observation and modeling of vertical heat exchange of lakes in namco basin, tibetan plateau
Binbin Wang is a PhD student in the Department of Water Resources. His supervisors are prof.dr. Z. Su from the Faculty of Geo-information Science and Earth Observation and prof.dr. Y. Ma from the Chinese Academy of Sciences.
Lakes are an important part of the landscape on the Tibetan Plateau (TP). They impact atmosphere boundary layer processes and are thus important for local climate modeling and catchment-scale water and heat budget analysis. However, due to lack of observational data and adequate modeling systems, studies that examine lake-atmosphere interaction and evaporation are very limited and the results show large uncertainties. Thus, observation and modeling of vertical heat exchange including the characteristics of lake-atmosphere interaction, the precise evaporation of water bodies, the energy budget and water balance, the lakes’ response to climate change show high priority and strong significance for research over high-elevation lakes of the TP.
To understand the vertical heat exchange in lakes over the TP, the closed lake basin of Nam Co is chosen as our study area, where meteorological and hydrological observations over a grass land station since 2005 could provide enormous climatological data. In addition, field experiments with eddy covariance (EC) observation systems, four component radiation sensors and instruments for measuring water temperature profiles were carried out in the small Nam Co lake (“small lake” for short hence force) during April 2012 to October 2014 and in the island of Nam Co lake (“large lake” for short hence force) during July 2015 to now. These comprehensive datasets together with the supplemental satellite data can be used to address for the above mentioned issues in Chapter 3 to Chapter 6, respectively.
Firstly, Chapter 3 focuses on observation and modeling of lake-atmosphere interaction processes of the “small lake”. We found that the lake-atmosphere temperature and humidity gradients are positive during the ice-free season, and unstable and neutral atmosphere conditions dominate in the “small lake”. The typical value of roughness lengths for momentum is , while the roughness length for water is larger than that for heat. As influenced by free convection, it gives a square root dependence of latent heat flux on wind speed. After that, by selecting observations of lake-atmosphere interaction through footprint and data quality control, two lake-atmosphere interaction methods (Bulk aerodynamic transfer method and multi-layer method, B method and M method for short hence force) are evaluated by in-situ eddy covariance observations. The results indicate that the two methods show high consistency, but both show underestimation of turbulent heat flux compared with observations. The underestimation of turbulent heat flux is found to be related to the underestimation of roughness lengths. After we optimized the Charnock number from 0.013 to 0.031 and the roughness Reynold number from 0.11 to 0.54 through EC observations, the simulated momentum roughness lengths are also much closer to the observations and the simulated friction velocity and turbulent heat flux also show obvious improvement.
Secondly, Chapter 4 deals with the precise estimation of evaporation, understands the physical controls on turbulent flux and finally analyzes lakes’ energy budget by using in-situ observations introduced in Chapter 2 and the B method evaluated in Chapter 3. Because of the limitation due to eddy covariance observations at the lake shore, the B method, with parameters optimized for momentum roughness length in the two water bodies, is used to provide reliable and consistent results for data interpolation of EC measurements with inadequate footprint or bad quality due to malfunction of the EC instrument. After that, we found that: the diurnal variation of sensible heat flux and latent heat flux have quite different diurnal patterns, with the former peaking in the early morning and the latter peaking in the afternoon. For the “small lake”, wind speed shows significance at temporal scales of half-hourly, whereas water vapor and temperature gradients have higher correlations over temporal scales of daily and monthly in lake-air turbulent heat exchange. For the “large lake”, temperature gradient has higher correlation coefficient to sensible heat flux than that by wind speed, while wind speed has larger correlation coefficient to latent heat flux than that by water vapor gradient. The evaporation during the ice-free season (April to November) of the “small lake” is approximately 812 mm while the evaporation during the ice-free season (May to January) of the “large lake” is around mm. The energy budget during the open water period of the “small lake” is generally closed, with a value of approximately 0.97; while the energy budget closure ratio during July to November of the “large lake” is 0.859.
After that, with the obtained data series of meteorological variables and turbulent heat flux in the two water bodies, Chapter 5 explores the differences of lake-atmosphere interaction parameters, meteorological variables and turbulent heat fluxes between the “small lake” and the “large lake”, and significant differences exist in their lake-atmosphere interaction processes due to differences in their inherent attributes and environmental backgrounds. Relative to the “small lake”, maximum surface temperature of the “large lake” is approximately 3 lower, in addition to a larger wind speed, a higher monthly average air temperature and delayed peaks of seasonal variations of water and air temperature. The typical values of roughness length and standard bulk transfer coefficient for momentum are about 80% and 21% higher respectively in the “large lake”. The typical values of roughness lengths for heat and water are one order of magnitude lower in the “large lake” while the corresponding standard bulk transfer coefficients are only 7% lower. The latent and sensible heat fluxes of the two lakes have quite different seasonal variations, with evaporation peaking in November over the “large lake” and in June over the “small lake”. The estimated evaporation during ice-free season of the “large lake” (around mm) is also higher than that (812 mm) in the “small lake”, attributing to its observed higher air temperature, longer ice free period, lower surface albedo, lower water surface temperature and also in relation to its lower Bowen ratio. Our results show evidences that it is inappropriate to evaluate lake evaporation by Pan observations, especially for its seasonal variation.
Lastly, Chapter 6 quantifies lake evaporation over the two high-elevation lakes, investigates lakes’ responses to climate change and determines the dominant driving forces behind. Two methods (one for traditional evaporation estimation method in the “small lake” and the other with Flake modeling in the “large lake”) are used for long term trend analysis:
For the first method, ten methods for estimating evaporation at a temporal resolution of 10 days over the “small lake” were evaluated by using eddy covariance (EC) observation-based reference datasets. After examination of the consistency of the parameters used in the different methods, the ranking of the methods under different conditions are shown to be inconsistent. The Bowen ratios derived from meteorological data and EC observations are consistent, and it supports a ranking of energy-budget-based methods (including the Bowen Ratio Energy Budget, Penman, Priestley-Taylor, Brutsaert-Stricker and DeBruin-Keijman methods) as the best when heat storage in the water can be estimated accurately. The elevation-dependent psychometric constant can explain the differences between the Priestley-Taylor and DeBruin-Keijman methods. The Dalton-type methods (Dalton and Ryan-Harleman methods) and radiation-based method (Jensen-Haise) all improve significantly after parameter optimization, with better performance by the former than the latter. The deBruin method yields the largest error due to the poor relationship between evaporation and the drying power of the air. The good performance of the Makkink method, with no significant differences before and after optimization, indicates the importance of solar radiation and air temperature in estimation of lake evaporation. The Makkink method was used for long-term evaporation estimation due to lack of water temperature observations in lakes on the TP. Lastly, long-term evaporation during the open-water period (April 6th to November 15th from 1979 to 2015) were obtained; the mean bias was only 6% for average over years 2012 and 2013. A decreasing-increasing trend in lake evaporation with a turning point in 2004 was noted, and this trend corresponds to the published decreasing-increasing trend in reference evapotranspiration on the TP and can be explained by variations in related meteorological variables.
For the second method, the performances of Flake modeling through ITP (Institute of Tibetan Plateau Research) forcing are evaluated by eddy covariance and meteorological data over the “small lake” during 2012-2013 and over the “large lake” during 2015-2016. The results indicate that the observed mixed layer depth () in the “large lake” show clearly diurnal variation with monthly averaged amplitude of approximately 8 m and it results from the significant surface warming during the day and surface cooling at night. Flake simulations could reproduce the seasonal variations of water surface temperature () and over daily and seasonal resolutions, but the amplitude of simulated and are significantly underestimated. Further, the seasonal variations of simulated sensible heat flux (H) and latent heat flux (LE) are close to the observations with a proper extinction coefficient and lake depth, with RMSE values of simulated daily , H and LE of only about 1 , 8 W m-2 and 22 W m-2 respectively. The simulated LE through land-dominated forcing shows clear underestimation compared to that by lake-dominated forcing, and the reasons result from the observed larger wind speed and warmer air temperature in the latter. In addition, no significant differences exist for the simulations with different forcing used in the “small lake”. Lake warming and increasing trends of simulated H and LE are found through long term simulations of corrected ITP forcing. Downward longwave radiation (), rather than air temperature, is considered to play the dominant role in lakes’ response to climate change. Our results found the significance of lake-dominant observations in lake modeling over the “large lake” and suggest the importance of in lake warming and in trends of simulated H and LE.
