Month: June 2024

Downscaling the MODIS land surface temperature using a trapezial concept applied to the MODIS and sentinel 2 images

Latest Results

Abstract  

Downscaling methods are crucial for accessing high-resolution thermal data simultaneously. The DisTRAD model is commonly used for downscaling thermal images, but changes in soil moisture, such as those caused by irrigation operations, can lead to errors in the process. This study investigated the potential use of TOTRAM and OPTRAM models to reduce errors in LST downscaling in irrigated fields. Sentinel satellite imagery was utilised to enhance the resolution of MODIS Land Surface Temperature (LST) from 1000 to 20 m in the fields of Megsal and Hezarjolfa agro-industrial company in Qazvin province. Soil moisture was estimated using the OPTRAM model, and the results were compared with observational data. The findings indicated that on days with NDVI greater than 0.6, the R2 value exceeded 0.88 and the RMSE value was less than 0.06 cm3/cm3. Then, MODIS LST images were downscaled to 20 m using codes in Google Earth Engine (GEE). Evaluation was conducted using observational data from collected land surface temperature data for 36 points. Comparison of the downscaled LST data with observational data on days with irrigation revealed a decrease in MAE and RMSE error indices by approximately 0.4 and 1.2 degrees Celsius, respectively, in the OPTRAM-TPTRAM model compared to the DisTRAD model. Consequently, the OPTRAM-TOTRAM model generally outperforms the DisTRAD model in LST downscaling. Lastly, it is recommended to assess the TOTARM and OPTRAM models for downscaling MODIS sensor LST in other irrigated fields.

Understanding the challenges affecting food-sharing apps’ usage: insights using a text-mining and interpretable machine learning approach

Latest Results

Abstract

Food waste is a serious problem affecting societies and contributing to climate change. About one-third of all food produced globally is wasted, while millions of people remain food insecure. Food-sharing apps attempt to simultaneously address ‘hunger’ and ‘food waste’ at the community level. Though highly beneficial, these apps experience low usage. Existing studies have explored multiple challenges affecting food-sharing usage, but are constrained by limited data and narrow geographical focus. To address this gap, this study analyzes online user reviews from top food-sharing apps operating globally. A unique approach of analyzing text data with interpretable machine learning (IML) tools is utilized. Eight challenges affecting food-sharing app usage are obtained using the topic modeling approach. Further, the review scores representing user experience (UX) are assessed for their dependence on each challenge using the document-topic matrix and machine learning (ML) procedures. Tree-based ML algorithms, namely regression tree, bagging, random forest, boosting, and Bayesian additive regression tree are employed. The best-performing algorithm is then complemented with IML tools such as accumulated local effects and partial dependence plots, to assess the impact of each challenge on UX. Critical improvement areas to increase food-sharing apps’ usage are highlighted, such as service responsiveness, app design, food variety, and unethical behavior. This study contributes to the nascent literature on food-sharing and IML applications. A significant advantage of the methodological approach utilized includes better explainability of ML models involving text data, at both the global and local interpretability levels, in terms of the associated features and feature interactions.

Artificial Intelligence on The Couch. Staying Human Post-AI

Latest Results

Abstract

This paper examines the human relationship to technology, and AI in particular, including the proposition that algorithms are the new unconscious. Key is the question of how much human ability will be duplicated and transcended by general machine intelligence. More and more people are seeking connection via social media and interaction with artificial beings. The paper examines what it means to be human and which of these traits are already or will be replicated by AI. Therapy bots already exist. It is easier to envision AI therapy guided by CBT manuals than psychoanalytic techniques. Yet, a demonstration of how AI can already perform dream analysis reaching beyond a dream’s manifest content is presented. The reader is left to consider whether these findings demand a new role for psychoanalysis in supporting, sustaining, and reframing our humanity as we create technology that transcends our abilities.

Machine learning-based spatial downscaling and bias-correction framework for high-resolution temperature forecasting

Latest Results

Abstract

High-resolution temperature forecasting plays a crucial role in various applications such as climate impact assessment, hydrological modeling, and localized weather forecasting. The existing low-resolution forecast data may not accurately capture the fine-grained temperature patterns required for localized predictions. These forecasts may contain biases that need to be corrected for accurate results. Therefore, there is a need for an effective framework that can downscale low-resolution forecast data and correct biases to generate high-resolution temperature forecasts. The paper proposes a machine learning-based spatial downscaling and bias-correction framework for high-resolution temperature forecasting. The framework utilizes low-resolution forecast data from the European Centre for Medium-Range Weather Forecasts and real-time 1km analysis product data from the National Meteorological Administration, to generate high-resolution 1km forecast temperature data. The framework consists of four modules: data acquisition module, data preprocessing module, downscaling and correction model module, and post-processing and visualization module. Through experiments, it demonstrated that the framework has superior performance and potential in meteorological data downscaling and correction and can be used to achieve real-time high-resolution temperature forecasting, which has important significance for various applications such as climate impact assessment, hydrological modeling, and localized weather forecasting.

Climate change trend analysis and future projection in Guguf watershed, Northern Ethiopia

Latest Results

Abstract

According to Intergovernmental panel on Climate Change (IPCC) Climate change is the weather characteristics such as precipitation, air temperature, humidity, wind, sunshine, cloud cover, and atmospheric pressure at a specific location determined over a long period of at least 30 years. The main objective of this study was to analyse the climate trend and future projection in Guguf watershed of Southern Tigray, Ethiopia. 32 years (1987–2018) Meteorological data were collected from the Ethiopian Meteorological Institute. Download canESM2 (Canadian Second Generation Earth System Model). The Mann-Kendal trend test was used to test for the presence of trends using XLSTAT. The SDSM 4.2.9 decision support tool was used to downscale large scale predictors and project future climate change. The period from 1987 to 2018 was considered as a base period, whereas the period from 2019 to 2100 was considered as future periods. Historically, from 1987 to 2018, there was an overall increase in the mean annual minimum and maximum temperatures by 0.016 °C and 0.048 °C, respectively, with a little decrease in the average annual rainfall (up to 0.685 mm). The highest increment of maximum temperature recorded in October month up to + 2.7 °C in RCP8.5 scenarios. The precipitation increases up to a maximum of 49% (2073–2100) for the RCP4.5 scenario and 66% (2073–2100) for the RCP4.5 (representative concentration pathway 4.5) scenario in the Belg (February to May). Precipitation decreases in the Kiremt (June to September) season by 8% (2019–2045) and 23% (2073–2100) for RCP4.5 scenarios. Future work needs to consider studying the effects of different climate change adaptation strategies.