{ "cells": [ { "cell_type": "markdown", "id": "0f872ec4", "metadata": {}, "source": [ "# Statistical comparison global gridded climate datasets and their influence on LPJ-GUESS model outputs\n", "Geocomputation course VT2021, Individual project\n", "\n", "Alexandra Pongracz \n", "\n", "alexandra.pongracz@nateko.lu.se" ] }, { "cell_type": "markdown", "id": "5d6cb96c", "metadata": {}, "source": [ "## Background & Introduction\n", "Gridded climate data products are imperative resources for global environmental research. These continuous datasets are prduced by compiling and interpolating observational datasets, reanalysis products or a hybrid product of these two approaches. Such climate datasets enable to investigate long term processes, causality and atmosphere-biosphere interactions on a global scale. They are widely used as forcing in models to investigate ecosystem response to changing climatic conditions (New et al. 2000). Nowadays there are several gridded climatic datasets available - for instance CEDA Archive, NOAA - and updated on a regular basis. \n", "\n", "Meteorological input variables have a large influence of simulated biogeochemical and hydrological model outputs, such as ecosystem productivity or the computed carbon balance (McGuire et al 2001, Wu et al. 2018). The choice of forcing climate dataset is a large source of uncertainty in ecosystem modelling (McGuire et al. 2018). Model intercomparison projects – such as the Coupled Model Intercomparison Project (CMIP) -,therefore, assign a common simulation protocol to avoid output bias due to the differences in the climatic inputs (Taylor 2012, CMIP5). Finding of state-of-the-art global modelling studies are presented in the IPCC (Intergovermental Panel on Climate Change) climate assessment reports and provide the basis for discussion on climate change adoptation and mitigation procedures. \n", "The chosen climate forcing does not only effect scientific research but has a direct impact on decision making. Assessing the the potential bias and incertainty in model outputs due to different climatic forcing is critical to evaluate.\n", "\n", "These issues concern my own PhD project where I investigate climate change impact on the Arctic carbon cycle using a dynamic vegetation model (LPJ-GUESS). Future climate change scenarios forecast largest changes in temperature and precipitation at northern high latitudes - the region that is already undergoing unprecedented environmental changes (Box et al.2019). It is important to get a more comprehensive understanding on the impact of changing environmental conditions on biogeochemical and hydrological cycles. As these cycles are strongly dependent on the climatic forcing, \n", "the choice of model climatic forcing needs to be carefully chosen. A number of studies evaluated the fit between observations and gridded climate datasets with the conclusion that depending on region and time period, there is a substantial bias between gridded climate data and observations (Mehran et al. 2014, Li et al. 2013). This bias may propagate to model uncertainty. Wu et al. (2018) suggested that the large climate bias induced uncertainties in modelled carbon balance estimated need to be further investigated and minimised. Similarily, the study of Ahlström et al. (2013) emphesised that the modelled carbon balance strongly depend on climatic forcing variables. It is useful to analyse the impact of different climatic forcing on the model simulations to (1) evaluate the outputs accordingly and to (2) keep the climate forcing derived differences in mind when comparing to modelling studies using different simulation set-ups. \n", "\n", "One of the most widely used global gridded climate datasets is the CRUNCEP (v7) product, that is combined from CRU TS 3.2 observational and NCEP reanalysis products (Viovy et al. 2016). CRUJRA (v.2.1) is a recently developed dataset, that is planned to replace CRUNCEP in the future (Harris et al. 2019). CRUJRA is constructed from the Japanese Reanalysis data (JRA) and CRU TS 4.03 datasets, and has a global coverage just as CRUNCEP. Currenly, LPJ-GUESS offline historical simulations are forced by CRUNCEP data, with plans to migrate to the CRUJRA dataset in the near future. Understanfing the differences between these two dataset would be valuable prior to this update.\n", "\n", "In this project, the focus is on investigating the spatio-temporal differences on current and future model climate forcing datasets and their impact on modelled variables. This project is purely geocomputational, as my aim is to analyse two well known gridded climate datasets and their impact on LPJ-GUESS model outputs. Details on the two studied datasets are shown in Table 1.\n", "\n", "Table 1. Description of studied global gridded climate datasets.\n", "\n", " - | CRUNCEP v.7 | CRU JRA 2.1 \n", "--|:---------:|:-----------:\n", "data coverage |1901-2015 | 1901-2019 \n", "spatial resolution | 0.5° x 0.5°| 0.5° x 0.5°\n", "temporal resolution | daily | daily\n", "reference| Viovy et al. 2016 | Harris et al. 2019\n", "available variables| air T, precipitation, humidity,| incoming solar radiation, surface winds, pressure" ] }, { "cell_type": "markdown", "id": "2351ca5c", "metadata": {}, "source": [ "The objectives of this project are:\n", "