North American Drought in the 21st Century Evaluation and Enhancement of Community Land Model Hydrology Dennis P. Lettenmaier1, Eric. F. Wood, Gordon B. Bonan3, Kaiyuan Y. Li1, and Justin Sheffield2 1. Department of Civil and Environmental Engineering, University of Washington 2. Department of Civil and Environmental Engineering, Princeton University; 3. National Center for Atmospheric Research 1 The HAPEX-MOBILHY Soybean Site (43.7ºN 0.1ºW) Introduction The Colorado basin (4) 10 CLM layers are aggregated back to 3 VIC layers and base flow is calculated using VIC drainage formulation; modified soil moisture is carried to the next time step, and the loop repeats with step 1. A key shortcoming of past studies of future climate and drought is that the land surface representations used in the climate models, in general, have not been able to produce realistic land surface hydrologic conditions. Furthermore, past studies that have evaluated the potential for future drought have not considered the possible role of vegetation change. We propose to use the Community Climate System Model (CCSM), in conjunction with a upgraded version of the Community Land Model (CLM) to evaluate the susceptibility of the U.S. to drought over the next century. Existing studies have shown that CLM has serious deficiencies in terms of hydrologic predictions , and therefore evaluation and enhancement of CLM hydrology are necessary before it is coupled into CCSM to predict the future drought. During the first phase of the project, we have made great efforts on evaluating and upgrading the CLM hydrology. The results summarized in this poster include: (1) Brief description of the basins and point sites, as well as datasets used for the model evaluation; (2) the model evaluation aimed at identifying the weakness of the CLM hydrology; (3) the methodology for incorporating VIC surface runoff and baseflow parameterization into CLM; and (4) the performance evaluation of the Upgraded CLM hydrologic prediction. VIC well reproduced the observed streamflow both in terms of magnitude and seasonality. CLM largely overestimated runoff, and the snow-melting dominated runoff peak came one month earlier than observation. VIC reproduced both the latent and sensible heat reasonably well, while CLM underestimated latent heat and overestimated sensible heat for the growing season. In the non-growing season (before sowing, Jan. – Apr., and after harvest, Oct – Dec) CLM markedly overestimated bare ground evaporation and hence underestimated soil moisture content as compared to the observation and VIC simulation. This underestimation of soil moisture by CLM for the period before planting (Jan. – Apr.) subsequently led to the underestimation of evapotranspiration for the growing period of early May through late August. The canopy interception for CLM is similar to the one for VIC. 10-Layer CLM 3-Layer VIC VIC Upper Layer 3.43 m VIC Lower Layer 2 Basin/site and dataset description We evaluate CLM by comparing with VIC in terms of their ability to reproduce observed water and energy fluxes in off-line tests for three large basins with contrasting hydroclimatic conditions spanning the range from temperate continental to Arctic, and four point (column flux) sites spanning the range from tropical to arctic. The three large basins are Arkansas-Red and Colorado basins in U.S., and Torne-Kalix in northern Scandinavia. The forcing data, soil and vegetation parameters are from LDAS and PILPS projects, and the naturalized streamflow data used to evaluate the model are from the Tulsa District of the U.S. Army Corps of Engineers for the Arkansas-Red basin and from the U.S. Bureau of Reclamation for the Colorado basin. The column flux evaluations are for a tropical forest site at Reserva Jaru (ABRACOS) in Brazil, a prairie site near Manhattan (FIFE), Kansas in central U.S., a soybean site at Caumont (HAPEX-Monbilhy) in France, and a small grassland catchment at Valdai in central Russia. Dynamic Depth Fixed Depth Diagram: Matching CLM layer scheme to VIC Layer scheme The Valdai Grassland Site (57.6ºN 33.1ºE) VIC parameters required in Upgraded CLM The FIFE Prairie Site (39.0ºN 96.5ºW) b_infil: Infiltration parameter; Depth: Upper layer depth (first plus second VIC layer depth); Ws: Fraction of maximum soil moisture content when baseflow occurs; Dsmax: Maximum velocity of baseflow; Ds: Fraction of Dsmax where non-linear baseflow occurs. The Upgraded CLM performed similarly to VIC in terms of surface and baseflow prediction, while the original CLM overestimated the surface runoff compared to VIC. The Upgraded CLM significantly reduced the surface runoff compared to the original CLM, thereby correcting the underestimation of the subsurface soil moisture and the latent heat, and the slight overestimation of the sensible heat by the original CLM. This site demonstrates that the runoff parameterization plays a critical role in water and energy balance prediction. Valdai site is a small catchment (0.36 km2) in central Russia, mainly covered by grassland. CLM and VIC performed similarly in the simulation of evapotranspiration except that CLM underestimated the peak value. Soil moisture contents were poorly simulated by CLM with lower peak value than observations. 3 Model evaluation of CLM The Arkansas-Red basin VIC well reproduced observed runoff in terms of both magnitude and seasonality. CLM well captured the runoff seasonality, but in general overestimated runoff, especially in the west portion of the basin. The FIFE Prairie Site (39.0ºN 96.5ºW) 5 Performance testing of Upgraded CLM The Arkansas-Red basin VIC simulated all four energy fluxes quite well, while CLM simulated net radiation, sensible and ground heat flux reasonably well but underestimated the latent heat due to the larger runoff estimation. VIC simulated soil moisture contents reasonably well both at surface and subsurface layer, while CLM largely undersimulated the soil moisture content at subsurface layer due to larger runoff and hence less infiltration. The canopy interception for CLM is similar to the one for VIC. The Upgraded CLM performed significantly better than the original CLM in terms of runoff prediction. The Upgraded CLM better captured the runoff peak, which, in general, was overestimated by the original CLM. The Upgraded CLM performed consistently well in both dry and wet areas of the basin, while the original CLM considerably overestimated the runoff in dry portion of the basin. Summary of the CLM model evaluation From our study: CLM tends to overestimate runoff peak, particularly in the relatively dry regions. (2) CLM tends to melt snow earlier than observation. CLM poorly simulates soil moisture profile due to poor soil moisture plant-water-relation function. CLM sometimes may underestimate evapotranspiration due to the poor parameterization for the soil-water-plant relationships. (5) CLM has similar canopy interception to the one of VIC (in terms of 1-hour time step). In general, VIC performs markedly better than CLM in terms of hydrologic predictions, including runoff, soil moisture and snow pack. Improvements of CLM are expected by incorporating some aspects of VIC hydrologic parameterizations into CLM. From other studies: (1) In coupled mode, CLM significantly underestimated runoff over central U.S., while overestimated runoff globally compared to observation [Bonan et al., 2002]. (2) The peak snow mass and thus the spring snowmelt of CLM lag observation by one month over Eurasia [Zeng et al., 2002]. Canopy interception of CLM is largely overestimated (we did not find this in our point sites as compared with VIC) [Lawrence et al. [2005]. This will be corrected by NCAR in the coming new version. The evapotranspiration partitioning is unrealistic when CLM is coupled to Community Atmosphere Model [Lawrence et al., 2005]. This unrealistic portioning of evapotranspiration and low precipitation contribute to dry soils that create a prominent warm summer temperature bias [Lawrence et al., 2005]. This will be fixed by NCAR in the coming new version. (5) Soil moisture content is poorly simulated by CLM due to the poor parameterization of soil-plant-water relationships (this is similar to our findings). This will be fixed by NCAR in the coming new version. (6) The study of muti-model ensemble by the University of Washington Land Surface Hydrology Group, has also found that CLM tends to melt snow earlier than observation and other land surface models in Arctic region. The ABRACOS Forest Site (10.1ºS 61.9ºW) The Upgraded CLM performed similarly to VIC in terms of surface runoff prediction, while the difference in the baseflow prediction are due to the different parameterization for the rooting depth by the Upgraded CLM (3.43m) and VIC (10m). The Upgraded CLM significantly improved the soil moisture, latent heat and evaportranspration prediction. Further improvement can be expected if the parameterizations are improved for soil-water-plant relationships and the rooting depth. The ABRACOS Forest Site (10.1ºS 61.9ºW) The Torne-Kalix basin CLM in general overestimated the sensible heat and underestimated the latent heat. The poor performance of CLM are attributed to the larger estimation of runoff and poor simulation of soil moisture content. The soil moisture content was very well estimated by VIC-10m (10 m rooting depth but poorly estimated by VIC-3.43m (3.43 m rooting depth), indicating that the 3.43m by CLM is too shallow, and 10m is approprirate. The evapotranspiration by CLM is extremely low (close to 0) in dry seasons, while the corresponding soil moisture contents are much higher than those simulated by VIC-3.43 in dry seasons. This indicates that soil-plant-water relationships and thus soil moisture content are poorly represented in CLM. The canopy interception for CLM is similar to the one for VIC. The VIC better captured the runoff magnitude and seasonality. CLM overestimated the runoff peak value for most subbasins except Paktfors, Pajala Rumphus and Pallo, where runoff peak was largely underestimated. The runoff peak of CLM, in general, came earlier than VIC and observation, implying that the CLM snow melt earlier than observation. 5 Conclusions 4 Enhancement of CLM soil hydrology The Colorado basin CLM tends to overestimate runoff, in particular in relatively dry areas. CLM tends to melt snow earlier, and thus the snow-melting dominated runoff peak comes one month earlier than observation. CLM simulates soil moisture contents unrealistically due to poor parameterization of soil-plant –water relationships. The poor simulation of the latent heat by CLM is due to the poor runoff parameterization. The Upgraded CLM, into which VIC runoff parameterization is incorporated, performs significantly better than original CLM. The Upgraded CLM requires only 5 VIC parameters, which are transferable to CLM without massive calibration although some systematical adjustment may be required for some basins. The enhancement of the CLM snow model remains to be done. Our studies have shown that VIC performs significantly better than CLM in terms of hydrologic predictions. As shown in the below diagram, we have upgraded CLM by incorporating the 3-layer VIC surface runoff and baseflow schemes into the 10-layer CLM. The procedures of runoff and baseflow calculation in the Upgraded CLM are as follows: In the time order during a model time step: Surface runoff is calculated based on the 3-layer VIC scheme; Infiltration is based on 10-layers (CLM layer scheme) using information for VIC layers interpolated to specified CLM layers; Soil evaporation, root-water-uptake and water redistribution (upward and downward), soil thermal states and fluxes are based on 10 CLM methodology layers; The Upgraded CLM, in general, performed markedly better than the original CLM in the runoff prediction. The Upgraded CLM largely corrected the runoff overestimation by the original CLM in the summer period. The Upgraded CLM did not largely reduce the runoff peak overestimation in the spring time by the original CLM, where the runoff peak is dominated by snow melting. Further improvements for the spring runoff prediction are expected when the CLM snow model is enhanced. The runoff seasonality is also expected to be improved after the CLM snow model is improved.