Proceedings of the Technical Sessions, 42 (2026) 90-95Institute of Physics, Sri Lanka 94Machine learning assisted identification of multi-layer 2D materials using optical microscopyb. Dataset Limitations and False NegativesA significant factor affecting the reported accuracy was the \"partial annotation\" of the ground truth dataset. As seen in the confusion matrices and visual outputs (Fig. 25), there were instances where the models correctly identified valid graphene flakes that were unlabelled in the ground truth. These \"correct\" predictions were mathematically penalized as false positives during evaluation, artificially lowering the precision scores. This suggests the actual capabilities of both models may be higher than the measured metrics indicate.Figure 3: (a) ground truth (b) Predicted using Yolov8 - Example of non-annotated data and predictions6. CONCLUSIONThis study demonstrates that deep learning significantly accelerates the characterization of 2D materials, offering a viable alternative to subjective human visual inspection. By automating the identification process, the proposed method reduces the inconsistencies and time constraints associated with manual microscopy, while remaining far more cost-effective than advanced techniques like AFM or Raman spectroscopy.The results indicate a clear operational distinction: YOLOv8 is recommended for highthroughput screening due to its superior 0.757 IoU score and detection speed, whereas Mask R-CNN remains essential for applications requiring detailed morphological analysis. A critical insight from this work is that model performance was constrained not by architecture, but by dataset quality; the models frequently detected valid flakes that were absent in the ground truth annotations. Consequently, future work should prioritize the development of a fully annotated, high-density dataset to eliminate false negatives. Furthermore, future research will explore lightweight segmentation models that may offer a middle ground between the speed of YOLOv8 and the pixel-level precision of Mask R-CNN.
Proceedings of the Technical Sessions, 42 (2026) 90-95Institute of Physics, Sri Lanka 95Machine learning assisted identification of multi-layer 2D materials using optical microscopy6. REFERENCES[1] L. Rodríguez-Pérez, M. A. Herranz, and N. Martín, The chemistry of pristine graphene, Chem. Commun., 49 (2013) 3721. [2] K. I. Bolotin et al., Ultrahigh electron mobility in suspended graphene, Solid State Commun., 146 (2008) 351-355. [3] S. Masubuchi et al., Deep-learning-based image segmentation integrated with optical microscopy, Npj 2D Mater. Appl., 4 (2020). [4] S. Masubuchi et al., Data for: Deep-Learning-Based Image Segmentation, figshare. Dataset (2020). [5] R. Y. Ju and W. Cai, Fracture detection in pediatric wrist trauma X-ray images using YOLOv8 algorithm, Sci. Rep., 13 (2023) 20077.
Proceedings of the Technical Sessions, 42 (2026) 96-105Institute of Physics, Sri Lanka 96DEVELOPMENT OF A FULLY INKJET PRINTED ANTENNA ON POLYETHYLENE TEREPHTHALATE FOR ENERGY HARVESTING PURPOSES IN SMART LABELLINGDEVELOPMENT OF A FULLY INKJET PRINTED ANTENNA ON POLYETHYLENE TEREPHTHALATE FOR ENERGY HARVESTING PURPOSES IN SMART LABELLINGW. A. I. B. J. Wickramasinghea, S. W. D. K. R. M. Manamendra, W. L. P. K. Wijesinghe, G.C. Wickramasinghe, D. L. Weerawarne*Department of Physics, Faculty of Science, University of Colombo, Sri [email protected]*[email protected]. ABSTRACTFlexible printed electronics offer a revolutionary approach to smart labelling. However, achieving energy autonomy remains a critical hurdle. While passive Near Field Communication (NFC) enables battery-free operation via electromagnetic energy harvesting, peak performance necessitates precise resonance around 13.56 MHz. This study investigates the fabrication of single-sided flexible antenna using inkjet printing by a regular office inkjetprinter, specifically addressing the usefulness of such printers in electronics fabrication and addressing the geometric deviations inherent in the process. Analysis of silver (Ag) nanoparticle conductive traces printed on polyethylene terephthalate (PET) substrate revealed that line-width deviations are highly dependent on motion of the print-head orientation. To mitigate this, a directional compensation strategy was developed to offset fabrication deviations during the digital design phase. The optimized single-layer antenna on PET achieved a harvested voltage of 3.220 ± 0.001 V, comparable to a standard rigid PCB antenna, yet exhibited a high series resistance of 583.2 ± 0.1 Ω with a low Q-factor. To overcome these limitations, a multilayer printing technique was implemented. The resulting three-layer inkjetprinted Ag antenna on PET reduced series resistance to 190.5 ± 0.1 Ω and increased harvested voltage to 3.250 ± 0.001 V with higher resonance characteristics. These results demonstrate that geometrically compensated, multilayer inkjet-printed antennas on PET substrates can achieve the electrical performance necessary for reliable, battery-free smart labelling in highimpedance, low-power sensing applications.Keywords: Inkjet Printing, Near Field Communication, Energy Harvesting, Smart Labelling2. INTRODUCTIONPrinted electronics enables the development of \"smart labels\" for packaging by creating circuits on a flexible substrate [1], [2]. To achieve energy autonomy, these devices often utilize Passive Near Field Communication (NFC) to harvest energy from electromagnetic fields [3], [4]. However, efficient harvesting requires the antenna to resonate at 13.56 MHz, which depends critically on the geometric accuracy of the printed inductance [5]. This poses a significant challenge for inkjet printing, particularly when using general purpose office inkjet printers. Unlike the subtractive etching processes used for rigid PCBs, inkjet printing is an additive process susceptible to ink spreading and geometric anisotropy [6]. This anisotropy arises because deposition accuracy is governed by the line's orientation relative to the print head’s scanning axis. While lines parallel to the head motion (0°) are deposited continuously, lines perpendicular to this motion (90°) are constructed via rasterized segments or discontinuous
Proceedings of the Technical Sessions, 42 (2026) 96-105Institute of Physics, Sri Lanka 97DEVELOPMENT OF A FULLY INKJET PRINTED ANTENNA ON POLYETHYLENE TEREPHTHALATE FOR ENERGY HARVESTING PURPOSES IN SMART LABELLINGfiring. This mechanical difference causes deviation to increase as the print angle diverges from the scanning axis, significantly altering the antenna's impedance and resonant frequency.This aims to establish the suitability of regular office inkjet printer in fabricating electronics and quantifies geometric deviations on PET substrate to develop a directional compensation strategy that offsets fabrication deviations in the digital design phase. We validate this approach by comparing the resonance quality and energy harvesting voltage of the optimized flexible antenna against single sided PCB antenna.3. THEORYSeries LCR Resonance frequency, Inductance and Q-FactorResonance occurs in an LC circuit when the inductive reactance (?? = 2???) equals the capacitive reactance (?? = 1/2??? ). At this point, denoted as the resonant frequency (?0) in Equation 1, the circuit oscillates at its natural frequency determined by its inductance and capacitance:?0 =12?√??(1)The energy in the circuit swings between the inductor and the capacitor. Ideally, if there were no resistive losses, the energy would oscillate indefinitely. Real circuits, however, have inherent resistance that dissipates energy, damping the oscillations over time.Ideally, the NFC antenna functions as a lossless LC tank circuit oscillating at ???? =13.56 ???. However, printed antennas possess significant internal resistance, effectively forming a series LCR circuit. The quality factor (Q-factor) in Equation 2 of such a system, which defines the sharpness of the resonance peak and the efficiency of energy storage, is given by:? =1?√??(2)where R is the total series resistance, L is the equivalent inductance, and C is the tuning capacitance. As indicated by the equation, the Q-factor is inversely proportional to resistance. Consequently, antennas with high internal resistance exhibit a low Q-factor, resulting in rapid energy dissipation and a broad, shallow resonance curve (S11) that may be difficult to detect via inductive coupling methods.Antennas operating at 13.56 MHz which are used for NFC communication, can be designed in various forms for different application needs. The focus of this study is solely on square-shaped antennas. One of the crucial parameters for these antennas is the equivalent inductance (L) of the antenna. Therefore, the Inductance of squared antennas can be calculated using the Equation 3 according to their physical dimensions.
Proceedings of the Technical Sessions, 42 (2026) 96-105Institute of Physics, Sri Lanka 98DEVELOPMENT OF A FULLY INKJET PRINTED ANTENNA ON POLYETHYLENE TEREPHTHALATE FOR ENERGY HARVESTING PURPOSES IN SMART LABELLINGLant = K1 ∙ μ0 ∙ N2∙ [dout + din21 + K2 ∙ (dout − dindout + din)] (3)?? = 2.34?? = 2.75?? = 4 π × 10–7 H/m? = number of turns???? = outer diameter of the antenna??? = inner diameter of the antenna???? = self-inductance of the loop antennameasured in Henry at 13.56 MHz4. METHODOLOGY AND INSTRUMENTATIONa. Development of Near Field Communication (NFC) antennaThe antenna design began by identifying the internal tuning capacitance of the M24LR04E-R IC, which was determined to be 27.5 pF from the datasheet [7]. Subsequently, following the guidelines in the application note and applying the resonant frequency formula in Equation 1for 13.56 MHz, the essential equivalent inductance for the antenna was calculated to be 5.0 μH. An antenna with an approximate inductance of 5.0 μH was designed utilizing the eDesignSuite[8], [9]. This online tool, developed by STMicroelectronics NV, facilitates the creation of NFC antenna designs tailored to specific requirements. Utilizing this tool is a more time-efficient option than manually calculating antenna dimensions using the Equation (3). However, the tool estimates equivalent inductance based on rigid PCB parameters and corrections were essential for flexible materials due to inductance differences. The antenna designed via eDesignSuite featured dimensions of 42 mm × 42 mm with eleven turns, a conductor width of 0.6 mm, and a spacing of 0.6 mm as shown in Figure 2. To enhance practicality for printing and connectivity, the design was further refined using Inkscape to incorporate modifications such as soldering pads.Development of the flexible antenna on the PET substrate was done by using an office inkjetprinter and the performance was compared to a rigid PCB antenna. The microprocessor M24LR04E-R was used for energy harvesting through NFC communication capabilities [7]. The circuit was designed without an onboard antenna and with the ability to connect antennas externally. The circuit was configured using the \"M24LR.h\" Arduino library to enable its energy harvesting mode. Following fabrication, the firmware required for NFC communication and energy harvesting was uploaded to the custom board via an Arduino UNO.Figure 1: Physical dimensions of a square antenna taken into account to calculate self -inductance.
Proceedings of the Technical Sessions, 42 (2026) 96-105Institute of Physics, Sri Lanka 99DEVELOPMENT OF A FULLY INKJET PRINTED ANTENNA ON POLYETHYLENE TEREPHTHALATE FOR ENERGY HARVESTING PURPOSES IN SMART LABELLINGi. Fabrication of single-side rigid antenna The antenna was fabricated as-designed on a rigid PCB, and its inductance was measured using a Keysight Technologies U1731C LCR meter at 100 kHz, the maximum frequency setting available on the device. Afterwards, the antenna was connected to the previously developed circuit to evaluate its NFC functionality and energy harvesting performance. To verify NFC communication, a test data string was written to the board’s memory via the firmware. An NFCenabled mobile phone (Nokia 6.1) served as both the electromagnetic energy source and the NFC reader. The test string was successfully retrieved by the phone to confirm connectivity, while the maximum harvested voltage output from the board was measured using a multimeter. Finally, the antenna's return loss curve (?11) was assessed using an AURSINC-manufactured NanoVNA (Nano Vector Network Analyzer).ii. Fabrication of single-side flexible antenna Achieving precise geometric dimensions is crucial for fabrication an NFC flexible antenna on a PET substrate. However, inkjet printing, the actual dimensions of the printed lines often deviate from the intended design, particularly with a domestic printer. This deviation is primarily attributed to printer characteristics and the orientation of the lines relative to the print head's motion [10]. To investigate the impact of print orientation on this deviation, the following methodology was employed.(a) (b)Figure 2: (a) interface of eDesign suit with the proposed antenna design and (b) proposed NFC antenna design after enhancing using Inkscape.
Proceedings of the Technical Sessions, 42 (2026) 96-105Institute of Physics, Sri Lanka 100DEVELOPMENT OF A FULLY INKJET PRINTED ANTENNA ON POLYETHYLENE TEREPHTHALATE FOR ENERGY HARVESTING PURPOSES IN SMART LABELLING4.1.2.1 Microscopic analysis of inkjet printed linesTo investigate dimensional deviation in inkjet-printed NFC antenna patterns, test coupons in Figure 3 (a) were designed using Inkscape software. The design featured conductive lines with intended widths and separation gaps as 0.9 mm, 0.6 mm, and 0.3 mm. The test coupon in Figure 3 (a) was printed using an EPSON L130 printer with NovaCentrix Metalon® JS-B25P silver nanoparticle ink on NovaCentrix: Novele™ IJ-220 PET. The coupons were printed at orientation angles ranging from 0° to 180° in 18° increments relative to the print head motion.After printing, these lines on PET were sintered at 110 °C for four hours.The printed lines in shown in figure 4 were imaged using an ANDONSTAR AD 249s digital microscope with a calibration scale. Analysis of these images was performed using a custom Python script that used Otsu’s thresholding method to separate the printed lines from the background. The script calculated the mean width and standard deviation of the lines by measuring the pixel width at every position along the line length.Subsequent antenna designs were adapted for PET based on the deviation data obtained from the above line analysis phase, including the design fabricated on a rigid PCB. Three designs in table 1 were fabricated on PET, using silver ink by EPSON L130 printer and sintered under 110 °C for 4 hours.The design 1 on PET showed resistance of 516.5 ± 0.1 Ω. Inductance measurements taken (Keysight U1731C) revealed discrepancies between the simulated and actual values. While, interfaced with the NFC circuit, this flexible antenna failed to achieve successful communication, likely due to inductance mismatches caused by substrate flexibility and ink spreading. To address this, two additional single - side antennas of design 2 and design 3 were developed with adjusted inductances and dimensions according to the dimensions in Table 1.(a) (b)Figure 4: (a) Detecting edges of the printed lines using python script and (b) detecting edges of the printed line and gaps using python script.(a) (b)Figure 3: (a) The test coupon design that use for line analysis and (b) dimensions of single cell of the test coupon
Proceedings of the Technical Sessions, 42 (2026) 96-105Institute of Physics, Sri Lanka 101DEVELOPMENT OF A FULLY INKJET PRINTED ANTENNA ON POLYETHYLENE TEREPHTHALATE FOR ENERGY HARVESTING PURPOSES IN SMART LABELLINGTable 1 - dimensions of the antenna designs (The following dimensions were adjusted in the vector file according to the results obtained in the ‘Inkjet Printed Line Analysis’ phase).DesignNo.Size(mm × mm)Turns (N)Trace Width (mm)Spacing (mm)Est. Inductance (μH)1 42 × 42 11 0.6 0.6 5.012 40 × 40 14 0.4 0.4 9.123 33 × 33 15 0.3 0.3 8.61Then all these antennas were connected to a custom PCB that was designed and evaluated for NFC communication using a mobile phone, and maximum harvested voltage outputs were measured using a multimeter. Then, the return loss curves were observed using NanoVNA.Based on the results from the above experiment, it was necessary to identify a method to reduce the internal resistance of the antennas fabricated on PET. One potential solution to achieve this is printing multiple conductive layers on the same trace. Consequently, a three-layer version of Antenna Design 2 was fabricated, and its internal resistance and performance were evaluated following the experimental procedure described above.5. RESULTS AND DISCUSSIONa. Geometric Deviation AnalysisTo quantify fabrication tolerances, conductive lines were printed on PET at orientation angles ranging from 0° (parallel to print head motion) to 180°. Microscopic analysis revealed a significant correlation between print orientation and dimensional accuracy; specifically, the results showed minimum deviation from the expected dimensions at 0° (and 180°) and maximum deviation at 90°.Visual Inspection: Lines printed at 0°, as indicated in Figure 5(a), exhibited smooth, welldefined edges. In contrast, lines printed at 90° shown in Figure 5(b) increased roughness and ink dispersion perpendicular to the print direction. Figure 5: Microscopic image of silver lines (a) printed line with 0° angle relative to the motion of the printer head on PET and (b) printed line with 90° angle relative to the motion of the printer head on PET.(a) (b)
Proceedings of the Technical Sessions, 42 (2026) 96-105Institute of Physics, Sri Lanka 102DEVELOPMENT OF A FULLY INKJET PRINTED ANTENNA ON POLYETHYLENE TEREPHTHALATE FOR ENERGY HARVESTING PURPOSES IN SMART LABELLINGQuantitative Deviation: Image analysis (using the Otsu thresholding method) confirmed that dimensional deviation peaks when lines are printed perpendicular to the print head motion.Table 2 illustrates the variations of the lines widths and gaps vary as following ranges for PET. However, these deviation values were independent from the target line widths and gaps values.Table 2 - Measured maximum geometric deviations for printed lines and separation gaps at 0° and 90° anglesSubstrateTarget Line Widths(mm)Max Width Increase at0° (mm)± 0.005 mmMax Gap Reduction at 0°(mm)± 0.005 mmMax width Increase at 90° (mm)± 0.01 mmMax Gap Reduction at 90° (mm)± 0.01 mmPET 0.3, 0.6, 0.9From+ 0.040to+ 0.050From- 0.042to- 0.085From+0.10to+0.12From-0.12to-0.18Design Compensation Strategy: Based on these results, a compensation factor was applied to the digital designs. For example, to achieve a target trace width of 0.9 mm in a vertical (90°) orientation on PET, the digital design was pre-compensated to 0.79 mm (0.9 mm - 0.11 mm offset). This calibration ensured that the final antenna dimensions matched the inductive requirements for resonance at 13.56 MHz. Table 3 - Design compensation factors applied to counteract ink spreading.SubstrateCompensation Factor Applied for width at 0° (mm)Compensation Factor Applied for gap at 0° (mm)Compensation Factor Applied for width at 90° (mm)Compensation Factor Applied for gap at 90° (mm)PET - 0.045 + 0.064 - 0.11 + 0.15b. Flexible antenna performance analysisThe electrical characteristics and NFC performance of the three single-sided antenna designs fabricated on PET are summarized in Table 4. Further, the single side rigid PCB antenna performance was compared with antennas on PET.
Proceedings of the Technical Sessions, 42 (2026) 96-105Institute of Physics, Sri Lanka 103DEVELOPMENT OF A FULLY INKJET PRINTED ANTENNA ON POLYETHYLENE TEREPHTHALATE FOR ENERGY HARVESTING PURPOSES IN SMART LABELLINGTable 4 - key parameters and the performances of the antenna designsSubstratePCB PETDesign 1(1 layer)Design 1(1 layer)Design 2(1 layer)Design 2(3 layers)Design 3(1 layer)Est. Inductance from the eDesign suit at13.56 MHz (μH)5.01 5.01 9.12 9.12 8.61Actual Inductance (μH) 4.82 ± 0.01 3.00 ± 0.01 6.20 ± 0.01 6.20 ± 0.01 5.20 ± 0.01Antenna resistance (Ω) 3.387 ± 0.001 516.5 ± 0.1 583.2 ± 0.1 190.5 ± 0.1 752.7 ± 0.1Maximum voltage of the harvested energy (V)3.255 ± 0.001 0.048 ± 0.001 3.220 ± 0.001 3.250 ± 0.001 2.575 ± 0.001NFC communication compatibility Yes No Yes Yes YesThe flexible antennas fabricated on PET demonstrated functional inductance values (6.20 μH), while bit higher than the target, remained within the tuning range of the M24LR04E-R IC when combined with its internal capacitance and significantly higher resistance (⁓ 500 – 750 Ω). Consequently, designs 2 and 3 on PET supported NFC communication. Notably, Design 2 achieved a harvested voltage of 3.220 V, which is comparable to the reference rigid PCB (3.25 V). However, the resistance still remained significantly high relatively to PCB antenna, resulting in a low Q-factor. To address this, a three-layer version of Design 2 was fabricated on PET. This modification reduced the resistance by approximately a factor of three, from 583.2 ± 0.1 Ω to 190.5 ± 0.1 Ω, while achieving a comparable harvested voltage of 3.250 V.Return Loss (S11) Analysis: Figure 6 show the return loss curves for the copper PCB antennas with distinct resonance dips around 13.56 MHz. However, the flexible antenna (Design 2) lacks this distinct dip despite functional NFC coupling. This is attributed to the antenna's relatively high series resistance (≈ 583 Ω), which drastically lowers the Q-factor of the LC circuit [11], [12]. The low Q-factor broadens the resonance peak, making the power absorption dip indistinguishable during non-contact measurements, even though sufficient energy transfer occurs for device operation. However, the three-layer flexible antenna (Design 2) exhibits a shallow and broad, yet observable, dip in the curve. This indicates that the reduced internal resistance of the multilayer antenna (≈ 190 Ω) has increased the Q-factor of the LC circuit.
Proceedings of the Technical Sessions, 42 (2026) 96-105Institute of Physics, Sri Lanka 104DEVELOPMENT OF A FULLY INKJET PRINTED ANTENNA ON POLYETHYLENE TEREPHTHALATE FOR ENERGY HARVESTING PURPOSES IN SMART LABELLINGFigure 6: Return-Loss (S11) curves of the fabricated antennas6. CONCLUSIONThe study successfully characterized and optimized inkjet-printed NFC antennas on PET. Geometric analysis revealed significant anisotropy, with conductive lines widening by up to 0.12 mm at a 90° print orientation while remaining relatively accurate at 0°. By implementing a directional compensation strategy, specifically reducing vertical trace widths in the digital design, we achieved the geometric precision required for resonance around 13.56 MHz for antenna on PET. The reference rigid PCBs established a voltage benchmark of 3.25 – 3.33 V, the optimized flexible antenna (Design 2 on PET) achieved a comparable 3.220 ± 0.001 V. Crucially, despite inherent material challenges, the flexible antenna successfully established reliable NFC communication with the reader, validating the design's functionality. However, a trade-off was observed between voltage generation and power delivery. The single-layer flexible antenna on PET exhibited a shallower return loss S11 dip compared to its rigid counterparts. This is attributed to the high series resistance (583.2 ± 0.1 Ω), which lowers the circuit's Q-factor and limits current sourcing capability due to ohmic losses. To address this, multilayer printing was explored as a solution. A three-layer version of the antenna design 2 fabricated on PET demonstrated much more performance improvements. This approach reduced the series resistance by a factor of three (to 190.5 ± 0.1 Ω) compared to the singlelayer counterpart, thereby increasing the antenna's Q-factor. This improvement was evidenced by a shallow and broad but distinct dip in the return loss curve around the NFC resonance frequency. In conclusion, this work demonstrates that inkjet-printed antennas fabricated by regular desktop inkjet printer are viable alternatives to rigid PCBs for high-impedance, lowpower sensing nodes. Future work should focus on further reducing trace resistance to support current-intensive loads. This could be achieved by refining multilayer alignment techniques to allow for higher layer counts or by investigating alternative inks with inherently lower resistivity to enhance radiation efficiency.
Proceedings of the Technical Sessions, 42 (2026) 96-105Institute of Physics, Sri Lanka 105DEVELOPMENT OF A FULLY INKJET PRINTED ANTENNA ON POLYETHYLENE TEREPHTHALATE FOR ENERGY HARVESTING PURPOSES IN SMART LABELLING7. REFERENCES[1] C. Yue, J. Wang, Z. Wang, B. Kong, and G. Wang, “Flexible printed electronics and their applications in food quality monitoring and intelligent food packaging: Recent advances,” Food Control, vol. 154, p. 109983, Dec. 2023, doi: 10.1016/j.foodcont.2023.109983.[2] S. Tripathi, S. Bisht, P. Kumar, Md. R. Mia, and K. K. Gaikwad, “Printable and flexible electronics for smart packaging applications: status, challenges, and opportunities,” Journal of Materials Science: Materials in Electronics, vol. 36, no. 25, p. 1583, Sep. 2025, doi: 10.1007/s10854-025-15626-w.[3] H. Kang et al., “Fully roll-to-roll gravure printable wireless (13.56 MHz) sensor-signage tags for smart packaging,” Sci. Rep., vol. 4, Jun. 2014, doi: 10.1038/srep05387.[4] C. L. Baumbauer et al., “Printed, flexible, compact UHF-RFID sensor tags enabled by hybrid electronics,” Sci. Rep., vol. 10, no. 1, p. 16543, Oct. 2020, doi: 10.1038/s41598-020-73471-9.[5] Q.-T. Luu, S. Koulouridis, A. Diet, Y. Le Bihan, and L. Pichon, “Investigation of inductive and radiating energy harvesting for an implanted biotelemetry antenna,” in 2017 11th European Conference on Antennas and Propagation (EUCAP), Paris: IEEE, Mar. 2017, pp. 160–163. doi: 10.23919/EuCAP.2017.7928620.[6] R. Vyas et al., “Inkjet printed, self powered, wireless sensors for environmental, gas, and authentication-based sensing,” IEEE Sens. J., vol. 11, no. 12, pp. 3139–3152, 2011[7] STMicroelectronics, “M24LR04E-R: Dynamic NFC/RFID tag IC with 4-Kbit EEPROM, energy harvesting, I2C bus and ISO 15693 RF interface,” 2017. Accessed: Jan. 23, 2026. [Online]. Available: https://www.st.com/resource/en/datasheet/m24lr04er.pdf[8] STMicroelectronics, “Application Note - How to design a 13.56 MHz customized antenna for ST25 NFC / RFID Tags,” no. AN2866, Rev. 5, Aug. 2021, Accessed: Jan. 23, 2026. [Online]. Available: https://www.st.com/resource/en/application_note/an2866.pdf[9] STMicroelectronics, “NFC Inductance Calculator - eDesignSuite.” Accessed: Jan. 23, 2026. [Online]. Available: https://eds.st.com/antenna/#/[10] C. Buga and J. C. Viana, “Inkjet Printing of Functional Inks for Smart Products,” in Production Engineering and Robust Control, IntechOpen, 2022. doi: 10.5772/intechopen.104529.[11] Ralph Jacobi and Eddie LaCost, “Antenna Design Guide for the TRF79xxA,” 2017. Accessed: Jan. 15, 2026. [Online]. Available: https://www.ti.com/lit/an/sloa241b/sloa241b.pdf[12] NXP Semiconductors, “AN13219 PN7160 antenna design and matching guide Rev. 1.5-27 February 2024 Application note Document information Information Content PN7160 antenna design and matching guide,” 2024. Accessed: Jan. 15, 2026. [Online]. Available: https://www.nxp.com/docs/en/application-note/AN13219.pdf
Proceedings of the Technical Sessions, 42 (2026) 106-113Institute of Physics, Sri Lanka 106Investigation of Micro-cracks in Inkjet-Printed Silver Traces on Polyethylene Terephthalate Substrates for Flex SensorsInvestigation of Micro-cracks in Inkjet-Printed Silver Traces onPolyethylene Terephthalate Substrates for Flex SensorsW.A.C. Perera, S. W. D. K. R. M. Manamendra, W. L. P. K. Wijesinghe, G. C. Wickramasinghe, D. L. Weerawarne*Department of Physics, Faculty of Science, University of Colombo, Sri Lanka [email protected]. ABSTRACTFlexible electronics represent a transformative advancement in modern technology, enabling the integration of functional electronic components into pliable substrates that can bend, stretch, and conform to complex surfaces. This study explores the differences between disordered and ordered micro-crack arrangements in flex sensors for soft robotics, manufactured via an easy, affordable process. Conversely, creating ordered cracks generally requires more intricate, expensive techniques. Using inkjet-printed silver nanoparticle films on polyethylene terephthalate (PET) substrates, we systematically evaluate the effect of dimensional, mechanical, and thermal parameters, such as conductive path width, applied load, and sintering conditions affect sensor performance. Our findings reveal that wider conductive paths and moderate applied loads optimize sensor sensitivity and reliability, while controlled sintering at 100°C for two hours ensures consistent electrical behavior. Although disordered cracks demonstrate strong potential, their inherent variability currently limits their ability to fully replace ordered configurations, underscoring the need for further refinement in fabrication techniques to achieve scalable, reproducible sensor solutions for real-world applications.Keywords: Silver Inkjet Printed, Polyethylene Terephthalate, Disordered Crack, Flex sensors2. INTRODUCTIONSoft robotics is a rapidly developing field that focuses on the designs of robotic systems composed of soft and compliant materials, enabling safe interaction with humans and adaptability to complex and unstructured environments. Unlike conventional rigid robots, soft robots can undergo large deformations such as stretching, bending, and twisting, making them suitable for wearable devices, human–robot interaction, and adaptive manipulation tasks [1][2]To enable effective control and perception, soft robotic systems rely on strain (flex) sensors capable of real-time monitoring of deformation, motion, and force. These sensors are essential for shape sensing, proprioception, and tactile feedback. Flexible strain sensors often operate based on the piezoresistive effect, where applied strain alters the conductive network within the sensing material, producing measurable changes in electrical resistance. Common sensing mechanisms include resistance change due to material piezo resistivity, disconnection of conductive pathways, tunnelling effects, and crack propagation in thin films [3] .Micro-crack-based sensing mechanisms have been widely investigated to enhance strain sensor performance. As strain increases, cracks open and propagate, sharply reducing conductive paths and resulting in large relative changes in resistance. This mechanism can produce ultrahigh sensitivity, wide working ranges, and excellent motion detection capabilities. [4][5]Microcrack structures can be broadly categorized as ordered or disordered. Ordered microcracks exhibit controlled orientation, spacing, and geometry, which produce uniform and predictable sensor responses, and can be fabricated via advanced patterning techniques or lithographically defined gaps; however, these processes are often complex, expensive, and difficult to scale for large-area flexible electronics.
Proceedings of the Technical Sessions, 42 (2026) 106-113Institute of Physics, Sri Lanka 107Investigation of Micro-cracks in Inkjet-Printed Silver Traces on Polyethylene Terephthalate Substrates for Flex SensorsIn contrast, disordered microcracks form randomly within the conductive layer through simpler fabrication approaches such as pre-stretching, microgroove templates, solvent evaporation, or mechanical deformation. The generation of microcracks significantly improves sensitivity by amplifying strain effects through the disconnection and reconnection of conductive networks, demonstrating their effectiveness for flexible strain sensing in soft robotics and wearable platforms. [6]. This study explored the electrical properties and fabrication potential of disordered micro-cracks using inkjet-printed silver nanoparticles on PET substrates for flexible electronics.3. METHODOLOGYi. Selecting the materials and fabrication The first step involves selecting materials and fabrication. Polyethylene terephthalate (PET) was chosen for its flexibility and stability. Silver (Ag) nanoparticle ink (Novacentrix Metalon JS-B25P) was selected for its electrical conductivity and applied via inkjet printing on an EPSON L-130 printer. ii. Designing Silver traces The second step involves designing the conductive traces in Figure 1 with a length of 6 cm and widths of 3 mm, 5 mm, 7 mm, and 10 mm to fit human fingers.Figure 1: Designs for silver traces for Inkjet printing iii. Inkjet Printing and SinteringAn EPSON L130 desktop document printer was used to print the silver traces. After printing, the traces were oven-sintered at 100°C and 50°C for four different time periods: one hour, two hours, three hours, and four hours. iv. Micro Crack FabricationMicrocracks were generated using a scientific method known as the crease test [7]. Weights of 20 g, 100 g, 500 g, 1 kg, and 2 kg were applied to induce cracks in the silver trace. The process involved placing the printed trace on a flat surface, bending it halfway with the printed side inward, and then rolling the selected weight over the trace without applying any additional hand pressure.
Proceedings of the Technical Sessions, 42 (2026) 106-113Institute of Physics, Sri Lanka 108Investigation of Micro-cracks in Inkjet-Printed Silver Traces on Polyethylene Terephthalate Substrates for Flex Sensorsv. CharacterizationsFigure 1: An automated angle-bending machine for the printed silver traces on a substrate.The process indicated in Figure 2 was controlled using a Servo Motor and an Arduino, ensuring precise and repeatable movements. The moving slide advances the printed path from 0° to 140° at 5 s intervals. To monitor resistance variations over time, a Keithley 2450 source meter was connected to the end of the conductive path. Hence, resistance measurements were made using a Keithley meter employing the 4-probe method before and after crack formation. The automated anglebending machine performed cycling tests at 5 s intervals to assess changes in resistance over time. Crack morphology was analyzed using a digital microscope. Data analysis focused on calculating percentage changes in resistance and plotting variations across bending cycles to explore the relationship between micro-crack formation and electrical behavior in flexible conductive systems.4. RESULTS AND DISCUSSIONFigure 3: Printed samples using silver on PET with a 10 mm width and 6 cm length.Figure 3 shows the printed silver traces on a PET substrate used in this study. Each sample has a length of 6 cm and a width of 10 mm, and the images depict the traces before the integration of micro-cracks. These samples were fabricated using inkjet printing with a conductive silver ink, resulting in uniform and well-defined rectangular traces. The smooth surface and consistent dimensions confirm successful deposition on the flexible PET film, which serves as the base for subsequent mechanical or electrical testing. Analysis of Applied Load Variation.
Proceedings of the Technical Sessions, 42 (2026) 106-113Institute of Physics, Sri Lanka 109Investigation of Micro-cracks in Inkjet-Printed Silver Traces on Polyethylene Terephthalate Substrates for Flex SensorsFigure 4: Variation of the resistance with time, for the sample (6 cm x 10 mm) creased using different weights. Sintered for 4 hours and 100 oC a.) sample creased using 20 g, b.). sample creased using 100 g, c.) sample creased using 500 g, d.) sample creased using 1 kg, and e.) sample creased using 2 kg.Figure 4 shows the results obtained for the sample creased using different weights. Load-optimization studies identified 1 kg as the optimal creasing weight, effectively balancing crack density with structural integrity. Both lower loads (100 g, 500 g) and higher loads (2 kg) resulted in diminished performance due to insufficient or excessive cracking. The reason for choosing the 10 mm for the initial investigation was to match the width of the human finger. i. Analysis of Dimensional variation.Figure 5: Variation of the resistance with time for the sample creased using 1 kg for different lengths: a.) 6 cm x 3 mm, b.) 6cm x 5mm, c.) 6 cm x 7 mm, and d.) 6 cm x 10 m.
Proceedings of the Technical Sessions, 42 (2026) 106-113Institute of Physics, Sri Lanka 110Investigation of Micro-cracks in Inkjet-Printed Silver Traces on Polyethylene Terephthalate Substrates for Flex SensorsInvestigations in Figure 5 revealed the significant relationships between sensor design parameters and performance in disordered micro-crack configurations. Dimensional analysis demonstrated that wider conductive paths (10 mm) substantially outperformed narrower variants (3 mm, 5 mm, 7 mm, 10 mm) by facilitating multiple stable conductive routes, thereby enhancing sensitivity and reliability.ii. Analysis of sintering temperature dependency.Figure 6: 6 cm x 10 mm samples sintered at 50 0C and 100 0C for four hours and creased using 1 kg.The resistance–time response of the silver traces sintered at 100 °C and 50 °C is shown in the figure. The 100 °C sample exhibits a stable and repeatable resistance variation of around 2–3 Ω, while the 50 °C sample shows very high resistance with significant noise and poor repeatability, indicating inferior electrical performance at the lower sintering temperature.iii. Analysis of Sintering Time DependencyFigure 7: Variation of the resistance with time for the different sintering times for three samples 6 cm x 10 mm sample, and a creased sample using 1 kg, respectively.
Proceedings of the Technical Sessions, 42 (2026) 106-113Institute of Physics, Sri Lanka 111Investigation of Micro-cracks in Inkjet-Printed Silver Traces on Polyethylene Terephthalate Substrates for Flex SensorsIn addition, the current study identified the optimal sintering time to propagate microcracks. The results were observed four different times: one hour, two hours, three hours, and four hours. Figure 7 shows the effects of sintering temperature for Ag inkjet printer trace on PET for the formation of the microcrack pattern during angle bending and the conductivity of the silver trace with micro-cracks. According to these Figures, samples sintered for 2 hours were able to produce a more repetitive pattern compared to other time periods. Longer sintering times lead to greater evaporation, which can increase crack formation. This can be controlled by choosing the correct sintering time to achieve the optimal crack density for better results.iv. Digital microscopic analysis of the Results.Figure 8: a.) Photograph of creased Inkjet printed silver trace on PET, b) Digital microscopic image of cracks c) Circuit for resistance variation for formation of micro cracks on a sample creased using 1 kg, sintered at 100 °C for four hours.Digital microscopic analysis confirmed the random nature of crack formation, explaining performance variability. The crack pattern observed here resembles the “network pattern” on inkjet-printed Ag on PET in Figure 8 (b) reported by [8]. Such network cracks enable a wider operational range by preserving conductive pathways through bridges or by distributing stress via out-of-plane deflection. Figure 8 (c) illustrates the resistance variations according to the pattern of formation of microcracks on Ag silver traces. However, their slower resistance increase results in lower sensitivity.v. Repeatability of the ResultsTo check whether the result can be reproducible, we compared three samples that produced microcracks under the same conditions. However, despite maintaining identical fabrication parameters, the microcrack patterns obtained from the three samples exhibited significant variability, indicating a lack of repeatability under the tested conditions.
Proceedings of the Technical Sessions, 42 (2026) 106-113Institute of Physics, Sri Lanka 112Investigation of Micro-cracks in Inkjet-Printed Silver Traces on Polyethylene Terephthalate Substrates for Flex SensorsFigure 11: Variation of the resistance with time for three samples (6cm x 10 mm) sintered at 100 °C for 2 hours and creased using 1kg.5. CONCLUSIONWider conductive paths enhanced sensor performance by providing stable conduction routes, while narrower paths limited effectiveness. Optimal results were achieved at moderate loads (approximately 1 kg), whereas extreme loads negatively affected sensor responsiveness. Although higher crack density generally improves performance, irregular crack geometry can sometimes allow smaller or lightly loaded samples to outperform larger ones, demonstrating the influence of unpredictable crack patterns. Larger samples (10 mm × 6 cm) yielded more accurate results due to their greater capacity for conducting current. However, the randomness of crack formation hinders reproducibility, as similar conditions may yield different results based on crack connectivity. Sintering for 2 hours at 100 °C provided the best balance between bonding and crack formation, whereas lower temperatures or longer sintering times resulted in poorer sensor characteristics. In conclusion, while disordered cracks from creasing and stretching show promise under certain conditions, their unpredictability limits their ability to replace ordered cracks in strain sensors inkjet printed with silver on a PET substrate. Future research should aim to enhance the control and reproducibility of crack formation to enable reliable, costeffective flex sensor manufacturing.6. REFERENCES[1] M. Iqra, F. Anwar, R. Jan, and M. A. Mohammad, “A flexible piezoresistive strain sensor based on laser scribed graphene oxide on polydimethylsiloxane,” Sci. Rep., vol. 12, no. 1, Dec. 2022, doi: 10.1038/s41598-022-08801-0.[2] M. Amjadi, K.-U. Kyung, I. Park, and M. Sitti, “Stretchable, Skin-Mountable, and Wearable Strain Sensors and Their Potential Applications: A Review,” Adv. Funct. Mater., vol. 26, no. 11, pp. 1678–1698, 2016, doi: https://doi.org/10.1002/adfm.201504755.[3] F. Han, M. Li, H. Ye, and G. Zhang, “Materials, electrical performance, mechanisms, applications, and manufacturing approaches for flexible strain sensors,” May 01, 2021, MDPI AG. doi: 10.3390/nano11051220.[4] N. Kim, D. Yun, I. Hwang, G. Yoon, S. M. Kang, and Y. W. Choi, “Crack-Based Sensor with Microstructures for Strain and Pressure Sensing,” Sensors, vol. 23, no. 12, Jun. 2023, doi: 10.3390/s23125545.
Proceedings of the Technical Sessions, 42 (2026) 106-113Institute of Physics, Sri Lanka 113Investigation of Micro-cracks in Inkjet-Printed Silver Traces on Polyethylene Terephthalate Substrates for Flex Sensors[5] H. Sun et al., “An ultrasensitive and stretchable strain sensor based on a microcrack structure for motion monitoring,” Microsyst. Nanoeng., vol. 8, no. 1, Dec. 2022, doi: 10.1038/s41378-022-00419-6.[6] K. Gao, Z. Zhang, S. Weng, H. Zhu, H. Yu, and T. Peng, “Review of Flexible Piezoresistive Strain Sensors in Civil Structural Health Monitoring,” Applied Sciences (Switzerland), vol. 12, no. 19, Oct. 2022, doi: 10.3390/app12199750.[7] G. C. Wikaramasinghe, R. M. Manamendra, K. R. J. Manuda, B. Nissanka, D. L. Weerawarne, and D. R. Jayasundara, “Binder-free conductive graphite coatings on polyimide substrates for applications in flexible electronics,” J. Coat. Technol. Res., vol. 22, no. 2, pp. 825–838, 2025, doi: 10.1007/s11998-024-01012-4.[8] C. Song et al., “Advances in Crack-Based Strain Sensors on Stretchable Polymeric Substrates: Crack Mechanisms, Geometrical Factors, and Functional Structures,” Apr. 01, 2025, Multidisciplinary Digital Publishing Institute (MDPI). doi: 10.3390/polym17070941.
Proceedings of the Technical Sessions, 42 (2026) 114-121Institute of Physics, Sri Lanka 114Diurnal and Seasonal Asymmetry in Urban Thermal Discomfort in Colombo, Sri Lanka: Evidence from Daytime and Nighttime THI Analysis (1988–2018)Diurnal and Seasonal Asymmetry in Urban Thermal Discomfort in Colombo, Sri Lanka: Evidence from Daytime and Nighttime THI Analysis (1988–2018)Saumya Chathuranga*and K.P.S. Chandana JayaratneAstronomy and Space Science Unit, Department of Physics, Faculty of Science, University of Colombo, Colombo 00300, Sri [email protected]. ABSTRACTThis study examines long-term trends and seasonal variability in thermal discomfort across the Colombo Metropolitan Area (CMA), Sri Lanka, using the Temperature-Humidity Index (THI) from 1988 to 2018. Particular emphasis is placed on the diurnal Temperature-Humidity Index difference ΔTHI, quantifying the asymmetry between nighttime and daytime conditions. Results indicate a significant increase in thermal discomfort across all seasons for both day and night. Notably, nighttime THI trends are consistently steeper than daytime trends, particularly during the March-April-May (MAM) and June-July-August (JJA) seasons, providing evidence of an intensifying nocturnal Urban Heat Island (UHI) effect. While absolute warming is significant, long-term trends in ΔTHI remain statistically insignificant (p > 0.05), indicating stability in the relative day-night thermal contrast. Analysis of ΔTHI anomalies reveals that apparent extreme deviations in the mid-1990s and 2005 were primarily data artifacts resulting from incomplete seasonal records, rather than sustained directional change. The findings highlight growing public health risks in Colombo due to enhanced nocturnal heat stress and reduced nighttime thermal recovery.Keywords: Thermal Discomfort, Temperature–Humidity Index, Colombo2. INTRODUCTIONThe combination of rapid urbanization and climate change has increased thermal discomfort globally, especially in tropical coastal cities experiencing its impacts significantly. In such environments, both high humidity and high air temperature notably increase perceived heat stress during nighttime when radiative cooling is restricted [1], [2]. The most industrialized and commercialized city of Sri Lanka, Colombo Metropolitan Area (CMA), is an example of these issues because of drastic Land Use Land Cover (LULC), compacted population density, and enduring warm, humid weather.High ambient temperatures and consistently high relative humidity are highlighted as climate factors in the Sri Lankan lowland coastal region. Previous studies pointed out that [3] rapid urban growth has significant changes in surface energy balances, intensifying heat retention and reducing nocturnal cooling in the CMA. Furthermore, urban architecture such as building density, street orientation, and limited ventilation corridors causes outdoor thermal discomfort, especially during the nocturnal period [4].The Temperature–Humidity Index (THI), initially created for agricultural and physiological purposes [5], is currently used for tropical metropolitan settings where humidity significantly affects temperature perception. In cities like Colombo, THI measurements have demonstrated that urbanized areas often remain in a state of \"thermal discomfort\" even during evening hours due to the combined effect of high humidity and trapped urban heat [3]. It measures human
Proceedings of the Technical Sessions, 42 (2026) 114-121Institute of Physics, Sri Lanka 115Diurnal and Seasonal Asymmetry in Urban Thermal Discomfort in Colombo, Sri Lanka: Evidence from Daytime and Nighttime THI Analysis (1988–2018)heat stress by combining air temperature and humidity. Recent studies provide evidence for an urban warming tendency through a remarkable increment in THI over the past several decades [6], [7].Surface materials, vegetation cover, and building orientation are critical factors that amplify microclimates in urban centers. Evapotranspiration and shading are significantly improved by green cover, which mitigates heat stress by reducing ambient temperatures [7]. However, the prevalence of impervious surfaces enhances heat storage and nighttime thermal radiation, leading to intensified urban heat islands [8]. Local research further emphasizes that the choice of roofing materials and building ventilation are vital in determining the level of thermal exposure for residents in tropical climates like Sri Lanka [7], [8].The Urban Heat Island (UHI) intensity, which quantifies the Urban Heat Island (UHI) effect. Due to delayed heat release from the built environment materials, nocturnal UHI typically exceeds daytime UHI [1]. Even though several studies have documented Land Surface temperature change with respect to LULC dynamics in the CMA, poorly examined long-term human-centric heat stress indices, such as THI. Hence, this research briefly highlights the following gaps:• Quantification of long-term seasonal trends in daytime and nighttime THI in the CMA (1988–2018)• Assessing the magnitude and statistical significance of the mean THI difference (ΔTHI) between day and night trends• Identifying extreme ΔTHI anomalies and evaluating their climatic and surface-related drivers.3. METHODOLOGY3.1 Data and ProcessingMeteorological data (air temperature and relative humidity) were collected from the Department of metrology Sri Lanka, and the National Oceanic and Atmospheric Administration to cover the three-decade (1988–2018) period in the CMA. After separatinginto day and night datasets, they were aggregated annually and seasonally.3.2 THI (Temperature–Humidity Index) CalculationTHI is known as one of the bioclimatic indices attempting to measure human comfort based on temperature and relative humidity. However, the original index combined the wet and dry bulb temperatures to produce the THI [9] modified version of the index was proposed as equation (1). ??? = 0.8? +??×?500 (1)where T is the temperature, and RH% is the relative humidity.
Proceedings of the Technical Sessions, 42 (2026) 114-121Institute of Physics, Sri Lanka 116Diurnal and Seasonal Asymmetry in Urban Thermal Discomfort in Colombo, Sri Lanka: Evidence from Daytime and Nighttime THI Analysis (1988–2018)3.3. Seasonal Grouping and AggregationThe data set was categorized into four meteorological seasons as follows:• “DJF” represents the December, January, and February monsoon season (Northeast Monsoon period)• “MAM” represents the March, April, and May monsoon season (First Inter-monsoon period)• “JJA” represents the June, July, and August monsoon season (Southwest Monsoon period)• “SON” represents the September, October, and November monsoon season (Second Inter-monsoon period)Seasonal mean THI and Standard Error of the mean were calculated for each year for further analysis.3.4. ∆THI and Anomaly CalculationDiurnal thermal discomfort asymmetry was quantified from equation (2).∆??? = ??????ℎ?̅̅̅̅̅̅̅̅̅̅̅− ??????̅̅̅̅̅̅̅̅̅̅ (2)Where ??????ℎ?̅̅̅̅̅̅̅̅̅̅̅ and ??????̅̅̅̅̅̅̅̅̅̅ provide both night and day seasonal average THI values,respectively. This equation follows established approaches used to characterize diurnal asymmetry in urban thermal environments [1], [2]. Seasonal ΔTHI anomalies were computed relative to the 30-year seasonal climatological mean to isolate interannual variability.3.5 Trend and Variability AnalysisSeasonal daytime THI, nighttime THI, and ΔTHI time series were subjected to linear regression to determine trends. To identify the multi-year persistence in anomalies, five-year running means were used. For that reason, this step was applied to highlight low-frequency variability and suppress short-term fluctuations.4. RESULTS AND DISCUSSION4.1 Absolute THI TrendsThe results show a clear trend of asymmetric warming, where the Nighttime THI (Figure 2)slopes are larger than the Daytime THI (Figure 1) slopes across all seasons except marginally for SON (September-October-November) season, confirming that the nocturnal UHI is strengthening faster than the diurnal UHI. The JJA (June-July-August) and MAM (MarchApril-May) seasons, which correspond to the hottest and most humid periods in Sri Lanka,
Proceedings of the Technical Sessions, 42 (2026) 114-121Institute of Physics, Sri Lanka 117Diurnal and Seasonal Asymmetry in Urban Thermal Discomfort in Colombo, Sri Lanka: Evidence from Daytime and Nighttime THI Analysis (1988–2018)exhibit the steepest and most significant warming trends, posing the greatest threat to thermal comfort.Figure 1: Seasonal Variation and trends of daytime THI over 30 years Figure 2: Seasonal Variation and trends of nighttime THI over 30 years The following Table 1 summarizes that all four seasons show statistically significant trends in thermal discomfort for both day and night, which supports the conclusion of widespread regional warming.
Proceedings of the Technical Sessions, 42 (2026) 114-121Institute of Physics, Sri Lanka 118Diurnal and Seasonal Asymmetry in Urban Thermal Discomfort in Colombo, Sri Lanka: Evidence from Daytime and Nighttime THI Analysis (1988–2018)Table 1: Linear regression summary of all four seasonsSeason Nighttime THI Trend(????? ?℃/????)Daytime THI Trend(????? ?℃/????)THI LevelContextDJF(December, January, and February monsoon season)0.064 (? = 8.42 × 10−4) 0.050 (? = 6.17 × 10−1)Moderate THIlevels, less steepwarming.MAM(March, April, and May monsoon season)0.134 (? = 4.57 × 10−5) 0.073 (? = 4.58 × 10−2)High THI levels,the highestNighttimewarming rate.JJA(June, July, and August monsoon season)0.141 (? = 1.43 × 10−4) 0.081 (? = 8.27 × 10−6)Highest THIlevels, steepestoverall warming.SON(September, October, and November monsoon season)0.102 (? = 6.30 × 10−2) 0.099 (? = 2.66 × 10−3)Significantwarming in bothperiods.4.2 ΔTHI Trends and StabilityAnalysis of the seasonal trends in ΔTHI (???ℎ????? ??? − ??????? ???) reveals (Figure 3) a critical result: none of the four seasonal linear trends are statistically significant, as indicated by consistently high p-values (e.g., DJF ? = 0.89 ; MAM ? = 0.237 ).This suggests that while both night and day THI may be increasing, the long-term relative imbalance between them is stable. The Daytime THI remains consistently higher than the Nighttime THI (as indicated by the negative values), but the rate of change of this differential is negligible. This points to a mature, systematic effect where the relative partitioning of heat between day and night has reached an equilibrium relative to the warming baseline.4.3 Extreme ΔTHI AnomaliesThe seasonal ΔTHI anomaly series (Figure 4) indicates moderate interannual variability in day–night thermal imbalance over the study period. Consistency of the incomplete seasonal data set, no extreme anomaly events were detected, and the magnitude of deviations remained within realistic climatic bounds.Across all four seasons (DJF, MAM, JJA, SON), linear trend analysis revealed no statistically significant long-term changes in ΔTHI anomalies (? > 0.05). The slope coefficients were close to zero, suggesting that the relative difference between nighttime and daytime thermal discomfort has remained broadly stable over the study period.
Proceedings of the Technical Sessions, 42 (2026) 114-121Institute of Physics, Sri Lanka 119Diurnal and Seasonal Asymmetry in Urban Thermal Discomfort in Colombo, Sri Lanka: Evidence from Daytime and Nighttime THI Analysis (1988–2018)Figure 3: Seasonal trends of ΔTHI over 30 yearsInterannual fluctuations were most noticeable during the late 1990s and early 2000s; however, these deviations were episodic rather than indicative of sustained directional change. The absence of significant monotonic trends implies that diurnal thermal imbalance in this tropical coastal city is governed primarily by short-term climatic variability rather than progressive long-term forcing.Apparent large deviations visible during the mid-1990s and around 2005 were initially identified as extreme anomalies. However, subsequent verification revealed that these spikes were associated with incomplete seasonal records rather than genuine climatic variability. After restricting the analysis to seasons with complete three-month observations, these extreme values disappeared, confirming that they represent data artifacts rather than physical events.Figure 4: Seasonal ΔTHI anomalies over 30 years
Proceedings of the Technical Sessions, 42 (2026) 114-121Institute of Physics, Sri Lanka 120Diurnal and Seasonal Asymmetry in Urban Thermal Discomfort in Colombo, Sri Lanka: Evidence from Daytime and Nighttime THI Analysis (1988–2018)Seasonal ΔTHI anomalies are characterized by pronounced multi-year oscillations rather than a clear monotonic long-term trend. The strongest positive departures occur in DJF (DecemberJanuary-February) during the early to mid-2000s, while JJA (June-July-August) exhibits marked negative anomalies in the mid-1990s. The 5-year running mean (Figure 5) further reveals sustained positive ΔTHI phases during the late 1990s and early 2000s. This trend is particularly evident during DJF (December-January-February) and MAM (March-April-May), suggesting periods of enhanced nocturnal thermal stress relative to daytime conditions.Figure 5: 5-year running mean of seasonal ΔTHI anomalies over 30 years 5. CONCLUSIONSThis study demonstrates that Thermal discomfort has increased significantly across all seasons in Colombo, with nighttime warming exceeding daytime warming in almost all four seasons.Long-term ΔTHI trends are statistically insignificant, indicating stability in the relative day–night thermal contrast. Extreme ΔTHI anomalies are driven by transient surface disturbances and climate variability rather than gradual warming. Persistent nocturnal thermal stress poses growing public health risks in tropical urban environments. ΔTHI emerges as a valuable indicator for assessing diurnal thermal discomfort and highlights the critical importance of nighttime conditions in urban heat risk assessments.
Proceedings of the Technical Sessions, 42 (2026) 114-121Institute of Physics, Sri Lanka 121Diurnal and Seasonal Asymmetry in Urban Thermal Discomfort in Colombo, Sri Lanka: Evidence from Daytime and Nighttime THI Analysis (1988–2018)6. REFERENCES[1] T.R. Oke, The energetic basis of the urban heat island, Quart. J. R. Met. Soc, 108 (1982) 551-551.[2] C. Heaviside, H. Macintyre, S. Vardoulakis, The Urban Heat Island: Implications for Health in a Changing Environment, Curr. Environ. Health Rep., 4 (2017) 296-305. Available from: https://doi.org/10.1007/s40572-017-0150-3 [Accessed 23rd February 2026][3] R. Emmanuel, Thermal comfort implications of urbanization in a warm-humid city: The Colombo Metropolitan Region (CMR), Sri Lanka, Build. Environ., 40 (2005) 1591-1601. Available from: https://doi.org/10.1016/j.buildenv.2004.12.004 [Accessed 23rd February 2026][4] E. Johansson, R. Emmanuel, The influence of urban design on outdoor thermal comfort in the hot, humid city of Colombo, Sri Lanka, Int. J. Biometeorol., 51 (2006) 119-133. Available from: https://doi.org/10.1007/s00484-006-0047-6 [Accessed 23rd February 2026][5] E.C. Thom, The Discomfort Index, Weatherwise, 12 (1959) 57-61. Available from: https://doi.org/10.1080/00431672.1959.9926960 [Accessed 23rd February 2026][6] M.K.N.D.T. Bandara, An Analysis of Temporal Changes of Human Comfort Index (HCI) in Colombo Metropolitan Area, Sri Lanka (1997-2022), Proc. Int. For. Environ. Symp., 29 (2025) 8196. Available from: https://doi.org/10.31357/FESYMPO.V29.8196 [Accessed 23rd February 2026][7] S. Simath, R. Emmanuel, Urban thermal comfort trends in Sri Lanka: the increasing overheating problem and its potential mitigation, Int. J. Biometeorol., 66 (2022) 1865-1876. Available from: https://doi.org/10.1007/s00484-022-02328-9 [Accessed 23rd February 2026][8] M. Ranagalage, R.C. Estoque, Y. Murayama, An urban heat island study of the Colombo Metropolitan Area, Sri Lanka, based on Landsat data (1997–2017), ISPRS Int. J. Geo-Inf., 6 (2017) 189-206.[9] S. Nieuwolt, Tropical Climatology: An Introduction to the Climates, John Wiley & Sons, Chichester UK, 1997.
Proceedings of the Technical Sessions, 42 (2026) 122-132Institute of Physics, Sri Lanka 122Assessing the Suitability of NASA POWER Temperature Products for Filling Climatic Data Gaps in Sri LankaAssessing the Suitability of NASA POWER Temperature Products for Filling Climatic Data Gaps in Sri LankaJ K H Madushan, K P S C Jayaratne, and A L K WijemannageAstronomy and Space Science Unit, Department of Physics, University of Colombo, Colombo - 03, Sri [email protected]. ABSTRACTThe NASA Prediction of Worldwide Energy Resources (NASA POWER) project provides accessible, satellite-derived daily meteorological datasets at a 0.5° × 0.625° spatial resolution, serving as a vital resource for environmental and agricultural modeling in regions with sparse ground observations. This study investigates the precision of NASA POWER’s daily temperature estimates by comparing them with calculated and observed ground temperatures from 11 stations across Sri Lanka over a 30-year span. Five statistical assessments were conducted across four parameters: Daily Maximum Temperature (????), Daily Minimum Temperature (????), Daily Mean Temperature (?????), and Daily Temperature Range (???). The statistical metrics used were the Coefficient of Determination (?2), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Bias Error (MBE), and Willmott Index of Agreement (d). Validation revealed a relative performance hierarchy, determined by a combined assessment of correlation, absolute errors, and systematic biases. While ????showing the highest correlation (mean ?2 = 0.447), though the average RMSE was substantial at 2.72 °C. In contrast, ???? exhibited a weaker correlation (mean ?2 = 0.393) with an average RMSE of 2.12 °C. Furthermore, the Mean Bias Error revealed significant regional systematic biases: while ???? showed the highest random error (mean RMSE =2.72 °C), DTR showed the largest systematic bias (mean MBE = -1.79 °C) alongside a large RMSE (2.68 °C). ????? showed moderate correlation (mean ?2 = 0.444) with a smaller positive bias (MBE = +0.37 °C). Overall, the results suggest that NASA POWER data contain significant random errors across all parameters and exhibit systematic biases, particularly struggling with ???? and ??? in high-elevation stations (e.g., Nuwara Eliya and Badulla). NASA POWER temperature data require parameter- and location-specific bias correction before use in applications demanding high accuracy in Sri Lanka.Keywords: NASA POWER, Temperature, Remote Sensing, Satellite Reanalysis2. INTRODUCTIONAccurate near-surface air temperature (≈2 m) data are fundamental for climate analysis, agriculture, hydrology, and environmental assessment[1–3]. Daily maximum temperature (????) is critical for estimating evapotranspiration and heat stress, minimum temperature (????) for cold stress and dew formation, and the Daily Temperature Range (DTR) for evaluating surface energy balance and boundary-layer processes[4,5]. In Sri Lanka, strong
Proceedings of the Technical Sessions, 42 (2026) 122-132Institute of Physics, Sri Lanka 123Assessing the Suitability of NASA POWER Temperature Products for Filling Climatic Data Gaps in Sri Lankamonsoon influences, coastal effects, and complex topography produce pronounced spatial variability, increasing the need for reliable temperature datasets[6–9].Although ground-based observations provide direct measurements, sparse station networks, data gaps, and operational limitations restrict their spatial and temporal representativeness[10]. As a result, satellite-based and reanalysis products such as NASA Prediction of Worldwide Energy Resources (NASA POWER) are widely used, offering daily ???? and ???? at 0.5° × 0.625° resolution and enabling derivation of ????? and DTR in datascarce regions[11].Previous validation studies reported generally strong agreement between NASA POWER temperature estimates and observations, though systematic biases persist, often with higher accuracy for ???? than ???? and errors influenced by elevation, seasonality, and coastal proximity[12–15]. In Sri Lanka, limited evaluations suggest that NASA POWER temperatures are suitable for gap-filling despite identifiable biases[16], but assessments remain spatially and temporally constrained and largely exclude ????? and DTR.This study addresses these gaps by evaluating NASA POWER daily ????, ????, ?????, and DTR against observations from 11 meteorological stations across Sri Lanka’s major climatic zones, using standard statistical metrics (?2, RMSE, MAE, MBE, and Willmott’s d) to assess performance and applicability.3. METHODOLOGY3.1 Study Area Sri Lanka is an island nation situated in the Indian Ocean, approximately between latitudes 5°55' N and 9°51' N and longitudes 79°42' E and 81°53' E. The country features a tropical monsoon climate characterized by distinct climatic zones influenced by two main monsoon periods (Southwest : May-September, Northeast : December-February) and two intermonsoonal periods [17]. Sri Lanka is classified into three principal climatic zones based primarily on annual average rainfall: the Wet Zone (>2500 mm), the Intermediate Zone (1750 mm - 2500 mm), and the Dry Zone (<1750 mm)[9]. Table 1: Weather stations used in this study, climatic zone, elevation, latitude, and longitudeClimate Zone Weather Station Elevation (m) Latitude (°N) Longitude (°E)Wet ZoneColombo 7 6.90 79.87Katugastota 417 7.33 80.63Katunayaka 8 7.17 79.88Nuwara Eliya 1894 6.97 80.77Ratmalana 5 6.82 79.88Intermediate Zone Badulla 670 6.98 81.05Dry ZoneAnuradhapura 92 8.35 80.38Batticaloa 8 7.72 81.70Mannar 4 8.98 79.92Pottuvil 4 6.88 81.83Puttalam 2 8.03 79.83
Proceedings of the Technical Sessions, 42 (2026) 122-132Institute of Physics, Sri Lanka 124Assessing the Suitability of NASA POWER Temperature Products for Filling Climatic Data Gaps in Sri Lanka3.2 DataDaily temperature data (1988–2018) were analyzed from 11 meteorological stations representing diverse elevations and geographical settings across three climatic zones (Figure 1; Table 1). Ground-based daily ???? and ???? provided by the Department of Meteorology, Sri Lanka, were compared with corresponding 2 m air temperatures retrieved from the NASA POWER data access viewer (v2.5.22; single-point method).Figure 1: Spatial distribution of ground weather stations used in this study along with climatic zonesAll datasets were processed to ensure temporal consistency. This involved merging the NASA and ground datasets based on a common 'Date' index and excluding all days with missing data in either the ground or POWER records. Subsequently, a rigorous quality control step was applied using the Interquartile Range (IQR) method to mitigate the influence of outliers on statistical metrics[18,19]. Data points that fell outside the range of ?1 −1.5 × ??? and ?3 + 1.5 × ??? for both the predicted and observed temperature series were excluded from the final correlation analysis for each parameter.
Proceedings of the Technical Sessions, 42 (2026) 122-132Institute of Physics, Sri Lanka 125Assessing the Suitability of NASA POWER Temperature Products for Filling Climatic Data Gaps in Sri LankaTo evaluate the statistical agreement between the NASA POWER estimates and the ground observations for the temperature parameters, five statistical metrics were employed. These metrics provide a comprehensive assessment of correlation, overall error magnitude, and systematic bias. The statistical tests and their standard formulas are listed in table 2.Table 2: Statistical tests, formulas and their ideal conditionsStatistical Test Formula Ideal Conditions Reference(s)Coefficient of Determination(?2)( )( )22 1211ni iiniiO PRO O==−= −− (1) Close to 1(Range: 0 – 1) [20]Root Mean Square Error (RMSE)( )21ni iiO PRMSEn=−= (2)Close to 0(Range: 0 – ∞) [21,22]Mean Bias Error (MBE)1( )ni iiP OMBEn=−= (3)0 indicates no bias (Range: −∞ – +∞) [23]Willmott Index of Agreement (d)2121( )1( )ni iini iiO PdP O O O==−= −− + −(4) Close to 1(Range: 0 – 1) [24,25]Mean Absolute Error (MAE)11ni iiMAE P On == − (5) Close to 0(Range: 0 – ∞) [26]Note: Oidenotes the observed (ground), Odenotes the average of observed, and Pidenotes the estimated/predicted (POWER) values.4. RESULTS The accuracy of the NASA POWER products was assessed against ground station observations across four parameters: ???? , ????, ????? and ???. The overall mean values for the five statistical metrics (?2, MAE, MBE, RMSE, and d) across the 11 stations are summarized in Table 3.Because the correlation metrics (?2and d) exhibited a narrow range of variation, the relative performance of each parameter was determined by evaluating the combined magnitude of absolute errors (RMSE, MAE) and systematic biases (MBE). Based on this combined assessment, ???? and ????? exhibited the strongest performance (?2 ≈0.447 and 0.444) with moderate errors (RMSE 2.72°C and 1.97°C), despite a negative bias in ???? (MBE = −0.59°C) and slight positive bias in ????? (+0.37°C). Conversely, ???? and DTR performed poorly. ???? showed the lowest correlation (R2 = 0.393) and systematic overestimation (MBE = +1.25°C), while DTR reflected compounded errors (MAE = 2.26°C) and strong underestimation (MBE = −1.79°C). As shown in Figures 6–10, ????? generally had the lowest error magnitudes, while DTR had the largest.As shown in figures 2,3,4,5 and 6 spatially, NASA POWER performed best at low-elevation Dry Zone stations (e.g., Batticaloa, Mannar, Anuradhapura) but struggled significantly at coastal (Colombo, Ratmalana) and high-elevation sites. Nuwara Eliya emerged as an extreme
Proceedings of the Technical Sessions, 42 (2026) 122-132Institute of Physics, Sri Lanka 126Assessing the Suitability of NASA POWER Temperature Products for Filling Climatic Data Gaps in Sri Lankaoutlier with minimal predictive reliability, recording RMSE > 6.9°C across all variables(except DTR), and massive overestimation of ????. While the dataset captures lowland variability reasonably well, it fails to accurately represent nocturnal cooling and diurnal amplitude in complex terrains.Table 3: Summary of statistical test results for daily temperature parameters (overall)Parameter ?2(Mean)MAE(Mean)(°C)RMSE(Mean)(°C)MBE(Mean)(°C)d(Mean) Performance Assessment???? 0.447 2.40 2.72 -0.59 0.63Highest relative correlation; moderate absolute errors with a negative bias.????? 0.444 1.75 1.97 0.38 0.68 Lowest absolute errors; slight positive bias.???? 0.393 1.87 2.12 1.25 0.66 Lowest correlation; strong systematic overestimation.??? 0.423 2.26 2.68 -1.79 0.68Highest absolute errors; severe systematic underestimation.Figure 2: R2comparison by temperature parameter across all 11 weather stations
Proceedings of the Technical Sessions, 42 (2026) 122-132Institute of Physics, Sri Lanka 127Assessing the Suitability of NASA POWER Temperature Products for Filling Climatic Data Gaps in Sri LankaFigure: 3 RMSE comparison by temperature parameter across all 11 weather stationsFigure 4: MAE comparison by temperature parameter across all 11 weather stations
Proceedings of the Technical Sessions, 42 (2026) 122-132Institute of Physics, Sri Lanka 128Assessing the Suitability of NASA POWER Temperature Products for Filling Climatic Data Gaps in Sri LankaFigure 5: Willmott's d comparison by temperature parameter across all 11 weather stationsFigure 6: MBE comparison by temperature parameter across all 11 weather stations
Proceedings of the Technical Sessions, 42 (2026) 122-132Institute of Physics, Sri Lanka 129Assessing the Suitability of NASA POWER Temperature Products for Filling Climatic Data Gaps in Sri Lanka5. DISCUSSIONThe analysis confirmed that the accuracy of satellite-derived temperature estimates is highly variable, driven by parameter type and local geography. Comparing raw NASA POWER data against ground observations revealed larger errors and stronger systematic biases whichcannot be neglected. Because correlation metrics (R2 and d) were narrowly clustered across all parameters, the hierarchy was defined primarily by error magnitudes and systematic biases. While ???? and ????? led slightly in correlation, the error magnitudes (RMSE, MAE) favored ?????, followed by ???? and ????, with the DTR exhibiting the highest overall error and largest systematic bias.???? showed the highest mean correlation (R2 ≈ 0.447), likely due to the dominance of solar radiation during daytime, which satellite sensors capture effectively. However, substantial random errors (RMSE ≈ 2.72 °C) and a negative bias (MBE ≈ -0.59°C) indicate a tendency to underestimate peak temperatures, particularly at coastal sites. In contrast, ???? presented the greatest modeling challenges. It recorded the lowest correlation (R2 = 0.393) and a systematic positive bias (MBE ≈ +1.25 °C), reflecting the limitations of reanalysis models in capturing nocturnal boundary-layer processes and terrestrial cooling.These opposing biases—underestimation of ???? and overestimation of ????—resulted in a severe compression of the DTR. DTR recorded the highest systematic error (MBE ≈ -1.79°C) and substantial random error (RMSE ≈ 2.68 °C), confirming that the dataset fails to capture the full diurnal amplitude of Sri Lanka's climate. When compared to international literature, where R2values typically exceed 0.80 for continental regions, the correlations in this study are notably lower, and RMSE values are significantly higher.Geographically, performance was markedly lower in complex terrain. High-elevation stations like Nuwara Eliya were outliers, with RMSE exceeding 7 °C for extreme variables and massive overestimation of ????. Conversely, flat Dry Zone regions (e.g., Mannar, Anuradhapura) showed superior agreement. While NASA POWER captures broad seasonal patterns, the prevalence of large random errors and systematic biases limits its daily reliability. For applications requiring high precision, such as agricultural modeling, stationspecific bias correction is essential.6. CONCLUSIONSIn this study, the accuracy of NASA POWER daily temperature products, ????, ????, ?????, and ??? was evaluated against daily ground-based observations from 11 meteorological stations across Sri Lanka using statistical indicators, including ?2, RMSE, MAE, MBE, and Willmott’s ?. The analysis provided key insights into the model’s performance and limitations after correcting the validation methodology to compare raw datasets directly. The results revealed a clear performance hierarchy among temperature parameters, determined by a combined evaluation of correlation, absolute error, and systematic bias. While ???? and ????? showed the highest correlations (?2 ≈ 0.45),their relative superiority is largely due to having lower bias and error magnitudes than the other parameters, though substantial absolute errors persisted across all variables, with mean RMSE values exceeding 1.90 ∘C for every parameter (???? ≈ 2.72 ∘C, ???? ≈ 2.12 ∘C, ????? ≈ 1.97 ∘C, and DTR ≈ 2.68 ∘C). Significant systematic biases were also identified. ???? exhibited a strong positive bias (overestimation) of approximately +1.25 ∘C. Conversely, ??? showed a pronounced
Proceedings of the Technical Sessions, 42 (2026) 122-132Institute of Physics, Sri Lanka 130Assessing the Suitability of NASA POWER Temperature Products for Filling Climatic Data Gaps in Sri Lankanegative bias (underestimation) of approximately −1.79 ∘C, primarily resulting from the combined effects of underestimating ????(mean MBE ≈ −0.59 ∘C) and overestimating ????. ????? displayed a smaller overall positive bias (mean MBE ≈ +0.37 ∘C). Geographical influences were also found to be critical. Stations located in high-elevation terrain, such as Badulla and Nuwara Eliya, showed the poorest agreement and largest errors. In contrast, flat, Dry Zone stations such as Mannar and Anuradhapura displayed relatively better correlations, though errors remained significant.The magnitude of these biases is notably larger than suggested by traditional regression-based validation approaches. Despite the easy accessibility of the data, for applications demanding high temporal and spatial precision, such as precision agriculture, hydrological forecasting, and local climate impact assessments, parameter-specific and location-specific bias correction is strongly recommended rather than using the data directly. Future work should extend validation to humidity and radiation parameters, assess temporal stability across monsoonal transitions, and explore machine learning-based correction frameworks to enhance the reliability of NASA POWER products for tropical island environments.7. REFERENCES [1] Jones PD, New M, Parker DE, Martin S, Rigor IG. Surface air temperature and its changes over the past 150 years. Rev Geophys 1999;37:173–99. https://doi.org/10.1029/1999RG900002.[2] Z Ustrnul, A Wypych DC. Air Temperature Change. Springer Clim., 2021, p. 45–68. https://doi.org/10.1007/978-3-030-70328-8_4.[3] Aweda FO, Samson TK. Relationship between Air Temperature and Rainfall Variability of Selected Stations in Sub-Sahara Africa. Iran J Energy Environ 2022;13:248–57. https://doi.org/10.5829/IJEE.2022.13.03.05.[4] Stone DA, Weaver AJ. Daily maximum and minimum temperature trends in a climate model. Geophys Res Lett 2002;29:70-1-70–4. https://doi.org/10.1029/2001GL014556.[5] Caesar J, Alexander L, Vose R. Large‐scale changes in observed daily maximum and minimum temperatures: Creation and analysis of a new gridded data set. J Geophys Res Atmos 2006;111. https://doi.org/10.1029/2005JD006280.[6] Chandrasekara S, Prasanna V, Kwon H-H. Monitoring Water Resources over the Kotmale Reservoir in Sri Lanka Using ENSO Phases. Adv Meteorol 2017;2017:1–9. https://doi.org/10.1155/2017/4025964.[7] Naveendrakumar G, Vithanage M, Kwon H-H, Iqbal MCM, Pathmarajah S, Obeysekera J. Five Decadal Trends in Averages and Extremes of Rainfall and Temperature in Sri Lanka. Adv Meteorol 2018;2018:1–13. https://doi.org/10.1155/2018/4217917.[8] Shelton S, Pushpawela B, Liyanage G. The long-term trend in the diurnal temperature range over Sri Lanka from 1985 to 2017 and its association with total cloud cover and rainfall. J Atmos Solar-Terrestrial Phys 2022;227:105810. https://doi.org/10.1016/j.jastp.2021.105810.[9] Madushan H, Jayaratne C, Wijemannage A. Evaluation of NASA POWER Daily Rainfall Products Against Ground Observations in Sri Lanka. Remote Sens Earth Syst
Proceedings of the Technical Sessions, 42 (2026) 122-132Institute of Physics, Sri Lanka 131Assessing the Suitability of NASA POWER Temperature Products for Filling Climatic Data Gaps in Sri LankaSci 2025;8:264–72. https://doi.org/10.1007/s41976-024-00173-5.[10] Sun Y-J, Wang J-F, Zhang R-H, Gillies RR, Xue Y, Bo Y-C. Air temperature retrieval from remote sensing data based on thermodynamics. Theor Appl Climatol 2005;80:37–48. https://doi.org/10.1007/s00704-004-0079-y.[11] Paul W, Stackhouse, Jr.,Bradley Macpherson, Madison Broddle, Chequel McNeil, A. Jason Barnett, Colleen Mikovitz TZ. Introduction to the POWER Project Prediction Of Worldwide Energy Resource ( POWER ) 2021:21.[12] Halimi AH, Karaca C, Büyüktaş D. Evaluation of NASA POWER Climatic Data against Ground-Based Observations in the Mediterranean and Continental Regions of Turkey. J Tekirdag Agric Fac 2023;20:104–14. https://doi.org/10.33462/jotaf.1073903.[13] Kheyruri Y, Sharafati A, Ahmadi Lavin J. Performance assessment of NASA POWER temperature product with different time scales in Iran. Acta Geophys 2023;72:1175–89. https://doi.org/10.1007/s11600-023-01186-2.[14] Kadhim Tayyeh H, Mohammed R. Analysis of NASA POWER reanalysis products to predict temperature and precipitation in Euphrates River basin. J Hydrol 2023;619:129327. https://doi.org/10.1016/j.jhydrol.2023.129327.[15] White JW, Hoogenboom G, Stackhouse PW, Hoell JM. Evaluation of NASA satelliteand assimilation model-derived long-term daily temperature data over the continental US. Agric For Meteorol 2008;148:1574–84. https://doi.org/10.1016/j.agrformet.2008.05.017.[16] Gunaratne M.D.N, De Silva S.H.N.P AR. Can NASA Power Climatic Data Fill the Gap of Climatic Data Required for Agriculture and Forest Ecosystems Modeling? Proc 26th Int For Environ Symp 2022:21.[17] Burt TP, Weerasinghe KDN. Rainfall distributions in Sri Lanka in time and space: An analysis based on daily rainfall data. Climate 2014;2:242–63. https://doi.org/10.3390/cli2040242.[18] Vinutha HP, Poornima B, Sagar BM. Detection of Outliers Using Interquartile Range Technique from Intrusion Dataset. Theor. Popul. Biol., vol. 701, Springer Singapore; 2018, p. 511–8. https://doi.org/10.1007/978-981-10-7563-6_53.[19] Tukey JW. Exploratory data analysis. 1977.[20] Di Bucchianico A. Coefficient of Determination ( R 2 ). Encycl. Stat. Qual. Reliab., Wiley; 2007, p. 2. https://doi.org/10.1002/9780470061572.eqr173.[21] Wang W, Lu Y. Analysis of the Mean Absolute Error (MAE) and the Root Mean Square Error (RMSE) in Assessing Rounding Model. IOP Conf Ser Mater Sci Eng 2018;324:012049. https://doi.org/10.1088/1757-899X/324/1/012049.[22] Rodrigues GC, Braga RP. Evaluation of nasa power reanalysis products to estimate daily weather variables in a hot summer mediterranean climate. Agronomy 2021;11:1–17. https://doi.org/10.3390/agronomy11061207.[23] Jacovides CP, Kontoyiannis H. Statistical procedures for the evaluation of evapotranspiration computing models. Agric Water Manag 1995;27:365–71. https://doi.org/10.1016/0378-3774(95)01152-9.[24] Willmott CJ. ON THE VALIDATION OF MODELS. Phys Geogr 1981;2:184–94. https://doi.org/10.1080/02723646.1981.10642213.[25] Willmott CJ, Robeson SM, Matsuura K. A refined index of model performance. Int J
Proceedings of the Technical Sessions, 42 (2026) 122-132Institute of Physics, Sri Lanka 132Assessing the Suitability of NASA POWER Temperature Products for Filling Climatic Data Gaps in Sri LankaClimatol 2012;32:2088–94. https://doi.org/10.1002/joc.2419.[26] Hodson TO. Root-mean-square error (RMSE) or mean absolute error (MAE): when to use them or not. Geosci Model Dev 2022;15:5481–7. https://doi.org/10.5194/gmd-15-5481-2022.8. ACKNOWLEDGEMENTS The data was obtained from the NASA POWER Project’s Data Access Viewer (DAV) (POWER single point daily) version v2.5.22 on 2025/10/25. This work is financially supported by the National Research Council (NRC) of Sri Lanka. Grant No 20–119.
Proceedings of the Technical Sessions, 42 (2026) 133-139Institute of Physics, Sri Lanka 133The Role of Physicists in Shaping the Future of Higher Education Reforms in Sri LankaThe Role of Physicists in Shaping the Future of Higher Education Reforms in Sri LankaPresidential Address 2026, Institute of Physics, Sri LankabyProf. C. Mahesh EdirisinghePresident, Institute of Physics, Sri Lanka1. IntroductionSri Lanka stands today at a critical crossroads in shaping the future of its education system.Education has always been one of the strongest pillars of our nation’s social progress and democratic development. The commitment to free education, introduced decades ago, opened the doors of opportunity to generations of Sri Lankans and played a transformative role in building our human capital.Yet today, as the world undergoes profound technological, economic, and social transformations, our tertiary education system must evolve to meet new realities. The demands of the knowledge economy, rapid technological advancement, and the aspirations of our young people require us to rethink how we organize, expand, and strengthen higher education in Sri Lanka. The development of a ‘National Policy on Higher Education', therefore, represents not merely a sectoral reform but a national priority aimed at empowering our youth, strengthening innovation, and positioning Sri Lanka as a competitive knowledge hub in the region.Several years ago, a bright student from a rural school in Sri Lanka achieved excellent results at the G.C.E. Advanced Level examination in the physical science stream. His teachers believed he had the potential to become a scientist. His family had only one dream: that he would enter a state university and pursue physics. However, despite qualifying for university admission, he could not secure a place due to the limited capacity of our university system. Eventually, he enrolled in a different program through an alternative pathway. Like many talented young Sri Lankans, he slowly lost his dream of becoming a physicist.Now imagine this story repeated tens of thousands of times every year across our country.Every year, roughly 160,000 students qualify for university admission through the Advanced Level examination, yet only about 40,000 to 42,000 students are admitted to state universities.Behind these numbers are not just statistics. They represent dreams postponed, talent underutilized, and potential that the nation may never fully benefit from. Some of those students could have become physicists, engineers, entrepreneurs, innovators, or policymakers who might contribute to solving the challenges facing our society.This reality compels us to ask an important question. Are we fully utilizing the intellectual potential of our nation? And more importantly, how should Sri Lanka reform its higher education system to unlock that potential? It is within this broader context that the role of physicists in shaping higher education reforms in Sri Lanka becomes particularly significant.
Proceedings of the Technical Sessions, 42 (2026) 133-139Institute of Physics, Sri Lanka 134The Role of Physicists in Shaping the Future of Higher Education Reforms in Sri Lanka2. Industrial Revolutions and the Transformation of KnowledgeTo understand the urgency of reforming higher education, it is useful to reflect on the broader historical forces that have shaped the evolution of knowledge, industry, and society. Over the past three centuries, the world has experienced a series of industrial revolutions that fundamentally transformed economies, technologies, and education systems.The First Industrial Revolution, which began in the late 18th century, was driven by steam power and mechanization. It marked the transition from agrarian economies to industrial production. Scientific understanding of mechanics, thermodynamics, and materials played a critical role in enabling these transformations. The Second Industrial Revolution, emerging in the late 19th and early 20th centuries, introduced electricity, mass production, and modern manufacturing systems. Advances in electromagnetism, materials science, and engineeringfields deeply rooted in physics enabled the development of electrical grids, communication systems, and industrial machinery that reshaped modern economies. The Third Industrial Revolution, often referred to as the digital revolution, began in the latter half of the 20th century. It was driven by semiconductors, computing technologies, and the internet. Quantum physics and solid-state physics laid the foundation for the invention of the transistor, integrated circuits, and the digital technologies that underpin today’s information society. Today, we stand in the midst of the Fourth Industrial Revolution, characterized by artificial intelligence, quantum technologies, advanced materials, robotics, biotechnology, and data-driven innovation. The boundaries between physical, digital, and biological worlds are rapidly converging. In this new era, knowledge itself has become the most valuable economic resource. These technological transformations are not merely industrial developments; they are deeply connected to the evolution of universities and research institutions. Each industrial revolution has required new forms of knowledge, new skills, and new models of education. Universities around the world have continually adapted their curricula, research priorities, and institutional structures to meet these changing demands. For Sri Lanka, this moment presents both a challenge and an opportunity. If our higher education system evolves in alignment with these global transformations, it can empower our youth to participate in the knowledge economy and drive national innovation. If it does not, we risk falling further behind in an increasingly competitive and technology-driven world. Physicists occupy a particularly important place in this transformation.Many of the foundational technologies that define modern industry, from semiconductors and lasers to medical imaging and quantum computing, emerged from advances in physics. As such, physicists are uniquely positioned to guide the scientific, technological, and educational directions necessary for the future. Therefore, as Sri Lanka considers reforms to its higher education system, it is essential that we recognize the strategic importance of fundamental sciences, particularly physics, in shaping the technological and intellectual foundations of the nation.3. Why Physicists Matter in Shaping Higher Education ReformThe historical trajectory of industrial revolutions reveals a clear pattern; transformative technological shifts are almost always rooted in advances in fundamental science. Among these sciences, physics has consistently served as the intellectual engine that powers technological progress. From the discovery of the laws of motion that enabled early mechanical engineering, to the understanding of electromagnetism that made modern electrical systems possible, and
Proceedings of the Technical Sessions, 42 (2026) 133-139Institute of Physics, Sri Lanka 135The Role of Physicists in Shaping the Future of Higher Education Reforms in Sri Lankafrom quantum mechanics that gave birth to the semiconductor industry, to modern advances in photonics, nanotechnology, and quantum computing, physics has repeatedly reshaped the technological landscape of the world.However, the contribution of physicists extends far beyond the development of technology itself. Physicists are trained to think analytically, approach complex systems systematically, and develop solutions based on evidence and fundamental principles. These qualities are precisely what are required when societies attempt to reform complex systems such as higher education. Higher education reform is not merely about administrative restructuring or policy adjustments. It requires a deeper understanding of how knowledge is created, transmitted, and applied in a rapidly changing world. In this process, the intellectual traditions of physics, rigorous reasoning, interdisciplinary thinking, and innovation driven by curiosity can provide valuable guidance.For Sri Lanka, the involvement of physicists in national conversations on higher education reform is therefore not incidental; it is essential. Physicists can contribute not only as scientists but also as educators, innovators, and policy thinkers who understand the deep relationship between scientific discovery, technological development, and societal progress. As Sri Lanka seeks to position itself within the knowledge-driven global economy, the voice of the scientific community, particularly physicists, must play a central role in shaping the vision, structure, and priorities of our higher education system.4. The Historical Strength of Sri Lanka’s Education SystemSri Lanka has long been admired for its commitment to free education. The education reforms introduced in the mid-20th century transformed access to schooling and created one of the most literate societies in Asia. Even today, Sri Lanka maintains a literacy rate above 92 percent, one of the highest in the region. This remarkable achievement enabled generations of students, from rural villages to urban centres, to enter universities and contribute to national development.However, while access to primary and secondary education expanded significantly, expansion in higher education capacity has been comparatively slower.Today, Sri Lanka’s tertiary education system consists of 17 state universities and several higher education institutes under the purview of the University Grants Commission, alongside five state-owned universities that operate outside the UGC framework. The system further includes the National Institute of Education, which is authorized by an Act of Parliament to award degrees, as well as a growing number of non-state higher education institutions granted degreeawarding status (32 institutions and 293 degree programs as of March 13, 2026), collectively contributing to the expansion and diversification of higher education opportunities in Sri Lanka.Sri Lanka’s current education pipeline clearly demonstrates the urgency of expanding and diversifying the tertiary education sector. In 2023, a total of 229,057 students sat for the G.C.E. Advanced Level examination, of whom 151,343 students (66%) qualified for university admission, while 77,714 students (34%) did not meet the required threshold. At an earlier stage of the pipeline, 311,321 students sat for the G.C.E. Ordinary Level examination in 2021, with 231,982 (75%) passing and 79,325 (25%) not qualifying for progression to Advanced Level studies. Despite this large and growing pool of students progressing through the education system, the state university sector collectively admits only about 40,000 to 45,000 students annually. Consequently, a significant number of qualified students seek alternative pathways
Proceedings of the Technical Sessions, 42 (2026) 133-139Institute of Physics, Sri Lanka 136The Role of Physicists in Shaping the Future of Higher Education Reforms in Sri Lankato higher education. Each year, it is estimated that nearly 25,000 to 30,000 Sri Lankan students leave the country to pursue higher education abroad, resulting in a considerable outflow of both financial resources and human capital. In addition, thousands of students enroll in non-stateand transnational education programs within the country.These trends clearly indicate that the demand for higher education in Sri Lanka far exceeds the capacity of the traditional university system, highlighting the need to expand and strengthen diverse tertiary education pathways, including technical and vocational education, non-state higher education institutions, and transnational education programs. This situation, therefore,presents both a critical challenge and a strategic opportunity for reform.5. Global Transformation of UniversitiesAround the world, universities are undergoing significant transformation. Three major forces are reshaping higher education globally. First, the rise of the knowledge economy, where economic growth is increasingly driven by knowledge, research, and innovation. Second, the rapid advancement of digital technologies, including artificial intelligence, data science, and advanced computing. Third, the need for interdisciplinary solutions to complex global challenges, such as climate change, energy sustainability, global health, and technological disruption. Universities are no longer simply institutions that transmit knowledge.They are increasingly expected to generate knowledge, innovate technologies, and contribute directly to economic and societal development. Countries that successfully align their higher education systems with these transformations become centres of innovation and economic growth. Sri Lanka cannot remain outside this global transformation. If we aspire to become a knowledge hub in the Indian Ocean region, our universities must evolve into research-driven, innovation-oriented institutions.6. Structural Challenges in Sri Lanka’s Higher Education SystemWhile Sri Lanka’s higher education system has many notable strengths and a proud legacy, it continues to face several structural challenges that must be addressed if the system is to support innovation-driven development and fully harness the nation’s scientific and intellectual potential.6.1 Limited AccessThe most visible challenge is limited access to university education. When nearly threequarters of qualified students cannot enter state universities, the system inevitably creates frustration and inequality in opportunity. Expanding access while maintaining quality is therefore a central challenge of higher education reform.6.2 Disciplinary ImbalanceAnother issue is the distribution of students across academic disciplines. In many countries, national development strategies emphasize STEM fields: science, technology, engineering, and mathematics. While Sri Lanka has produced outstanding graduates in many disciplines, there remains a need to strengthen scientific and technological capacity to support innovation-driven development. Within STEM itself, physics plays a foundational role. Physics is often described
Proceedings of the Technical Sessions, 42 (2026) 133-139Institute of Physics, Sri Lanka 137The Role of Physicists in Shaping the Future of Higher Education Reforms in Sri Lankaas the fundamental science that underpins modern technology. The technologies that shape our modern world, such as semiconductors, lasers, telecommunications, renewable energy systems, satellite technology, and quantum computing, all trace their origins to fundamental discoveries in physics. Strengthening physics education and research, therefore, strengthens the entire technological ecosystem of a nation.6.3 Research and Innovation CapacityAnother challenge relates to research intensity and innovation ecosystems. Modern universities are not only teaching institutions; they are also centres of discovery and innovation. Countries that invest in research universities often experience significant growth in technology development, industrial innovation, and high-value economic sectors. Sri Lanka’s universities have produced important research contributions, yet there remains considerable room to strengthen research funding, industry collaboration, and technology commercialization.7. Why Physicists Matter in Higher Education ReformAt this point, one might ask, why specifically emphasize the role of physicists in higher education reform? The answer lies in the intellectual culture of physics itself. Physics trains individuals to approach problems with precision, curiosity, and analytical rigor. Physicists are trained to understand complex systems, analyze patterns in data, and derive insights from empirical evidence. These skills are highly relevant not only in scientific research but also in policy design, institutional planning, and systemic reform. A few ways in which physicists can contribute are stated below:7.1 Evidence-Based Policy DesignPhysics is fundamentally an evidence-based discipline. Physicists rely on measurement, experimentation, quantitative analysis, etc. These same principles can strengthen policy decisions in higher education. Instead of relying solely on tradition or opinion, reforms should be guided by data on student outcomes, research productivity, graduate employability, and institutional performance. Physicists can contribute to designing analytical frameworks that support evidence-based decision-making in higher education governance.7.2 Strengthening STEM EcosystemsPhysics forms the conceptual foundation for many STEM disciplines. A strong physics ecosystem strengthens engineering, computing, materials science, and emerging technologies.Advances in quantum physics are driving new developments in computing and secure communication. Research in materials physics is enabling breakthroughs in renewable energy technologies. Developments in medical physics are transforming diagnostics and cancer treatment. Thus, investing in physics education and research strengthens multiple sectors of the knowledge economy.7.3 Interdisciplinary LeadershipPhysicists often work at the interface of disciplines. In modern research environments, physicists collaborate with engineers, chemists, computer scientists, and biologists. Many of today’s global challenges require interdisciplinary solutions. Universities, therefore, need academic leaders who are comfortable crossing disciplinary boundaries and integrating diverse
Proceedings of the Technical Sessions, 42 (2026) 133-139Institute of Physics, Sri Lanka 138The Role of Physicists in Shaping the Future of Higher Education Reforms in Sri Lankaperspectives. Physicists can play an important role in designing interdisciplinary academic programs and research initiatives.7.4 Innovation in Teaching and LearningPhysics education research has produced valuable insights into how students learn complex scientific concepts. Innovative approaches such as problem-based learning, inquiry-based laboratories, and computational simulations have improved student engagement and conceptual understanding. These pedagogical innovations can inform broader reforms in teaching and learning across universities.7.5 Systems Thinking in Institutional ReformPerhaps most importantly, physics cultivates systems thinking. Physicists study systems where multiple variables interact in dynamic ways. Universities themselves are complex systems, involving governance structures, funding models, research ecosystems, and student experiences. Reforming such systems requires holistic thinking and careful analysis of interactions between different components. Physicists are uniquely trained to think in these terms.8. The Way Forward for Sri LankaIf Sri Lanka is to build a higher education system capable of meeting the challenges of the 21stcentury, several strategic priorities must guide reform. First, expand access to higher education through a diversified ecosystem of public universities, non-state institutions, private institutions, and international collaborations. Second, strengthen research universities that can drive innovation and knowledge creation. Third, promote STEM education and interdisciplinary learning aligned with the needs of the modern economy. Fourth, encourage stronger university–industry partnerships that translate research into real-world impact. Finally, empower academics and scientists to play a greater role in policy dialogue, institutional leadership, and national development strategies. Higher education reform should not be viewed merely as an administrative exercise. It is fundamentally about shaping a nation's intellectual future.9. The Responsibility of the Physics CommunityPhysics teaches us that within every complex system there exists a critical threshold, a moment when gradual change leads to sudden transformation. In physics, we call this a phase transition.Water slowly heats, degree by degree, until suddenly it becomes steam. For a long time, change may appear slow and incremental. But when the right conditions emerge, a new state becomes possible. I believe Sri Lanka’s higher education system may be approaching such a moment.As members of the physics community represented by the Institute of Physics, Sri Lanka, we have a collective responsibility that goes beyond our laboratories and classrooms. We must contribute to national dialogue on science education, advocate for evidence-based policy, inspire the next generation of scientists, and help shape institutions that nurture discovery and innovation.There is a well-known saying that “Mathematics is the queen of the sciences.” And as history reminds us, even the most powerful king cannot rule wisely without a queen beside him. If we imagine physicists as the king in the great kingdom of knowledge, driven by curiosity to
Proceedings of the Technical Sessions, 42 (2026) 133-139Institute of Physics, Sri Lanka 139The Role of Physicists in Shaping the Future of Higher Education Reforms in Sri Lankaunderstand the universe, we must also recognize that no king rules alone. Our queens are many: mathematicians, chemists, biologists, engineers, data scientists, medical scientists, environmental scientists, economists, sociologists, psychologists, policy analysts, and educationalists. Without this court of knowledge, even the most brilliant king would struggle to govern. The message is clear: the future of higher education cannot be shaped by any single discipline, but through a vibrant multidisciplinary collaboration where diverse fields work together to create new knowledge and new possibilities.10. A Call for Scientific Leadership in Educational TransformationAt this pivotal moment in history, higher education systems across the world are being reshaped by rapid technological change, new forms of knowledge production, and evolving societal needs. Sri Lanka is no exception. The challenge before us is not simply to reform universities, but to reimagine them as dynamic institutions that drive innovation, nurture creativity, and empower future generations.In this transformation, the role of physicists and the broader scientific community becomes critically important. Physicists have long been explorers of the fundamental laws of nature, but in today’s world, they must also become architects of the intellectual and technological future of our society. By contributing to policy dialogue, curriculum innovation, interdisciplinary collaboration, and national research priorities, physicists can help guide Sri Lanka toward a higher education system that is resilient, forward-looking, and globally competitive.If the industrial revolutions of the past were powered by scientific discovery, the revolutions of the future will depend on how effectively nations mobilize their knowledge communities. The question before us, therefore, is not whether change will come but whether we are prepared to shape that change.11. A Call to Collective ResponsibilityWe have a proud tradition of education. We have talented students. We have dedicated academics. What we need now is vision, courage, and collaborative leadership to guide the next phase of transformation. If we succeed, the reforms we discuss today will not simply expand universities. They will expand opportunities for an entire generation. And perhaps somewhere in a classroom in Sri Lanka today, there is a young student who dreams of becoming a scientist. Our responsibility is to ensure that the system we build does not limit that dreambut enables it.As physicists, educators, and scholars, we carry both the responsibility and the opportunity to ensure that Sri Lanka’s higher education system evolves in ways that serve not only our universities but the broader aspirations of our nation.Let us work together to build a higher education system where no dream is limited by opportunity, and where the curiosity of every young mind can illuminate the future of our nation.— ∞ —