A região do Vale do Paraíba, situada em uma planície cortada pelo Rio Paraíba do Sul, entre as Serras da Mantiqueira e do Mar, possui alternância entre períodos secos e chuvosos, alcançando cerca de 1300 mm por ano. Todas estas características físicas somadas à disposição de moradias em várzeas e áreas com alta declividade contribuem para a ocorrência de desastres naturais. Deste modo, foi realizado o levantamento da vulnerabilidade climática aos desastres naturais da região do Vale do Paraíba. A pesquisa foi dividida em duas partes, passado e futuro. No passado (1990-2008) a região apresentou municípios vulneráveis aos desastres naturais como São José dos Campos, Jacareí, Campos dos Jordão, Taubaté e Aparecida. Os desastres naturais de maior ocorrência foram às inundações (54%) e as tempestades severas (25%) com maior frequência nos meses de janeiro, fevereiro e março. Para o futuro foram analisados dados de precipitação (2070-2100) para os cenários A2 e B2, a partir dos dados de simulação climática futura, modelo ETA/CCS, pela técnica de downscaling dinâmico, o modelo apontou para redução da precipitação na região, sendo de 44% para o cenário A2 e 35% para o cenário B2. Além da redução no total de precipitação, os dados futuros apontam para o aumento do período de dias de permanência de chuva, com predomínio de chuvas leves (0,1 a 5 mm), ou seja, haverá redução nos eventos extremos de precipitação, o que contribuiria para a diminuição de processos geradores de desastres naturais na região do Vale do Paraíba. / The region of Paraíba Valley, situated on a plain crossed by Paraiba do Sul River, between Mantiqueira and Mar mountain rigdes, alternates dry and wet periods, getting about 1300 mm of rain per year. The physical characteristics combined with the location of homes in low and flat lands alongside a watercourse and in areas with steep slopes contribute to the occurrence of natural disasters. This study of climate vulnerability and natural disasters in the region of Paraíba Valley. This research is divided into two parts: past and future. In the past (1990-2008) the region has vulnerable cities to natural disasters. Such as São José dos Campos, Jacareí, Campos do Jordão, Taubaté, and Aparecida. In these cities, the predominant natural disasters were floods (54%) and severe storms (25%) that occur frequently in the months of January, February and March. For the future precipitation data modeled (2070-2100) were analyzed for the scenarios A2 and B2 of IPCC, from the data of future climate simulation, (ETA / CCS model) by applying the dynamic downscaling technique. The model indicates reduced precipitation in the region (44% for A2 scenario and 35% for scenario B2). Besides the reduction in total precipitation, the future data point to the increase in the number of rainy days with the predominancy of ligth rains (0.1 to 5 mm), so, it will have a reduction in extreme precipitation events that could contribute to a decrease of natural disaster generating processes in the region of Paraíba Valley.

4 Abstract Determination of wind direction and speed from free atmosphere data. This work deals with relationship between surface wind and large-scale predictors in the Czech Republic. Approach is statistical downscaling based on multiple linear regression with stepwise screening. Tested predictors include four geopotential heights (hgt 500, hgt 700, hgt 850 and hgt 1000), sea level pressure and surface wind components U and V. Observed surface wind speed and azimuth from 18 meteorological station ČHMÚ is used as predictand. Period of interest is define by used data on term 1961 - 2010. In three different methods are compared straight downscaling of wind components U and V, used wind speed calculate from downscaled components U and V, and straight downscaling of wind speed without distinguish wind direction. Resulting models are tested by cross-validation method. The resulting equation allows good simulation of surface wind at most of station. The better method to downscaling surface wind speed is prescribe method of calculate wind speed from downscaled components U and V.

The profitability from the production wind power energy is related to the quality of the wind speed forecasts. All wind predicting methods needs meteorological data, for the prevailing synoptic situation, as input. High quality input is wanted for a better result. In this study a new idea of a method for estimation of high resolution wind fields is examined. The idea is to use an existing database, containing simulations of high resolution wind fields, to estimate the actual wind by combining the simulations in a way fitting actual synoptic data. The simulations in the database have been produced by the mesoscale MIUU-model, which has been developed by Leif Enger at Uppsala University. The database contains simulations characterized by different geostrophic wind speeds and directions. There is also a separation into four seasons, where values which are typical for each season is put on meteorological parameters. Reanalysis data from NCEP-NCAR, containing 850 hPa geopotential heights describing actual synoptic situations, is used to calculate geostrophic wind speeds and directions. Three different geostrophic wind calculation methods, the triangle method, the small cross-method and the large cross-method, are tested. The calculated geostrophic wind is compared between the methods. The small cross-method is chosen and the main reason for that is the large amount of reanalysis information considered by this method and the use of a small calculation area. Measurements of the wind speed and direction are available from the tower at Utgrunden. The geostrophic wind speeds and directions are therefore calculated especially for the position of Utgrunden. This is done by a linear weighting of data, from several grid points close to Utgrunden, with respect to the distance to Utgrunden. Linear weighting is also used when estimating the wind speed for Utgrunden. The wind speed is estimated by weighting together MIUU-model simulations, for different geostrophic wind speeds and directions, so that they fit the geostrophic wind values calculated for Utgrunden. The calculated wind speed, measured wind speed and calculated geostrophic wind speed, for Utgrunden, are compared. The correspondence, between the calculated and measured wind speed, turns out to be quite good for many time periods. The diurnal variations in the measured wind speed are partly captured by calculated wind speed, but the diurnal variations tend to be larger in the measured wind speed then in the calculated. There are also cases where there are large differences between the measured and estimated wind speed. Many of these cases are probably cased by unusual weather situations. By considering additional parameters, as the temperature field, it is likely that these wind estimations can be improved. With more research it may be possible to produce high resolution wind fields with enough accuracy to be useful as inputs in wind prognostic systems. The advantage with such a method would be that accurate high resolution wind fields could be produced without the use of a time consuming numerical high resolution model. / Lönsamheten för produktion av vindkraft elektricitet bestäms delvis av förmågan att göra bra vindprognoser för nästkommande dygn. Alla metoder för vindprognostisering behöver meteorologisk indata som beskriver den rådande synoptiska situationen. Kvaliteten och upplösningen på dessa indata har stor betydelse för metodens resultat. I denna studie undersöks en alternativ metod för bestämning av högupplösta vind fält. Idén är att man ska försöka utnyttja en tillgänglig databas av högupplösta vindfält, producerade av den mesoskaliga MIUU – modellen som är utvecklad av Leif Enger på meteorologiska institutionen vid Uppsala Universitet. Tanken är att dessa vindfält ska kunna kombineras på ett sådant sätt att de överensstämmer med en given synoptisk situation. MIUU – modell körningarna, i databasen, är indelade i situationer karaktäriserade av olika värden på den geostrofiska vindstyrkan och vindriktningen. Körningarna är gjorda för fyra säsonger, för vilka typiska värden för säsongen är satta på styrande parametrar. För att kunna kombinera MIUU - modell körningarna beräknas den geostrofiska vinden från 850 hPa geopotential höjd återanalys data tillgänglig från NCEP-NCAR. Tre olika beräkningsmetoder för geostrofisk vind testas och jämförs. Den ”lilla korsmetoden” väljs för uppgiften beroende på att den utnyttjar en förhållandevis stor mängd återanalys data, för beräkning av geostrofisk vind, samt använder litet beräkningsområde. Automatiskt uppmätta värden över vindhastighet och vindriktning finns tillgängliga från en mast positionerad vid Utgrunden i Kalmar sund. Den geostrofiska vinden beräknas därför i Utgrundens position. Beräkningen utförs genom linjär viktning av data från de från Utgrunden sett fem närmaste gridpunkterna (i lilla korsmetodens gridfält). En linjär viktning används sedan även för att vikta ihop de MIUU – modell simulerade vindfälten så att de passar de beräknade värdena på geostrofisk vindhastighet och vindriktning. Jämförelser görs mellan den beräknade vinden, den uppmätta vinden samt den geostrofiska vinden, för Utgrunden. Korrelationen, mellan uppmätt och beräknad vind, visar sig vara ganska god periodvis. Den dagliga variationen i den uppmätta vindhastigheten fångas delvis av beräkningsmetoden, men dygnsvariationen är betydligt större i den uppmätta vinden än i den beräknade. Det noteras även att det finns situationer då det är stora skillnader mellan beräknad och uppmätt vind. Dessa situationer beror i många fall troligen på onormala vädersituationer. Studium av ytterliggare parametrar, som t.ex. temperaturfältet, skulle troligen leda till betydande förbättringar i vinduppskattningen. Ytterligare forskning och förbättring av metoden skulle kunna leda till produktion av högupplösta vindfält med tillräcklig kvalitet för användning i vindprognostiseringsmodeller. Fördelen skulle i så fall vara möjligheten att kunna producera högupplösta vindfält utan användning av tidskrävande numerisk modeller.

abstract: Soil moisture (θ) is a fundamental variable controlling the exchange of water and energy at the land surface. As a result, the characterization of the statistical properties of θ across multiple scales is essential for many applications including flood prediction, drought monitoring, and weather forecasting. Empirical evidences have demonstrated the existence of emergent relationships and scale invariance properties in θ fields collected from the ground and airborne sensors during intensive field campaigns, mostly in natural landscapes. This dissertation advances the characterization of these relations and statistical properties of θ by (1) analyzing the role of irrigation, and (2) investigating how these properties change in time and across different landscape conditions through θ outputs of a distributed hydrologic model. First, θ observations from two field campaigns in Australia are used to explore how the presence of irrigated fields modifies the spatial distribution of θ and the associated scale invariance properties. Results reveal that the impact of irrigation is larger in drier regions or conditions, where irrigation creates a drastic contrast with the surrounding areas. Second, a physically-based distributed hydrologic model is applied in a regional basin in northern Mexico to generate hyperresolution θ fields, which are useful to conduct analyses in regions and times where θ has not been monitored. For this aim, strategies are proposed to address data, model validation, and computational challenges associated with hyperresolution hydrologic simulations. Third, analyses are carried out to investigate whether the hyperresolution simulated θ fields reproduce the statistical and scaling properties observed from the ground or remote sensors. Results confirm that (i) the relations between spatial mean and standard deviation of θ derived from the model outputs are very similar to those observed in other areas, and (ii) simulated θ fields exhibit the scale invariance properties that are consistent with those analyzed from aircraft-derived estimates. The simulated θ fields are then used to explore the influence of physical controls on the statistical properties, finding that soil properties significantly affect spatial variability and multifractality. The knowledge acquired through this dissertation provides insights on θ statistical properties in regions and landscape conditions that were never investigated before; supports the refinement of the calibration of multifractal downscaling models; and contributes to the improvement of hyperresolution hydrologic modeling. / Dissertation/Thesis / Doctoral Dissertation Civil, Environmental and Sustainable Engineering 2018

Impacts of Climate-responsive Controls and Land Usage on Regional Climate and Air Quality Marcus A. Trail 201 pages Directed by Dr. Armistead G. Russell Regional air quality impacts public health, visibility and ecosystem health, and is significantly affected by changes in climate, land use and pollutant emissions. Predictions of regional air quality responses to such changes can help inform policy makers in the development of effective approaches to both reduce greenhouse gases and improve air quality. However, major sources of uncertainty exist in predicting future air quality including limitations in the tools used to project future emissions, land use changes and uncertainties associated with predicting future climate. Recently, technical advances in downscaling global climate simulations to regional scales, and, the development of bottom-up operational tools used to forecast emissions have enhanced our ability to account for the complex interactions between population, socio-economic development, technological change, and federal and regional environmental policies. The results show that emissions reductions strategies will continue to play a vital role in improving air quality over the U.S. while CO2 emission reduction policies can have mixed positive and negative impacts on air quality. However, additional costs will be necessary to reach air quality goals due to climate change because deeper emission reductions will be required to compensate for a warmer climate, even if current efforts are predicted to show improvement. The results of this study also show that regional climate and O3 and aerosol concentrations are highly sensitive to reforestation and cropland conversion in the Southeast and these land use changes should be considered in air quality management plans.

Climate change studies are crucial to assist decision-makers in understanding future risks and planning adequate adaptation measures. In general, Global/Regional Climate Models (GCMs/RCMs) achieve coarse resolutions, and are thus unable to provide sufficient information to conduct local climate assessments. Downscaling, defined as a method that derives local to regional-scale (10 to 100 km) information from larger-scale models or data analyses, is used to address this deficiency. In this thesis, a particular downscaling technique, known as the Quantile-Quantile transformation, was used to adjust the statistical distribution of RCM variables to match the statistical distribution of the observed variables generated by two RCMs: the Canadian Regional Climate Model version 3.7.1 and the ARPEGE model, on the historical period (1961-2001). The analyses presented in this study were applied to daily precipitation and maximum and minimum temperatures in the South Nation watershed in Eastern Ontario, Canada. The two-sample Kolmogorov–Smirnov test indicated that the Quantile-Quantile transformation improved the shape of the PDF of RCM-simulated climate variables. The results suggest that, under the A1B scenario, temperatures in the watershed would rise significantly and there would be an increment in precipitation occurrence and intensity. Trend analysis was performed on the 1961 to 2001 and 2041 to 2081 timeframes, using the Mann-Kendall test and the Sen's slope estimator. Discernible, often significant, increases of maximum and minimum temperatures were found for the 1961 to 2001 period, and stronger ascending slopes for the 2041 to 2081 period. However, there was marginal evidence of changes in the time series of maximum and accumulated annual precipitation for both periods. The study also outlined how the frequency and intensity of some extreme weather events will evolve in the 2041-2081 period in response to the rise in atmospheric GHG concentrations. Projected impacts were investigated by tracking future changes in four extreme temperature indices and three precipitation indices. It was predicted that heavy precipitation events and warm spells will occur more frequently and intensely, while extreme cold events will be weaker, and some will be hardly observed.

The knowledge of flow hydrodynamics within the next decades is of particular importance in many practical applications. In this study, a methodological improvement has been made to the evaluation of the flow hydrodynamics under climate change. This research, indeed, proposes an approach which includes the methods that can consider the climate change impact on the flow in estuaries, gulfs, etc. It includes downscaling methods to project the required climate variables through the next decades. Here, two statistical downscaling methods, namely, Nearest Neighbouring and Quantile-Quantile techniques, are developed and implemented in order to predict the wind speed over the study area. Wind speed has an essential role in flow field and wave climatology in estuaries and gulfs. In order to make the proposed methodology computationally efficient, the flow in the estuary is simulated by a large-scale model. The finite volume triangular C-grid is analysed and shown to have advantages over the rectangular (finite difference) one. The dispersion relation analysis is performed for both gravity and Rossby waves that have crucial effects in oceanic models. In order to study the unstructured characteristic of the triangular grids, various isosceles triangles with different vertex angles are considered. Moreover, diverse well-known second-order time stepping techniques such as Leap-Frog, Adams-Bashforth and improved Euler are studied in combination with the C-grid semi discrete method. The fully discrete method is examined through several numerical experiments for both linear and non-linear cases. The results of the large-scale model provide the boundary conditions to the local coastal model. In order to model the flow over a local coastal area, a well-balanced positivity preserving central-upwind method is developed for the unstructured quadrilateral grids. The quadrilateral grid can effectively simulate complex domains and is shown to have advantages over the triangular grids. The proposed central-upwind scheme is well-balanced and preserve the positivity. Therefore, it is capable of modelling the wetting and drying processes that may be the case in many local coastal areas. It is also confirmed that the proposed method can well resolve complex flow features. The local model incorporates the outputs of the downscaling and large-scale flow models and evaluates the flow hydrodynamics under changing climate.

An ensemble of 11 dynamically downscaled CMIP3 GCMs under A2 projection scenario are first bias corrected for the historic (1971-2000) and scenario (2041-2070) period using a Scaled Distribution Mapping (SDM) technique, that preserves the relative change in the monthly mean and variance of precipitation and any model trends in temperature to generate an ensemble of streamflow projections across 3 catchments in the Colorado River basin - Upper Colorado at Lees Ferry, Salt and Verde. The hydroclimatic projections obtained from this method are compared against an existing ensemble of 15 Bias Corrected and Spatially Disaggregated (BCSD) CMIP3 models under A2 projection scenario developed by the Bureau of Reclamation (BOR). The confidence in the DD Ens. stems from its ability to represent historical flow quantiles better than BCSD Ens. Across all three basins, the mean of the dynamically downscaled ensemble (DD Ens.) projects a decrease in both monsoon and winter projected precipitation as compared to mean of the statistically downscaled ensemble (BCSD Ens.). For the Upper Colorado, both Ens. show a shift in peak hydrograph from June to May due to earlier snowmelt, but a projected decrease in precipitation (-5%) by DD Ens. as compared to a slight increase (+2%) by BCSD Ens. results in a lower April snow water equivalent (SWE) and reduced streamflows (14% by DD Ens. as compared to 5% by BCSD Ens.). The streamflow decrease over the Upper Colorado River basin, quantified by both the mean and the spread of the ensemble. is representative in high flows and flows during moist conditions. For smaller basins like Salt and Verde, DD Ens. shows a greater decrease (-11%) in precipitation than BCSD Ens. (-2%), which results in lower peak hydrograph during March and significantly reduced streamflows (-20%&-14% for Salt and Verde by DD Ens. as compared to -3% by BCSD Ens.). This decrease is more substantial in high flows, but occurs across all streamflow quantiles. The future streamflow projection, quantified by the spread of the DD Ens. presents the shifting of the streamflow range downward to be drier in the future.

Increasing populations, shifting demographics and changes in diet are driving increases in crop production. However, any increases in food demand are ultimately limited by water availability, which is under pressure globally, but especially so in arid and semi-arid regions. To address this challenge, spatially distributed information on crop water use, vegetation health, soil moisture status and a range of other water, energy and carbon variables are all required. However, critical to the determination of many of these processes is an accurate characterization of the land surface temperature (LST). The only feasible manner by which to estimate this variable across a range of spatial and temporal scales is using thermal infrared (TIR) satellite data. Here we investigate the estimation of LST, focusing on its accurate retrieval across a range of different spatial scales. First, we examine the influence of atmospheric correction on retrieval accuracy by employing a radiative transfer model and Landsat data using a variety of available atmospheric profile data, with the aim of identifying an optimal product combination for retrieval. Using these results, we then investigate the potential to downscale coarse resolution (O~103 m) MODIS satellite data to scales appropriate for agricultural application (less than O~102 m), using a machine-learning approach. To further advance the downscaling technique, we explore the utility of novel Cubesat data to produce within-field scale (O~101 m) distributions of land surface temperature. Finally, to expand upon the multi-resolution/multi-satellite LST strategy explored here, we examine the capacity of ultra-high resolution (O~10-1 m) thermal imagery from an unmanned aerial vehicle to characterize surface temperature response and behavior, focusing on the retrieval accuracy and diurnal variability of these spatially and temporally varying land surface temperature estimates. The ultimate goal of this research is to advance the utility of LST for agricultural application by providing description and insights into product development, accuracy issues, and identifying some limitations and opportunities of both current and future remote observation platforms.

There is a strong interest in the climate community to improve the ability to accurately predict future trends of climate variables. Recently, machine learning methods have proven their ability to contribute to more accurate predictions of historical data on a variety of climate variables. There is also a strong interest in using statistical downscaling to predict local station data from the output of multi-model ensembles. This project looks at using the machine learning algorithm XGBoost and evaluating its ability to accurately predict historical monthly precipitation, with a focus of applying this method to simulate future precipitation trends. / Thesis / Master of Science (MSc)