Search

Lakes are significant emitters of methane to the atmosphere, and thus are important components of the global methane budget. Methane is typically produced in lake sediments, with the rate of methane production being strongly temperature dependent. Local and regional studies highlight the risk of increasing methane production under future climate change, but a global estimate is not currently available. Here, we project changes in global lake bottom temperatures and sediment methane production rates from 1901 to 2099. By the end of the 21st century, lake bottom temperatures are projected to increase globally, by an average of 0.86-2.60 degrees C under Representative Concentration Pathways (RCPs) 2.6-8.5, with greater warming projected at lower latitudes. This future warming of bottom waters will likely result in an increase in methane production rates of 13%-40% by the end of the century, with many low-latitude lakes experiencing an increase of up to 17 times the historical (1970-1999) global average under RCP 8.5. The projected increase in methane production will likely lead to higher emissions from lakes, although the exact magnitude of the emission increase requires more detailed regional studies. Funding Agencies|Deutsche Forschungsgemeinschaft [AD 91/22-1]; European Regional Development Fund [FEDER--MCIU-AEI/CGL2017-86788-C3-2-P]; FP7 Ideas: European Research Council [336642]; H2020 European Research Council [725546]; Knut och Alice Wallenbergs Stiftelse [KAW 2018-0191]; Natural Environment Research Council

Bacheloroppgave i ergoterapi, Høgskulen på Vestlandet BER332

Lakes and reservoirs are important emitters of climate forcing trace gases. Various environmental drivers of the flux, such as temperature and wind speed, have been identified, but their relative importance remains poorly understood. Here we use an extensive field dataset to disentangle physical and biogeochemical controls on the turbulence-driven diffusive flux of methane (CH4) on daily to multi-year timescales. We compare 8 years of floating chamber fluxes from three small, shallow subarctic lakes (2010–2017, n = 1306) with fluxes computed using 9 years of surface water concentration measurements (2009–2017, n = 606) and a small-eddy surface renewal model informed by in situ meteorological observations. Chamber fluxes averaged 6.9 ± 0.3 mg m−2 d−1 and gas transfer velocities (k600) from the chamber-calibrated surface renewal model averaged 4.0 ± 0.1 cm h−1. We find robust (R2 ≥ 0.93, p

Northern lakes are important sources of greenhouse gases carbon dioxide and methane to the atmosphere. Emissions are expected to increase as the climate continues to warm. Even so, lake carbon budgets are currently poorly constrained. This is in part because of a limited understanding of the processes that govern the flux. This thesis focuses on the physical and biogeochemical drivers of carbon trace gas emissions from three small, post-glacial lakes situated within the Stordalen Mire, a subarctic peatland underlain by thawing permafrost in northern Sweden. A unique, multiyear dataset is used to quantify the importance of different emission pathways – ebullition, turbulence-driven diffusion and release from storage – on short and long timescales. In summer and on seasonal to interannual timescales, emissions are robust functions of thermal energy input. Short-term storage-and-release cycles are governed by kinetic drivers, such as turbulence fuelled by wind shear and, to a lesser extent, by thermal convection. In winter, when the lakes are ice-covered, persistent anoxia and density-driven currents enable methane accumulation at rates exceeding summer emissions. Release at ice-off in spring can constitute the majority of annual methane emissions and scales predictably with ice-cover season length, except in warm winters when snowmelt displaces lake water. Most lake flux studies focus on the warmest summer months and omit the spring efflux, as well as emissions in the colder ice-free months which, because of the well-known temperature-dependency of carbon cycling processes, tend to be low. The latter sampling bias may lead to a substantial overestimation of the ice-free flux in regional and global lake emission budgets. Temperature proxies, potentially combined with gas transfer models, can efficiently gap-fill colder months to arrive at a more representative flux estimate, but important feedbacks, such as lake degassing with increasing wind speed, must be taken into account. The mechanisms emerging from intense study of the Stordalen lakes are likely to be found in a majority of northern lakes, which are small, seasonally ice-covered and of post-glacial origin. However, because gas transfer velocity and temperature sensitivity are spatiotemporally variable, field observations remain essential for the development and calibration of models, and to predict future emissions. At the time of the doctoral defense, the following papers were unpublished and had a status as follows: Paper 2: Submitted. Paper 3: Manuscript.

The area of Knowledge Representation and Reasoning, a subfield of Artificial Intelligence, studies how knowledge can be represented and how it can be used for automated reasoning. Several declarative programming paradigms implement this by developing a formal language to symbolically represent knowledge, as well as an associated form of inference to achieve the desired solution. Recently, the Knowledge Base System (KBS) paradigm has been proposed, based on the idea that knowledge is not inherently linked to a specific reasoning task. Instead, this paradigm proposes to express knowledge in a truly declarative manner. Additionally, to stress reusability of this knowledge, the KBS paradigm allows the knowledge to be combined with one out of a set of possible inferences, each providing a solution to some type of computational task. The initial implementation of this KBS paradigm, also known as the IDP3 system, provided a suitable laboratory to examine a new type of software engineering. State-of-the-art declarative programming systems such as IDP3, clasp, WASP, and lp2sat work using the ground-and-solve methodology. First, the high-level language is ground into a low-level propositional language. As a second step the grounding is used as input for general-purpose, low-level propositional (generally SAT-like) solvers. This thesis contains a thorough study of the impact of different grounding approaches on the solver behaviour. I.e., if the grounding process that is used is smart and results in a smaller low-level representation, does this impact the search behaviour of the underlying solver? The language supported by the IDP3 system contains constraints and definitions. For the purpose of this thesis, we split up definitions into two kinds: input* (also called intentional or stratified predicates, or domain atoms) and search definitions. Input* definitions are definitions that depend on concepts that are known beforehand and can be evaluated. One contribution of this thesis is the evaluation of input* definitions using Tabled Prolog. In order to use these techniques, definitions have to be transformed into a format usable by Prolog. The main challenge in this transformation is correctly projecting away the types and rich language constructs, whilst taking into acount Prolog's left-to-right execution mechanism. Search definitions on the other hand cannot be evaluated because they depend on unknown data. This text elaborates how the above translation to Prolog can be re-used: 1) to partially evaluate definitions, deriving as much information as possible, and 2) ground definitions using a Tabled Prolog execution mechanism. We developed a method to guide the solver to focus on relevant parts in the search space. Only the part of the specification that is linked with the current search branch of the problem has to be taken into account. The goal of this technique is to prevent certain decisions from being made once it is possible to prove that they will not influence the outcome of searching in the current search branch. Additionally, this leads us to an improved stopping criterion for SAT(ID) solvers since any state without relevant decisions is considered an end state, instead of only states in which all literals have been decided. An implementation of technique is presented. The implementation uses a incremental approach, adjusting certain data structures based on the changes in the solver state. With this work we intend to increase the usability of the IDP3 system and elevate it to a "mature" Knowledge Base System. Abstract iii Contents vii List of Acronyms xi List of Symbols xiii List of Figures xvii List of Tables xix 1 Introduction 1 2 Preliminaries: FO(·) 7 2.1 Some Mathematical Basics 7 2.1.1 Sets 7 2.1.2 Lists 8 2.1.3 Type 9 2.1.4 Typings and Tuples 9 2.1.5 Functions 9 2.1.6 Predicate 10 2.2 The Language FO(·) 10 2.2.1 Vocabulary Σ 11 2.2.2 Interpretation I 13 2.2.3 Theory T 17 2.2.4 Concluding Example: Sudoku 21 2.3 Semantics for FO(·) 23 2.3.1 Evaluating a Term in a Structure 24 2.3.2 Semantics of Formulas 24 2.3.3 Semantics of Definitions 24 2.4 Conclusion 26 3 Preliminaries: KBS and IDP 27 3.1 KBS paradigm: Inferences as Tools for Solving Problems 27 3.1.1 Inferences 28 3.1.2 Advantages of the KBS Paradigm 30 3.2 The IDP System 30 3.2.1 The Grounder of IDP 31 3.2.2 The Solver of IDP 40 3.3 Conclusion 43 4 Experimental Evaluation of a State-of-the-art Grounder 45 4.1 Grounding Technique Experiments 46 4.1.1 Experimental Evaluation of RED 49 4.1.2 Experimental Evaluation of LUP 51 4.1.3 Experimental Evaluation of GWB 52 4.2 Solver Behaviour on Optimized Ground Theories 53 4.3 Conclusion 60 5 Complementing the IDP3 system with XSB 63 5.1 Preliminaries 63 5.1.1 Workflow Analysis of IDP3 64 5.1.2 The Calculating Definitions Step 65 5.2 Evaluating input∗ Symbols With XSB 69 5.2.1 Translating FO(·) to a tabled Prolog Program 69 5.2.2 Choosing and Configuring a Prolog System 79 5.2.3 Calling XSB from IDP3 80 5.3 Experimental Evaluation 81 5.4 Refining Definitions With XSB 84 5.5 Conclusion 88 6 Relevance for SAT(ID) 91 6.1 Introduction 91 6.2 Preliminaries 93 6.2.1 PC(ID) 93 6.2.2 Justifications 94 6.3 Relevance 96 6.3.1 Observations 96 6.3.2 Exploiting Relevance 99 6.4 Implementing Relevance as part of an Existing SAT(ID) Solver 100 6.4.1 The Basic Framework 100 6.4.2 Deriving the Justification Status of Literals 101 6.4.3 Implementing the Relevance Tracker 103 6.5 Experimental Evaluation 109 6.6 Conclusion 114 7 Conclusion 117 7.1 Contributions and Conclusions 117 7.2 Future Research Ideas 119 A Additional XSB Prolog code 121 A.1 The forall/2 Predicate in XSB 121 A.2 Translation of IDP3 Built-in Arithmetic Operators to XSB Prolog Code 121 A.3 Translation of IDP3 Aggregate Terms to XSB Prolog Code 123 Bibliography 125 Curriculum Vitae 135 List of Publications 137 status: published

Answer Set Programming (ASP) and FO(·) are two similar state-of-the-art languages for declarative problem solving. This paper presents an automatic transformation from ASP programs to FO(·) programs for the IDP3 system. There are two main language differences that have to be overcome. Firstly there is the introduction of types, which are mandatory in FO(·) but not present in ASP. Secondly there are Herbrand function terms that are present in ASP but not sufficiently supported in IDP3. In this paper we give a short overview of both languages and introduce our transformation that overcomes these differences in the respective languages. ispartof: pages:1-14 ispartof: Proceedings of the 3d Workshop on Grounding, Transforming, and Modularizing Theories with Variables pages:1-14 ispartof: Workshop on Grounding, Transforming, and Modularizing Theories with Variables location:Lexington, KY, USA date:27 Sep - 27 Sep 2015 status: published

FO(·)IDP3 is a declarative modeling language that extends first-order logic with inductive definitions, partial functions, types and aggregates. Its model generator IDP3 grounds the problem into a low-level (propositional) representation and consequently use a generic solver to search for a solution. Recent work introduced a technique that evaluates all definitions that depend on fully known information before the grounding step. In this paper, we extend this technique, which allows us to refine the interpretation of defined symbols when they depend on information that is only partially given instead of completely given. We use our existing transformation of FO(·)IDP3 definitions to Tabled Prolog rules and extend it to support definitions that depend on information that is possibly partially unknown. In this paper we present an algorithm that uses XSB Prolog to evaluate these rules in such a way that we achieve the most precise possible refinement of the defined symbols. Experimental results show that our technique derives extra information for the defined symbols. ispartof: pages:15-29 ispartof: Proceedings of the international joint workshop on implementation of constraint and logic programming systems and logic-based methods in programming environments 2014 pages:15-29 ispartof: International Joint Workshop on Implementation of Constraint and Logic Programming Systems and Logic-based Methods in Programming Environments location:Vienna, Austria date:17 Jul - 18 Jul 2014 status: published