Studies of drought have increased in light of new data availability and advances in spatio-temporal analysis. However, the following gaps still need to be filled: 1) methods to characterise drought that explicitly consider its spatio-temporal features, such as spatial extent (area) and pathway; 2) methods to monitor and predict drought that include the above-mentioned characteristics and 3) approaches for visualising and analysing drought characteristics to facilitate interpretation of its variation. This research aims to explore, analyse and propose improvements to the spatio-temporal characterisation of drought. Outcomes provide new perspectives towards better prediction.
The following objectives were proposed. 1) Improve the methodology for characterising drought based on the phenomenon’s spatial features. 2) Develop a visual approach to analysing drought variations. 3) Develop a methodology for spatial drought tracking. 4) Explore machine learning (ML) techniques to predict crop-yield responses to drought. The four objectives were addressed and results are presented.
Finally, a scope was formulated for integrating ML and the spatio-temporal analysis of drought. Proposed scope opens a new area of potential for drought prediction (i.e. predicting spatial drought tracks and areas). It is expected that the drought tracking and prediction method will help populations cope with drought and its severe impacts.
Author(s): Vitali Díaz Mercado
Publisher: CRC Press
Year: 2022
Language: English
Pages: 160
City: Boca Raton
COVER
TITLE PAGE
COPYRIGHT PAGE
ACKNOWLEDGMENTS
SUMMARY
SAMENVATTING
CONTENTS
1 INTRODUCTION
1.1 Backgrmmd
1.1.1 Drought
1.1.2 Drought study
1.1.3 Machine learning
1.2 PROBLEM STATEMENT
1.3 RESEARCH HYPOTHESES
1.4 RESEARCH OBJECTIVES
1.5 DISSERTATION STRUCTURE
2 LITERATURE REVIEW
2.1 DROUGHT
2.2 DROUGHT INDICATORS
2.2.1 Meteorological drought indicators
2.2.2 Agricultural drought indicator
2.2.3 Hydrological drought indicator
2.2.4 Main difficulties of a standardized drought indicatorcomputation
2.3 DROUGHT CALCULATION
2.4 VISUALISATION AND ANALYSIS OF DROUGHT VARIATION
2.5 MACHINE LEARNING TECHNIQUES IN DROUGHT STUDIES
2.6 SUMMARY AND CONCLUSIONS
3 METHODOLOGICAL FRAMEWORK
3.1 IMPROVE THE METHODOLOGY FOR CHARACTERISING DROUGHT INSPACE AND TIME BASED ON THE PHENOMENON'S SPATIAL FEATURES,SUCH AS SPATIAL EXTENT AND LOCATION
3.2 DEVELOP A VISUAL APPROACH TO ANALYSING VARIATIONS OF SPATIOTEMPORALDROUGHT CHARACTERISTICS
3.3 DEVELOP A METHODOLOGY FOR MONITORING THE SPATIAL EXTENT OFDROUGHT, I.E. DROUGHT TRACKING
3.4 EXPLORE THE APPLICABILITY OF USING ML TECHNIQUES TO PREDICTCROP-YIELD RESPONSES TO DROUGHT BASED ON SPATIO-TEMPORALDROUGHT CHARACTERISTICS
4 CASE STUDIES AND DATA
4.1 MEXICO
4.1.1 Large scale
4.1.2 Catchment scale
4.2 INDIA
4.2.1 Large scale
4.2.2 Regional scale
5 SPATIO-TEMPORAL DROUGHT CHARACTERISATION
5.1 INTRODUCTION
5.2 SPATIO-TEMPORAL ANALYSIS OF DROUGHT (STAND)
5.2.1 Temporal analysis
5.2.2 Spatio-temporal analysis: first approach
5.2.3 Spatio-temporal analysis: second approach
5.2.4 Spatio-temporal drought patterns
5.3 SUMMARY AND CONCLUSIONS
6 COMPARISON OF DROUGHT INDICATORS
6.1 INTRODUCTION
6.2 METHODS AND DATA
6.3 RESULTS AND DISCUSSION
6.3.1 First method: catchment-aggregated drought indicator
6.3.2 Second method: drought areas
6.4 SUMMARY AND CONCLUSIONS
7 MACHINE-LEARNING APPROACHTO CROP YIELD PREDICTION
7.1 INTRODUCTION
7.2 ML MODELLING METHODOLOGY
7 .2.1 Step 1. Data preparation
7.2.2 Step 2. Input variable selection
7.2.3 Step 3. Polynomial regression models calculation
7.2.4 Step 4. Artificial neural network models calculation
7.2.5 Step 5. Models application and combination
7.3 RESULTS AND DISCUSSION
7.3.1 Correlation analysis
7.3.2 Polynomial regression (PR) models
7.3.3 Artificial neural network (ANN) models
7.3.4 ML modelling limitations
7.4 SUMMARY AND CONCLUSIONS
8 VISUAL APPROACHES TODROUGHT ANALYSIS
8.1 INTRODUCTION
8.2 METHODS AND DATA
8.2.1 Visual approaches to drought analysis
8.2.2 Experiment setup
8.3 RESULTS AND DISCUSSION
8.4 SUMMARY AND CONCLUSIONS
9 SPATIAL DROUGHT TRACKING DEVELOPMENT
9.1 INTRODUCTION
9.2 METHODS
9.2.1 S-TRACK: Spatial tracking of drought
9.2.2 Calculation of drought characteristics
9.3 EXPERIMENTAL SETUP
9.3.1 Drought indicator data
9.3.2 Drought areas and centroids
9.3.3 Tracking algorithm setup and evaluation
9.4 RESULTS
9.4.1 Drought areas and centroids
9.4.2 Drought tracks and characteristics (generic mode)
9.4.3 Qualitative evaluation (specific mode)
9.5 DISCUSSION
9.5.1 Drought indicator and areas
9.5.2 Drought tracking method
9.6 SUMMARY AND CONCLUSION
10 CONCLUSIONS AND RECOMMENDATIONS
10.1 GENERAL
10.2 (O1) SPATIO-TEMPORAL CHARACTERISATION OF DROUGHT BASED ONTHE PHENOMENON'S SPATIAL FEATURES
10.2 (O1) SPATIO-TEMPORAL CHARACTERISATION OF DROUGHT BASED ONTHE PHENOMENON'S SPATIAL FEATURES
10.3 (O2) VISUAL APPROACHES TO ANALYSING SPATIO-TEMPORALDROUGHT VARIATION
10.4 (O3) METHODOLOGY FOR DROUGHT TRACKING
10.5 (O4) ML MODELS TO PREDICT CROP YIELD RESPONSES TO DROUGHT
10.6 CONCLUSION IN BRIEF
REFERENCES
ANNEXES
ABOUT THE AUTHOR
