Measurements, Parameters and Model Uncertainties in Rainfall Interception Modelling
Interception field measurements are in practice uncertain. However and to our knowledge there is no study thoroughly devoted to the analysis of the uncertainties' influence on rainfall interception modelling. To cover this lack of information, the present work exams the uncertainties related to interception modelling in detail, according with two different aspects. In the first part, the standard errors of the measured precipitation, throughfall and stemflow are used to estimate the measurement uncertainty bounds of calculated rainfall interception loss. Confidence limits are estimated for the model parameters derived from lineal regressions. Rutter model simulations are validated not only by means of the fit to experimental data, but also by the use of the measurements uncertainties bounds. In the second part, Generalized Likelihood Uncertainty Estimation (GLUE) method is applied, accounting for experimental data, parameters and model uncertainty together, with the aim of estimating the predictive uncertainty associated with Rutter interception model.
A Method to Infer Interception Evaporation Using Eddy Covariance Measurements: Application Over an Eastern Amazon Old-growth Rain Forest
We develop a general methodology to estimate rainfall interception using eddy covariance data that are available at a large number of worldwide flux tower sites. This method is then demonstrated using data from an old-growth rain forest site in the eastern Amazon. The approach is to estimate the 'excess' evaporation that occurs during and following individual precipitation events, using baseline evaporation time series obtained from long time series of flux data and creating ensemble averages from these precipitation events and base- state dry days. One advantage of this method over the traditional techniques of estimating interception using rain gauges alone is that the interception evaporation is directly measured and not determined as the residual of incident precipitation and throughfall. This method would also be useful in cases where rain gauge measures of precipitation are suspect, such as in fog or wind-driven conditions. Furthermore, the large differences in interception that can occur on a site due to varying forest canopy density, structure and the appearance of canopy gaps is smoothed out using the eddy covariance method as the size of the flux footprint area incorporates these variations, and provides an average interception value over the flux footprint area. Identification of light rainfall events not detected by an on-site tipping bucket rain gauge was aided by the use of a ceilometer. Results from an eastern Amazon old-growth rain forest site (the km67 site in LBA-ECO) show that for daytime events, interception percentages decrease with rainfall intensity, with mean interception for light (0-2 mm/hr), moderate (2-16 mm/hr), and heavy (greater than 16 mm/hr) rainfall-rate events being 18.0, 9.9, and 7.8 percent of incoming precipitation respectively. The mean interception for all events in the study (daytime and nighttime) was 11.6 %. Energy balance comparisons between dry and afternoon rain-days show an approximately 15 % increase of evaporative fraction on the rain days, with the energy being supplied by a corresponding decrease in the canopy heat storage. Future work aims to apply this method to other flux tower sites in varying forest types and climates.
Precipitation partitioning for three land cover classes in the wet-dry tropics of Australia.
The savannas of the northern Australian tropics cover approximately a quarter of the Australian continent. These wet-dry tropical systems are driven by the north Australian monsoon, which has created an extreme environment, with the region's highest rainfall occurring during a 3-4 month wet season, followed by a sustained dry season that lasts the majority of the year. There is little knowledge about the runoff characteristics of most of the catchments in this region, and the impact that wide spread clearing of these savanna systems will have on the water balance is hard to predict. With the increasing pressure to develop the Daly River region in the Northern Territory, an ongoing Tropical Rivers and Coastal Knowledge (TRaCK) project was set up to look at the impact that clearing savanna woodland will have on the water balance of the Daly River catchment. Some of the results from this study are presented in this paper. Three sites within a sub-catchment of the Daly River system have been instrumented to monitor rainfall, evapotranspiration, soil moisture dynamics and deep drainage. The selected sites represent the major land use classes in the Daly River area. These sites are an uncleared woodland savanna, an improved pasture site cleared 25 years ago, and an unmanaged regrowth site that was a native grass pasture cleared 8 years ago. Using data from the past two wet seasons, we have partitioned the precipitation into its evapotranspiration and infiltration components and will show how clearing modifies these water balance components across 3 contrasting land cover types. Preliminary results show that the modified pasture site has the most seasonal evapotranspiration, and that the uncleared woodland savanna is stabilised by the influence of the evergreen eucalypt trees, resulting in the least variable evapotranspiration. We also see that the deep rooting trees in the uncleared savanna site significantly draw on soil water throughout the long dry season, resulting in lower soil moisture and decreased drainage to surface aquifers at this site. Further results will confirm that clearing of the native savanna in the Daly River area will result in an increase in the runoff of the catchment.
Multi-scale Analysis of Surface-Precipitation Feedbacks in the Central United States
Currently the central plains of the United States exhibit signs of a positive soil moisture-vegetation-precipitation feedback regime and understanding the responses of the feedback regimes to regional and global climate change is necessary for many applications. The existence and magnitude of feedback processes between the surface and the atmosphere are investigated using several approaches: low dimensional and regional climate modeling, remote sensing analysis and eddy covariance data. Information theory metrics such as entropy and mutual information content are used to quantify these interactions across spatial and temporal scales via wavelet analysis. Of specific interest are the roles of biosphere-atmosphere coupling, energy balance partitioning, boundary layer dynamics and soil moisture as the dominant controlling mechanisms over the feedback processes. In addition, the stability of this positive feedback regime under altered climate scenarios is investigated using downscaled global climate model output. These results indicate the susceptibility of this region to changes in both local and global climate processes.
The Role of Vegetation Response to Elevated CO2 in Modifying Land-Atmosphere Feedback Across the Central United States Agro-Ecosystem
Recent local-scale observational studies have demonstrated significant modifications to the partitioning of incident energy by two key mid-west agricultural species, soy and corn, as ambient atmospheric CO2 concentrations are experimentally augmented to projected future levels. The uptake of CO2 by soy, which utilizes the C3 photosynthetic pathway, has likewise been observed to significantly increase under elevated growth CO2 concentrations. Changes to the sensible and latent heat exchanges between the land surface and the atmospheric boundary layer (ABL) across large portions of the mid-western US has the potential to affect ABL growth and composition, and consequently feed-back to the near-surface environment (air temperature and vapor content) experienced by the vegetation. Here we present a simulation analysis that examines the changes in land-atmosphere feedbacks associated with projected increases in ambient CO2 concentrations over extended soy/corn agricultural areas characteristic of the US mid-west. The model canopies are partitioned into several layers, allowing for resolution of the shortwave and longwave radiation regimes that drive photosynthesis, stomatal conductance and leaf energy balance in each layer, along with the canopy microclimate. The canopy component of the model is coupled to a multi-layer soil-root model that computes soil moisture and heat transport and root water uptake. Model skill in capturing the sub-diurnal variability in canopy-atmosphere exchange is evaluated through multi-year records of canopy-top eddy covariance CO2, water vapor and heat fluxes collected at the Bondville (Illinois) FluxNet site. An evaluation of the ability of the model to simulate observed changes in energy balance components (canopy temperature, net radiation and soil heat flux) under elevated CO2 concentrations projected for 2050 (550 ppm) is made using observations collected at the SoyFACE Free Air Carbon Enrichment (FACE) experimental facilities located in central Illinois, by incorporating observed acclimations in leaf biochemsitry and canopy structure. The simulation control volume is then extended by coupling the canopy models to a simple model of daytime mixed-layer (ML) growth and composition, ie. air temperature and vapor content. Through this coupled canopy-ABL model we quantify the impact of elevated CO2 and vegetation acclimation on ML growth, temperature and vapor content and the consequent feedbacks to the land surface by way of the near-surface environment experienced by the vegetation. Particular focus is placed on the role of short-term drought, and possible changes in land cover composition between soy, a C3 crop, and corn, a more water-use efficient C4 crop, on modulating the strength of these CO2-induced feedbacks.
Implementation of MODIS Land Cover Distribution in the NCEP Mesoscale Model
The new MODIS satellites remote sensing provides high quality, consistent and well-calibrated land cover data at 1km resolution. This MODIS land cover data with the widely accepted IGBP (International Geosphere- Biospere Programme) classification was prepared by Boston University group Based on this data, we added 3 tundra classes and modified the parameters that govern the flow of water between the vegetation and the atmosphere, which is important for numerical weather predictions over a wide range of spatial and temporal scales. As a first attempt, we implement this MODIS land cover data with a new vegetation parameter table in the NWS mesoscale operational model (WRF). The resulting plot of daytime 2-m temperature averaged over CONUS agrees with observation very well, but the averaged nighttime temperature deviates from observation by almost a degree C. However, after the adjustment of the thermal roughness length that is related to stable layer turbulent transport, the predicted values at night agree with observations very well. This new model has been tested with many cases, and the resulting statistics are consistently better than the NWS mesoscale operational model. In addition, in some cases, the predicted convective instability indices differ from the operational model's result, which indicates the presence of the feedback mechanism by alternating the amount of precipitation as well as soil moisture content.