R²为96.1%,Nash—Suttcliffe系数为0.89;验证期的相对误差为8.8%,R²为91.5%,Nash—Suttcliffe系数为0.82。结果表明实测流量和模拟流量的线性回归系数和模型的效率系数满足模型模拟的要求,SWAT模型适用于该研究区域。     

基于3个站点校准与验证的晋江流域径流模拟

 应用分布式水文模型SWAT,对东南沿海晋江流域内3个水文站点的年、月径流和基流分别进行模拟。选取1972—1975年作为模型校准期,以效率系数、平均误差、相对误差和决定系数为评价指标,率定出6个模型敏感参数,并用1976—1979年的资料进行模型验证。结果表明:利用3个站点模拟的方法可以从流域尺度上更客观地率定模型参数,改善模型的模拟效果;模型在东南沿海流域的模拟效果较好,精度较高;模拟期内降水量的差异以及水库年径流调节均会对水量模拟结果造成影响。     

基于SWAT模型的径流与土壤侵蚀过程模拟

将可表征土地利用变化对水文响应影响的分布式水文模型SWAT应用于密云水库潮河子流域,对模型的敏感性进行了分析,并采用1986～1991年下会水文站数据对其径流量和泥沙负荷进行了率定和验证,模型评估结果表明该模型对径流的模拟效果非常好,对泥沙的模拟效果较好。通过计算34个子流域内的侵蚀模数,对流域内的土壤侵蚀强度进行了分级,并分析了不同土地利用方式对产沙的影响。结果表明,流域内的土壤侵蚀主要发生在农田,为保障密云水库的水质应对其采取有效的水土保持措施。     

河套灌区不同水肥处理对玉米生长影响的AquaCrop模型模拟

为探究内蒙古河套灌区玉米生长适宜的水肥管理方案,开展了不同灌水与施肥水平玉米田间试验。利用田间实测数据对AquaCrop模型进行了率定与验证,以此为基础,采用AquaCrop模型模拟分析了不同灌水和施肥条件下玉米产量和水分利用效率的变化规律。结果表明:(1)模型率定和验证过程中冠层覆盖度和生物量模拟值与实测值的R~2分别介于0.74～0.99,0.87～0.99;NRMSE分别介于4.55%～12.32%,5.77%～27.07%;E_(NS)分别介于0.90～0.99,0.85～0.99;各处理产量模拟值与实测值间R~2分别为0.99,0.97;NRMSE分别为4.59%,3.42%,E_(NS)分别为0.95,0.97;水分利用效率模拟值与实测值间R~2分别为0.81,0.86;NRMSE分别为6.75%,13.85%,E_(NS)分别为0.96,0.83。说明AquaCrop模型在河套灌区具有很好的适用性。(2)利用校验后的模型模拟了不同施肥水平下灌水量变化对玉米产量和水分利用效率的影响,当灌水量达270 mm后,继续增加灌水量,玉米产量和水分利用效率基本保持不变,甚至有所降低。(3)以稳产高效为目标,中肥条件下,灌水270 mm时玉米产量和水分利用效率均为各灌水量中的最优值。因此,推荐灌水270 mm、施肥375 kg/hm~2为研究区保障玉米稳产高效的最优灌溉施肥组合。     

How Will Climate Change Affect Road Salt Export from Watersheds?

Salt applied to roads for de-icing can persist for decades in watersheds, and understanding rates of salt accumulation and export is important for anticipating the magnitude and duration of its environmental impacts. Salt removal is affected by storm frequency and intensity, both of which are projected to change as a result of anthropogenic climate change. To examine the potential outcomes of changing storm regimes on rates of salt export from watersheds, we studied chloride concentrations in baseflow and surface runoff in 5 years of streamflow data (taken at 20-min intervals) in a headwater tributary of the Hudson River. Baseflow, with chloride concentrations of 200–230 mg/L year-round, dominated streamflow, except during and after large storms, and accounted for about 90% of chloride export from the watershed. Extreme precipitation events (e.g., tropical storms) increased salt export mainly by increasing baseflow discharge, which remained elevated long after storms ended. While intense events accomplished more salt export per storm, they did not produce disproportionately higher export per volume of precipitation, compared to smaller rain events. Chloride export in a wet year was almost twice that in a dry year, despite less than a doubling of rainfall. Under future climate conditions, if winter precipitation shifts toward a higher proportion of rainfall, or summers experience a greater frequency of tropical storms, then long-term residence time of road salt in watersheds may be shorter than previously anticipated. Conversely, if climate becomes drier, with increased evapotranspiration, reduced infiltration and baseflow discharge may prolong salt storage, which could exacerbate biotic stresses from high chloride concentrations in streams.     

气候变化对开都河流域径流的影响研究

利用SWAT模型模拟开都河流域的径流变化,并采用1990—2009年的水文站点径流数据进行精度验证,然后设定气候变化情景,模拟不同气候条件下径流的响应特征。结果表明:模拟结果与实测径流较吻合,剔除异常年份（1994年、1995年）后,校准期（1990—2000年）效率系数为0.58,平均相对误差为-5.7%,线性拟合度为0.8;验证期（2000—2009年）的结果与校准期接近,均达到了模型的评价标准,说明SWAT模型在开都河流域的适用性较好。基于此,采用任意情景设置方法,设置了25种气候变化（气温和降水）组合情景,研究了该流域对气候变化的响应,结果表明,气候变化对径流量的影响较为显著,降水增加或气温降低均会导致径流量增加,流域未来年均径流变化的主要影响因素是降水,温度的影响相对较弱。     

四川丘陵区农林系统和农草系统流域的径流和土壤水分特征

A study was conducted in a hilly area of Sichuan Province, Southwestern China, to compare the streamflow and soil moisture in two upland watersheds with different land use patterns. One was an agroforestry watershed, which consisted mainly of trees with alder (Alnus cremastogyne Burkill) and cypress (Cupressus funebris Endl.) planted in belts or strips with a coverage of about 46%, and the other was a grassland primarily composed of lalang grass (Imperata cylindrica var. major (Nees) C. E. Hubb.), filamentary clematis (Clematis filamentosa Dunn) and common eulaliopsis (Eulaliopsis binata (Retz.) C. E. Hubb) with a coverage of about 44%. Streamflow measurement with a hydrograph established at the watershed outlet showed that the average annual streamflow per 100 mm rainfall from 1983 to 1992 was 0.36 and 1.08 L s^-1 km^-2 for the agroforestry watershed and the grass watershed, respectively. This showed that the streamflow of the agroforestry watershed was reduced by 67% when compared to that of the grass watershed. The peak average monthly streamflow in the agroforestry watershed was over 5 times lower than that of the grass watershed and lagged by one month. In addition, the peak streamflow during a typical rainfall event of 38.3 mm in August 1986 was 37% lower in the agroforestry watershed than in the grass watershed. Results of the moisture contents of the soil samples from 3 slope locations (upper, middle and lower slopes) indicated that the agroforestry watershed maintained generally higher soil moisture contents than the grass watershed within 0-20 and 20-80 cm soil depths for the upper slope, especially for the period from May through July. For the other (middle and lower) slopes, soil moisture contents within 20-80 cm depth in the agroforestry watershed was generally lower than those in the grass watershed, particularly in September, revealing that water consumption by trees took place mainly below the plow layer. Therefore, agroforestry land use types might offer a complimentary model for tree-annual crop water utilization.     

基于混合产流模式的黄土高原流域水沙过程耦合模拟

为深入理解黄土高原干旱半干旱地区复杂地貌条件下流域水沙运移规律。基于垂向混合产流机理和运动波方程，构建分布式流域水文模型，耦合流域土壤侵蚀和泥沙输移过程模拟模块，并考虑梯田对水沙过程的影响，建立适用于黄土高原的分布式流域水沙过程模型。选取黄土高原延河支流西川河流域多年实测场次洪水过程的径流泥沙资料，对模型进行率定和验证。径流模拟的纳什效率系数在0.56以上，平均值超过0.70,模拟次洪过程的峰形、峰值、峰现时间与实测过程具有较好的一致性；侵蚀产输沙模拟精度较低，其纳什效率系数均值率定期为0.79,但验证期仅为0.45,模拟结果整体趋势与实测值较一致，但输沙量模拟峰值比实测值偏低。模型可以较精确地模拟黄土高原流域洪水产汇流过程，但输沙量模拟值偏低，一方面由于产汇流模块的误差传递；另一方面，对重力侵蚀考虑不足。因此，未来模型将考虑滑坡、崩塌等重力侵蚀过程，提升模拟精度和效率，为流域水沙过程模拟与流域综合治理提供有效工具。     

Hillslope nutrient flux during near-stream vegetation removal

At the Coweeta Hydrologie Laboratory in the southern Appalachians of western North Carolina, a near-stream vegetation manipulation experiment is being conducted to determine the effect of removal of streamside Rhododendron maximum L. on the export of hillslope nutrients (K, Na, Ca, Mg, N, P, S) and organic matter. Experimental hillslope transects that span topographical flowpaths from a local highpoint to the stream have been instrumented with lysimeters and TDR rods at two depths, as well as with streambed and streambank piezometers. We present a review of studies of nutrient flux in the riparian zone of forested watersheds. In the southern Appalachians, we hypothesize that R. maximum is a keystone species at the interface between terrestrial and aquatic systems, with extensive near-stream thickets having a possible impact on carbon and nutrient transport into streams. We present the conceptual basis and initial implementation of a model-based experimental design to test the effect of R. maximum removal on hillslope nutrient and organic matter export in upland watersheds. The model is terrain-based and will be used to extrapolate elemental flux measurements both spatially from the hillslope to watershed scale and temporally for various climate regimes. The model consists of three modules: (1) objective terrain analysis (TAPES-C); (2) a dynamic interception canopy module; (3) a hillslope hydrology module (IHDM4) with a 2-D Richard's equation of subsurface moisture dynamics. Calibration and validation of the model will occur at two scales: at the hillslope scale, using well, lysimeter, and TDR data; at the watershed scale, using streamflow measurements across a variety of storm types. We show watershed terrain analysis for the experimental watershed (WS56) and discuss use of the model for understanding effects of watershed management of riparian zone processes.     

概念性土壤侵蚀模型的建立及在紫色土小流域的应用

概念性土壤侵蚀模型在描述流域基本过程的同时对参数要求较低,现有概念性模型中计算产沙所需参数主要靠律定或借鉴经验值。该文介绍了1个无需流域出口产沙量长序列观测资料来律定且能适用于研究流域的概念性土壤侵蚀模型。模型中引入了分布式水文模型WetSpa Extension作为水文模块,结合流域内试验小区上建立的流量—产沙量经验关系计算侵蚀量,再结合泥沙输移比构建起产沙模块。通过在紫色土地区小流域的应用表明,模型能够得到较合理的流域出口产流量、产沙量以及侵蚀率的空间分布等模拟结果,在没有试验小区的邻近流域地区也具有推广性,且能作为评价水保措施效益的有力工具。该概念性模型对于流域出口泥沙资料稀少的地区具有很高的应用价值和协助流域管理的 潜力。     

基于SWAT模型的南渡江流域土地利用/覆被变化的径流响应

为了探究热带岛屿性流域土地利用/覆被变化(Land Use/Cover Change,LUCC)的水文效应,以海南岛典型热带岛屿性流域南渡江流域为研究区,构建本地化SWAT模型,模拟流域水文过程及其对LUCC的响应,揭示流域径流的变化规律,对比不同土地利用类型的水文过程及调蓄能力,探讨了典型热带岛屿性流域LUCC的径流效应。结果表明:SWAT模型在南渡江流域的适用性良好,径流模拟对植被蒸散发、产生径流的坡面状况、土壤状况和地下水过程较敏感。不同土地利用类型对流域的产流贡献顺序为耕地>其他林地>橡胶林地; 耕地产流能力最强,其他林地和橡胶林地具有一定的截流蓄水作用。综上所述,与1990年相比,2015年南渡江流域河流径流量减少,流域LUCC导致的地表径流减少是河流径流量减少的直接原因,而蒸散发加剧是地表径流减少的主要原因。     

Long-Term Hydrological Impacts of Land Use/Land Cover Change From 1984 to 2010 in the Little River Watershed,Tennessee

Assessing long-term hydrological impacts of land use/land cover (LULC) change is of critical importance for land use planning and water resource management. The Little River Watershed, Tennessee, is an important watershed supporting drinking water and recreational activities within and around the Great Smoky Mountains National Park in the Unites States. However, the potential hydrological impacts of LULC change, especially urbanization in recent decades, are not quantified. This paper assessed the long-term impacts of LULC change on streamflow and non-point source pollution using the Soil and Water Assessment Tool (SWAT) and a detailed LULC record from 1984 to 2010. The SWAT was first calibrated and validated using observed streamflowin 2010 and then simulated using different LULC patterns in 1984-2010 to quantify the long-term hydrological impacts caused by the LULC change. Simulated results indicated a minor 3% increase in streamflow for the whole watershed from 1984 to 2010, but with a distinct spatial pattern. The increase in streamflow is closely related to urban development. Almost no streamflow increase occurred in the upper watershed within the national park, whereas > 10% increase occurred in the lower watershed, especially in areas close to cities. Model simulation also suggested 34.6% reduction in sediment and about 10% reduction in nutrient loads from 1984 to 2010, closely related to the decrease in agricultural land. However, without calibration and validation, the simulated reduction in the sediment and nutrient loads may be problematic because SWAT mainly simulates the static LULC patterns, whereas LULC transitions, such as construction phases, may generate more sediment and nutrient loads. In addition, the simulation also did not account for the sediment and nutrients generated from stream bank erosion.     

Streamflow and Soil Moisture of Agroforestry and GrassWatersheds in Hilly Area

A study was conducted in a hilly area of Sichuan Province, Southwestern China, to compare the streamflow and soilmoisture in two upland watersheds with different land use patterns. One was an agroforestry watershed, which consistedmainly of trees with alder (Alnus cremastogyne Burkill) and cypress (Cupressus funebris Endl.) planted in belts or stripswith a coverage of about 46%, and the other was a grassland primarily composed of lalang grass (Imperata cylindricavar. major (Nees) C.E.Hubb.), filamentary clematis (Clematis filamentosa Dunn) and common eulaliopsis (Eulaliopsisbinata (Retz.) C. E. Hubb) with a coverage of about 440/o. Streamflow measurement with a hydrograph established atthe watershed outlet showed that the average annual streamflow per 100 mm rainfall from 1983 to 1992 was 0.36 and1.08 L s^-1 km^-2 for the agroforestry watershed and the grass watershed, respectively. This showed that the streamflowof the agroforestry watershed was reduced by 67% when compared to that of the grass watershed. The peak averagemonthly streamflow in the agroforestry watershed was over 5 times lower than that of the grass watershed and lagged byone month. In addition, the peak streamfiow during a typical rainfall event of 38.3 mm in August 1986 was 37% lowerin the agroforestry watershed than in the grass watershed. Results of the moisture contents of the soil samples from 3slope locations (upper, middle and lower slopes) indicated that the agroforestry watershed maintained generally highersoil moisture contents than the grass watershed within 0-20 and 20-80 cm soil depths for the upper slope, especially forthe period from May through July. For the other (middle and lower) slopes, soil moisture contents within 20-80 cm depthin the agroforestry watershed was generally lower than those in the grass watershed, particularly in September, revealingthat water consumption by trees took place mainly below the plow layer. Therefore, agroforestry land use types mightoffer a complimentary model for tree-annual crop water utilization.     

基于全局敏感性分析和贝叶斯方法的WOFOST作物模型参数优化

作物模型参数的敏感性分析、标定和验证可以提高模型的效率和精准度,进而为模型应用做好准备工作。该研究结合参数全局敏感性分析方法以及贝叶斯后验估计理论的马尔科夫蒙特卡洛(Markov Chain Monte Carlo,MCMC)方法,以华北栾城站三年的冬小麦观测数据(叶面积和地上生物量)为参照,对WOFOST模型的55个品种参数进行了敏感性分析、筛选和优化。发现:1)对叶面积影响较大的参数为:生育期为0、0.5、0.6和0.75时的比叶面积、生育期为1.5时的最大光合速率、叶面积指数最大增长率;对地上干物质影响较大的参数为:生育期为1.5时的最大光合速率、生育期为0时的比叶面积、35℃时叶面积的生命周期、生育期为0时的散射消光系数、生育期为1.8时的最大光合速率、储存器官的同化物转换效率。2)潜在和雨养产量水平下,最大叶面积和地上生物量对参数的敏感性差异不大。3)马尔科夫蒙特卡洛方法(MCMC)可以对WOFOST模型品种参数较好地优化;设计的3种校正-验证方案中,第1种方案(用1998-1999年作为校正年份,1999-2000年,2000-2001年作为验证年份)模拟效果最好。4)优化后的参数,模型对潜在产量水平模拟较好,一致性指数均大于0.9,相对均方根误差小于20%;而对有水分胁迫的雨养情况下比潜在产量水平的模拟结果差,表明模型对水分胁迫的模拟不足。该研究为WOFOST模型区域应用和模型调整优化提供科学理论依据。     

Vine Copula与贝叶斯模型平均结合的月径流预测及应用

准确可靠且预见期较长的月径流预测对水资源配置、防汛抗旱以及生态环境保护等具有重要意义。径流变化与降水、气温、潜在蒸散发以及前期径流等存在密切联系。鉴于Vine Copula可以灵活地将多个随机变量的边缘分布函数通过Copula对的形式联结起来构造多维联合分布函数以及贝叶斯模型平均(Bayesian Model Averaging,BMA)在处理多模型集合预报方面的优势,该研究基于BMA集合多个Vine Copula模型提出了一种BVC径流预测模型(简称BVC模型),应用于黄河流域上游4个水文站(唐乃亥站、民和站、红旗站和折桥站)的月径流预测,采用确定性系数(R²)、纳什效率系数(Nash-Sutcliffe Efficiency coefficient,NSE)和均方根误差(Root Mean Squared Error,RMSE)评价模型的预测性能。结果表明,验证期内预见期为1～3个月时,BVC模型在各水文站的R²均大于等于0.83、NSE均大于等于0.78且RMSE均维持在较低水平;与随机森林(Random Forest,RF)模型和长短...