GEOG 401 Climate Change

Transcription

1 GEOG 401 Climate Change Climate Downscaling GCMs have coarse resolu/on Spa<al resolu<on of global models con<nues to improve. But, they are s<ll not sufficiently resolved to accurately represent processes at regional and local scales. CMIP5 high-resolu/on models: 1 5 deg resolu/on 1

2 Downscaling Downscaling Strategies: Sta<s<cal Dynamical Ra/onale for downscaling in Hawai i: Complexi<es in rainfall genera<on processes and the resul<ng steep gradients cannot be represented in global models Oliver Elison Timm, Henry Diaz, Abby Frazier, Thomas Giambelluca Lauren Kaiser, Mami LeMaster Giambelluca, T.W., Q. Chen, A.G. Frazier, J.P. Price, Y.-L. Chen, P.-S. Chu, J.K. Eischeid, and D.M. Delparte, 2013: Online Rainfall Atlas of Hawai i. Bull. Amer. Meteor. Soc. 94, , doi: /BAMS-D Approaches to Downscaling Sta<s<cal downscaling: establish sta<s<cal rela<onships between weather pamerns in global model and varia<ons in weather at a point Dynamical downscaling: apply dynamical climate model at high spa<al resolu<on within a regional domain Intermediate complexity downscaling: subs<tute parameteriza<ons to represent some processes in dynamical model to reduce computer requirements 2

3 Sta<s<cal Downscaling Climate variability at a point is related to large-scale atmospheric pamerns of circula<on, moisture transport, and stability Global models are skillful at represen<ng the large scale pamerns By establishing sta<s<cal rela<onships between the circula<on/transport/stability pamerns and climate at a sta<on, projec<ons can be made of past or future climate varia<ons at the sta<on based on varia<ons in the pamerns. Assumes that the rela<onships between spa<al pamerns and climate at a point do not change as a result of climate change: sta<onarity assump<on Data Assimila<on Global climate models are not expected to reproduce actual sequences of hour to hour or day to day weather pamerns beyond forecast periods of a 1 to 2 weeks Beyond that, the models are expected to produce plausible sequences of weather pamerns with sta<s<cal proper<es similar to actual weather pamerns When global models are used for historical periods, they can be operated in the same way as model runs of the future, i.e., without benefit of observa<ons except for those used to specify the model ini<al condi<ons, or observa<ons can be systema<cally incorporated to keep the model on track to represent the actual sequence of weather pamerns: Data Assimila/on Reanalysis Data Sets Weather observa<ons are rela<vely sparse and irregularly located Many important variables, such as solar radia<on, are measured at only a few sta<ons Observa<ons at levels above the ground are available only at radiosonde sta<ons Reanalysis Data Sets are global gridded es<mates of past weather produced by global climate models constrained by assimila<ng observa<ons from ground sta<ons, radiosonde profiles, and satellite data Reanalysis provides spa<ally and temporally complete data sets of all weather variables for historical periods Reanalysis data sets are extremely valuable for numerous applica<ons including downscaling 3

4 Sta<s<cal Downscaling: Hawai i Example Elison Timm et al. (2015) Oliver Elison Timm, Henry Diaz, Abby Frazier, Thomas Giambelluca Lauren Kaiser, Mami LeMaster Elison Timm, O., Giambelluca, T.W. and Diaz, H.F Sta<s<cal downscaling of rainfall changes in Hawai i based on the CMIP5 global model projec<ons. Journal of Geophysical Research: Atmospheres 120: , doi: /2014JD Climate Change Circula<on pamern in the Pacific Sector around Hawai i Principal Component Analysis used to reduce complexity Dominant mode of variability in 32-member ensemble simula/on at RCP8.5 Geopoten/al height anomaly (pressure panern) at 500 hpa level 4

5 Climate Change Circula<on pamern in the Pacific Sector around Hawai i Dominant mode of variability in 32-member ensemble simula/on at RCP8.5 Moisture transport at 700 hpa level Climate Change Circula<on pamern in the Pacific Sector around Hawai i Dominant mode of variability in 32-member ensemble simula/on at RCP8.5 Temperature difference 1000hPa minus 500hPa Climate Change Circula<on pamern in the Pacific Sector around Hawai i Dominant mode of variability in 32-member ensemble simula/on at RCP8.5 height Temperature difference 1000hPa minus 500hPa Temp. 5

6 Climate Change Circula<on pamern in the Pacific Sector around Hawai i Dominant mode of variability in 32-member ensemble simula/on at RCP8.5 height Temperature difference 1000hPa minus 500hPa Temp. Transla<ng large-scale climate anomalies into rainfall es<mates. Timm and Diaz, J. Climate, 2009 Composite PaMern Geopoten<al Height 500hPa (1000hPa in coutours) (1) (2) (3) (4) 6

7 Composite PaMern 700hPa Moisture Transport (1) (2) (3) (4) Calibra<on Skill for Rainfall (Nov-Apr season) Oahu Correlation between observed & statistically downscaled rainfall Calibra<on Skill for Rainfall (Nov-Apr season) Correlation between observed & statistically downscaled rainfall 7

8 CMIP5 Future Circula<on Changes Projected onto the Composite PaMern (1) (2) (3) (4) Sta<s<cal downscaling results CMIP5 RCP yr average rainfall changes (Nov-Apr. season) Projected Change in Wet Season Rainfall Based on Sta<s<cal -158 Downscaling CMIP5 ECP8.5 ensemble median scenario for late average (Elison Timm et al. 2015). 8

9 Latest Projec<ons Elison Timm et al. (2015) produced maps of seasonalmean rainfall changes in Hawai i Wet-season results show higher skills than dry season Overall scenario for 21 st century: dry regions get drier, the wet regions remain wet or get wemer Dynamical Downscaling U<lizes the same type of numerical model used for global climate simula<ons and regional weather predic<on: a regional climate model. A regional domain is used allowing much higher spa<al resolu<on; nested domains with successively higher resolu<on ogen used. Psuedo Global Warming method is a common strategy; historical reanalysis is used to define the lateral boundary condi<ons; global warming increments used to modify condi<ons inside domain to represent future condi<ons. PGW approach assumes no change in climate variability. Computa<onally intensive, thus limi<ng number of test runs and global models used. Dynamical Downscaling Hawai i Example: and Zhang et al. (2016) Lauer, A., Zhang, C., Elison Timm, O., Wang, Y., and Hamilton, K Downscaling of climate change in the Hawaii region using CMIP5 results: on the choice of forcing fields. Journal of Climate 26: , doi: /JCLID s1. Zhang, C., Wang, Y., Hamilton, K., and Lauer, A., Dynamical downscaling of the climate for the Hawaiian Islands, Part II: Projec<on for the late twenty-first century. Journal of Climate, doi: /JCLI- D

10 SST Increment: RCP4.5 SST Increment: RCP8.5 Air Temperature Increment 10

11 4/17/18 Precipitable Water Vapor Increment Change in Mean Annual Precipita<on: RCP4.5 Change in Mean Annual Precipita<on: RCP8.5 11

12 4/17/18 Results Summary for RCP4.5 Results Summary for RCP8.5 Changes in TWI Height 12

13 4/17/18 Dynamical Downscaling Does the PGW approach adequately represent the important effects of global warming on regional and local precipita<on? What this shows is that the future projec<on based on the pseudo global warming approach is constrained by being <ed to the historical variability. The frequency of disturbances is determined by the historical data which are used to give the <medependent boundary condi<ons at the lateral boundaries of the model domain. Monthly mean rainfall data from the APDRC web page Chunxi Zhang (IPRC, UHM). One for present day, one for future. Averaged the monthly mean rainfall in the 3-km resolu<on data over the Hawai i region (160w-155W 18.5N-23.5N) Mean Annual Rainfall Change , RCP 8.5 Sta/s/cal Downscaling Elison Timm et al. (2015) 13

14 4/17/18 Rainfall Change , CMIP3, A1B Dynamical Downscaling Zhang et al. (2016) Characterizing Uncertainty in Downscaled Products Sources of uncertainty: Uncertainty in GCM simula<ons Uncertainty in GHG scenarios Added uncertainty in downscaling Characterizing uncertainty: Use many GCM simula<ons Use different RCPs Use different downscaling approaches Use different methods of es<ma<ng pamerns of present (historical) climate Characterizing uncertainty requires large number of downscaling simula<ons Easy to do with sta<s<cal downscaling Difficult to do with dynamical downscaling Need an intermediate method 14

15 Discrepancies Between Sta<s<cal and Dynamical Downscaling Results for Hawaii. Resource managers are frustrated Workshops held last September 2016 and April 2017 to help answer ques<ons Other alterna<ves being sought, e.g. addi<onal sta<s<cal and dynamical downscaling results and use of models of intermediate complexity AMS Mountain Net Conference presenta<on by Ethan Gutman: hmps://ams.confex.com/ams/16mountmet/webprogram/ Paper html 15