Bias correction in Statistical Downscaling

Recent perceived climate variability raises concerns with unprecedented hydrological phenomena and extremes. Distribution and circulation of the waters of the Earth become increasingly difficult to determine because of additional uncertainty related to anthropogenic emissions. The world wide observed changes in the large-scale hydrological cycle have been related to an increase in the observed temperature over several decades. Although the effect of change in climate on hydrology provides a general picture of possible hydrological global change, new tools and frameworks for modelling hydrological series with nonstationary characteristics at finer scales, are required for assessing climate change impacts. Of the downscaling techniques, dynamic downscaling is usually based on the use of Regional Climate Models (RCMs), which generate finer resolution output based on atmospheric physics over a region using General Circulation Model (GCM) fields as boundary conditions. However, RCMs are not expected to capture the observed spatial precipitation extremes at a fine cell scale or at a basin scale. Statistical downscaling derives a statistical or empirical relationship between the variables simulated by the GCMs, called predictors, and station-scale hydrologic variables, called predictands. The main focus of the paper is on the need for using statistical downscaling techniques for projection of local hydrometeorological variables under climate change scenarios. The projections can be then served as a means of input source to various hydrologic models to obtain streamflow, evapotranspiration, soil moisture and other hydrological variables of interest.

Hydrological implications of global climate change are usually assessed by downscaling appropriate predictors simulated by General Circulation Models (GCMs). Results from GCM simulations are subject to a number of uncertainties due to incomplete knowledge about the underlying geophysical processes of global change (GCM uncertainties) and uncertain future scenarios (scenario uncertainties). Disagreement between projections of regional climate change suggests that reliance on a single GCM with a few selected scenarios could lead to inappropriate planning and adaptation responses. This paper summarizes recent published work by the authors. The following methods and tools for statistical downscaling are discussed: (a) Fuzzy Clustering, (b) Relevance Vector Machine (RVM) and (c) Conditional Random Fields (CRFs). Uncertainty modelling with non-parametric methods and possibility theory are discussed. Applications of the methodologies are demonstrated by projection of the meteorological drought in the Orissa subdivision, India, and by predictions of the inflow to Hirakud Dam, Mahanadi River basin in India.

Water Resources Research, 2010

1] Regional impacts of climate change remain subject to large uncertainties accumulating from various sources, including those due to choice of general circulation models (GCMs), scenarios, and downscaling methods. Objective constraints to reduce the uncertainty in regional predictions have proven elusive. In most studies to date the nature of the downscaling relationship (DSR) used for such regional predictions has been assumed to remain unchanged in a future climate. However, studies have shown that climate change may manifest in terms of changes in frequencies of occurrence of the leading modes of variability, and hence, stationarity of DSRs is not really a valid assumption in regional climate impact assessment. This work presents an uncertainty modeling framework where, in addition to GCM and scenario uncertainty, uncertainty in the nature of the DSR is explored by linking downscaling with changes in frequencies of such modes of natural variability. Future projections of the regional hydrologic variable obtained by training a conditional random field (CRF) model on each natural cluster are combined using the weighted Dempster-Shafer (D-S) theory of evidence combination. Each projection is weighted with the future projected frequency of occurrence of that cluster ("cluster linking") and scaled by the GCM performance with respect to the associated cluster for the present period ("frequency scaling"). The D-S theory was chosen for its ability to express beliefs in some hypotheses, describe uncertainty and ignorance in the system, and give a quantitative measurement of belief and plausibility in results. The methodology is tested for predicting monsoon streamflow of the Mahanadi River at Hirakud Reservoir in Orissa, India. The results show an increasing probability of extreme, severe, and moderate droughts due to climate change. Significantly improved agreement between GCM predictions owing to cluster linking and frequency scaling is seen, suggesting that by linking regional impacts to natural regime frequencies, uncertainty in regional predictions can be realistically quantified. Additionally, by using a measure of GCM performance in simulating natural regimes, this uncertainty can be effectively constrained.

SN Applied Sciences

Water resources are naturally influenced by weather, topography, geology and environment. These factors cause difficulties in evaluating future water resources under changing climate. The Intergovernmental Panel on Climate Change (IPCC) reported that the increasing greenhouse gases which can cause sea level rise increase frequency of storms, heavy rainfall events and droughts. In order to quantify the future change of the hydrological cycle rainfall, different modeling approaches like the use of global climate models, regional climate models and hydrological models (for local scale about 200 km 2) are being used. This paper aims to investigate the challenges in modeling because of different parameter or variable or model conceptualization uncertainties using different scenarios, downscaling methods and climatic variables. This work helps to quantify future water resources to maintain quality of water to support aquatic life, agriculture and industrial needs. Challenges in climate change impact analysis in the present research and knowledge gaps are also discussed.

Climate change Daily rainfall Statistical downscaling Multiple GCMs Malaprabha catchment Modified Markov model s u m m a r y Impact of global warming on daily rainfall is examined using atmospheric variables from five General Circulation Models (GCMs) and a stochastic downscaling model. Daily rainfall at eleven raingauges over Malaprabha catchment of India and National Center for Environmental Prediction (NCEP) reanalysis data at grid points over the catchment for a continuous time period 1971-2000 (current climate) are used to calibrate the downscaling model. The downscaled rainfall simulations obtained using GCM atmospheric variables corresponding to the IPCC-SRES (Intergovernmental Panel for Climate Change -Special Report on Emission Scenarios) A2 emission scenario for the same period are used to validate the results. Following this, future downscaled rainfall projections are constructed and examined for two 20 year time slices viz. 2055 (i.e. 2046-2065) and 2090 (i.e. 2081-2100). The model results show reasonable skill in simulating the rainfall over the study region for the current climate. The downscaled rainfall projections indicate no significant changes in the rainfall regime in this catchment in the future. More specifically, 2% decrease by 2055 and 5% decrease by 2090 in monsoon (JJAS) rainfall compared to the current climate under global warming conditions are noticed. Also, pre-monsoon (JFMAM) and postmonsoon (OND) rainfall is projected to increase respectively, by 2% in 2055 and 6% in 2090 and, 2% in 2055 and 12% in 2090, over the region. On annual basis slight decreases of 1% and 2% are noted for 2055 and 2090, respectively. Crown

Adjustment of modeled values to reflect the observed distribution and statistics. Change factor (CF): Ratio between values of current climate and future GCM simulations. Climatology: Long-term average of a given variable, often over time periods of 20 to 30 years. For example, a monthly climatology consists of a mean value for each month computed over 30 years, and a daily climatology consists of a mean value for each day. Coastal breeze: Wind in coastal areas driven by differences in the rate of cooling/warming of land and water. Convective precipitation: Intense precipitation of short duration that characterizes most of the rainfall in the tropics. Direct and indirect effect of aerosols: Atmospheric aerosols are solid and liquid particles suspended in air that influence the amount of solar radiation that reaches the surface of the Earth. Aerosols can cool the surface of the Earth via reflection of solar radiation. This is termed the direct effect. The effect of aerosols on the radiative properties of Earth's cloud cover is referred to as the indirect effect. Downscaling: Derivation of local to regional-scale (10-100 kilometers) information from larger scale modeled or observed data. There are two main approaches: dynamical downscaling and statistical downscaling. Emissions Scenario: Estimates of future greenhouse gas emissions released into the atmosphere. Such estimates are based on possible projections of economic and population growth and technological development, as well as physical processes within the climate system. Feedback (climate): An interaction within the climate system in which the result of an initial process triggers changes in a second process that in turn influences the initial one. A positive feedback intensifies the original process and a negative one reduces it. Frequency distribution: An arrangement of statistical data that shows the frequency of the occurrence of different values. General Circulation Model (GCM): A global, three-dimensional computer model of the climate system that can be used to simulate human-induced climate change. GCMs represent the effects of such factors as reflective and absorptive properties of atmospheric water vapor, greenhouse gas concentrations, clouds, annual and daily solar heating, ocean temperatures, and ice boundaries. Grid cell: A rectangular area that represents a portion of the Earth's surface. Interannual variability: Year-to year change in the mean state of the climate that is caused by a variety of factors and interactions within the climate system. One important example of interannual variability is the quasi-periodic change of atmospheric and oceanic circulation patterns in the Tropical Pacific region, collectively known as El Niño-Southern Oscillation (ENSO). A Review of Downscaling Methods for Climate Change Projections vi Interpolation: The process of estimating unknown data values that lie between known values. Various interpolation techniques exist. One of the simplest is linear interpolation, which assumes a constant rate of change between two points. Unknown values anywhere between these two points can then be assigned. Land-sea contrast: Difference in pressure and other atmospheric characteristics that arises between the land and ocean, caused by the difference in the rate of cooling/warming of their respective surfaces. Large-scale climate information: Atmospheric characteristics (e.g., temperature, precipitation, relative humidity) spanning several hundred to several thousand kilometers. Lateral boundaries: Information about the air masses, obtained from GCM output or observations, used by RCMs to derive fine-scale information. Markovian process: When values of the future depend solely on the present state of the system and not the past. Predictand: The variable that is to be predicted. In downscaling, the predictand is the local climate variable of interest. Predictor: A variable that can be used to predict the value of another variable. In downscaling, the predictor is the large-scale climate variable. Regional Climate Model (RCM): High-resolution (typically 50 kilometers) computer model that represents local features. It is constructed for limited areas, run for periods of ~20 years, and driven by large-scale data. Spatial downscaling: Refers to the methods used to derive climate information at finer spatial resolution from coarser spatial resolution GCM output. The fundamental basis of spatial downscaling is the assumption that significant relationships exist between local and large-scale climate. Spatial resolution: In climate, spatial resolution refers to the size of a grid cell in which 10-80 kilometers and 200-500 kilometers are considered to be "fine" and "coarse," respectively. Stationarity: Primary assumption of statistical downscaling; as the climate changes, the statistical relationships do not. It assumes that the statistical distribution associated with each climate variable will not change, that the same large-scale predictors will be identified as important, and that the same statistical relationships between predictors and predictands exist. Stochastic: Describes a process or simulation in which there is some indeterminacy. Even if the starting point is known, there are several directions in which the process can evolve, each with a distinct probability. Synoptic: Refers to large-scale atmospheric characteristics spanning several hundred to several thousand kilometers. Systematic bias: The difference between the observed data and modeled results that occurs due model imperfections. Temporal downscaling: Refers to the derivation of fine scale temporal data from coarser-scale temporal information (e.g., daily data from monthly or seasonal information). Its main application is in impact studies when impact models require daily or even more frequent information. Temporal resolution: The time scale at which a measurement is taken or a value is represented. Daily and monthly resolutions denote one value per day and one value per month, respectively.

Progress in Physical Geography, 1999

The scientific literature of the past decade contains a large number of reports detailing the development of downscaling methods and the use of hydrologic models to assess the potential effects of climate change on a variety of water resource issues. This article reviews the current state of methodologies for simulating hydrological responses to global climate change. Emphasis is given to recent advances in climatic downscaling and the problems related to the practical application of appropriate models in impact studies. Following a discussion of the advantages and deficiencies of the various approaches, challenges for the future study of the hydrological impacts of climate change are identified.

J. Climate 12: 258-272 , 1999

Empirical downscaling procedures relate large-scale atmospheric features with local features such as station rainfall in order to facilitate local scenarios of climate change. The purpose of the present paper is twofold: first, a downscaling technique is used as a diagnostic tool to verify the performance of climate models on the regional scale; second, a technique is proposed for verifying the validity of empirical downscaling procedures in climate change applications. The case considered is regional seasonal precipitation in Romania. The downscaling model is a regression based on canonical correlation analysis between observed station precipitation and European-scale sea level pressure (SLP). The climate models considered here are the T21 and T42 versions of the Hamburg ECHAM3 atmospheric GCM run in ‘‘time-slice’’ mode. The climate change scenario refers to the expected time of doubled carbon dioxide concentrations around the year 2050. The downscaling model is skillful for all seasons except spring. The general features of the large-scale SLP variability are reproduced fairly well by both GCMs in all seasons. The climate models reproduce the empirically determined precipitation–SLP link in winter, whereas the observed link is only partially captured for the other seasons. Thus, these models may be considered skillful with respect to regional precipitation during winter, and partially during the other seasons. Generally, applications of statistical downscaling to climate change scenarios have been based on the assumption that the empirical link between the large-scale and regional parameters remains valid under a changed climate. In this study, a rationale is proposed for this assumption by showing the consistency of the 2 3 CO2 GCM scenarios in winter, derived directly from the gridpoint data, with the regional scenarios obtained through empirical downscaling. Since the skill of the GCMs in regional terms is already established, it is concluded that the downscaling technique is adequate for describing climatically changing regional and local conditions, at least for precipitation in Romania during winter.

Geophysical Research Letters, 2000

Daily rainfall and surface temperature series were simulated for the Animas River basin, Colorado using dynamically and statistically downscaled output from the National Center for Environmental Prediction/ National Center for Atmospheric Research (NCEP/NCAR) re-analysis. A distributed hydrological model was then applied to the downscaled data. Relative to raw NCEP output, downscaled climate variables provided more realistic simulations of basin scale hydrology. However, the results highlight the sensitivity of modeled processes to the choice of downscaling technique, and point to the need for caution when interpreting future hydrological scenarios.