Gay Jane P. Perez


Doctor of Philosophy in Physics, University of the Philippines, 2009

Master of Science in Physics, University of the Philippines, 2005

Bachelor of Science in Applied Physics, University of the Philippines 2003

Field of Specialization






Article title: Determination of Cloud-top Height through Three-dimensional Cloud Reconstruction using DIWATA-1 Data

Authors: Castro Ellison, Ishida Tetsuro, Yukihiro Takahashi, Kubota Hisayuki, et al.

Publication title: Scientific Reports (Nature Publisher Group) 10(1), May 2020



Cloud-top height is a useful parameter with which to elucidate cloud vertical growth, which often indicates severe weather such as torrential rainfall and thunderstorms; it is widely used in meteorological research. However, general cloud-top height estimation methods are hindered by observational and analytical constraints. This study used data from DIWATA-1, the Philippines’ first microsatellite, to overcome these limitations and successfully produce sophisticated three-dimensional cloud models via stereo-photogrammetry. High-temporal snapshot 200-ms-interval imaging of clouds over Iloilo, Philippines, is performed. Two types of telescopes were used to capture 30 stereoscopic cloud images at ~60- and ~3-m ground sampling resolutions; these were used to construct three-dimensional cloud models with 40- and 2-m vertical resolutions, respectively. The imaged clouds have heights of 2.0 to 4.8 km, which is below freezing level for the Philippines and typical of stratocumulus and cumulus clouds. The results are validated using cloud-edge heights determined by measuring the distance from the clouds to their ground shadows. An RMSE of 0.32 km and a maximum difference of 0.03 km are found for the low- and high-resolution telescopes, respectively. For further validation, the results are compared with cloud-top heights estimated from HIMAWARI-8 images captured on the same day, yielding an average vertical difference of 0.15 km and a maximum difference of 1.7 km.


Article title: Reforestation and Deforestation in Northern Luzon, Philippines: Critical Issues as Observed from Space

Authors: Gay Jane Perez, Josefino C Comiso, Lemnuel V Aragones, Harry C Merida, et al.

Publication title: Forests 11(10), October 2020



Among the richest in biodiversity globally has been the Philippine rainforest, which used to cover about 90% of the country’s land area. During the last few decades, the forest cover has been reduced to less than 10% of the original, only a fraction of which is old-growth forest. The negative impacts of deforestation led to the launching of the National Greening Program (NGP) that involved the planting of more than a billion seedlings over a few million hectares of land from 2011 to 2016. To assess the success of the NGP, satellite data from Landsat and the Moderate Resolution Imaging Spectroradiometer (MODIS) were analyzed before, during, and after the NGP. Reforestation in the NGP sites was examined concurrently with observed deforestation in Luzon using forest loss data derived from Landsat for the period 2001 to 2018. The results show that losses declined from 2011 to 2015 but increased from 2016 to 2018. Because of such losses, the net effect is a balance of reforestation and deforestation or no significant gain from the NGP. Case studies were done in three sites in the Sierra Madre forest, where half of the remaining old-growth forest is located, using a combination of Landsat and Very High Resolution (VHR) data. The Landsat data were classified into closed forest, open forest, and other vegetation cover types. The conversion from one vegetation cover type to another was evaluated through the use of the Sankey Diagram. While some non-forest types became open or closed forests, the loss of open or closed forests is more pronounced. VHR data reveal critical issues happening within the NGP sites during the NGP period. More comprehensive data from MODIS also confirm that there was no significant increase in the forest cover in Luzon, Sierra Madre, and Cordillera from 2001 to 2018.


Article title: A novel approach for vegetation classification using UAV-based hyperspectral imaging

Authors: Tetsuro Ishida, Junichi Kurihara, Fra Angelico Malicdin, Gay Perez, et al.

Publication title: Computers and Electronics in Agriculture 144, January 2018



The use of unmanned aerial vehicle (UAV)-based spectral imaging offers considerable advantages in high-resolution remote-sensing applications. However, the number of sensors mountable on a UAV is limited, and selecting the optimal combination of spectral bands is complex but crucial for conventional UAV-based multispectral imaging systems. To overcome these limitations, we adopted a liquid crystal tunable filter (LCTF), which can transmit selected wavelengths without the need to exchange optical filters. For calibration and validation of the LCTF-based hyperspectral imaging system, a field campaign was conducted in the Philippines during March 28–April 3, 2016. In this campaign, UAV-based hyperspectral imaging was performed in several vegetated areas, and the spectral reflectances of 14 different ground objects were measured. Additionally, the machine learning (ML) approach using a support vector machine (SVM) model was applied to the obtained dataset, and a high-resolution classification map was then produced from the aerial hyperspectral images. The results clearly showed that a large amount of misclassification occurred in shaded areas due to the difference in spectral reflectance between sunlit and shaded areas. It was also found that the classification accuracy was drastically improved by training the SVM model with both sunlit and shaded spectral data. As a result, we achieved a classification accuracy of 94.5% in vegetated areas.


Article title: Positive Trend in the Antarctic Sea Ice Cover and Associated Changes in Surface Temperature

Authors: Josefino C. Comiso, Robert A. Gersten, Larry V. Stock, Gay Perez, et al

Publication title: Journal of Climate 30(6), March 2017 



The Antarctic sea ice extent has been slowly increasing contrary to expected trends due to global warming and results from coupled climate models. After a record high extent in 2012 the extent was even higher in 2014 when the magnitude exceeded 20 × 106 km2 for the first time during the satellite era. The positive trend is confirmed with newly reprocessed sea ice data that addressed inconsistency issues in the time series. The variability in sea ice extent and ice area was studied alongside surface ice temperature for the 34-yr period starting in 1981, and the results of the analysis show a strong correlation of −0.94 during the growth season and −0.86 during the melt season. The correlation coefficients are even stronger with a one-month lag in surface temperature at −0.96 during the growth season and −0.98 during the melt season, suggesting that the trend in sea ice cover is strongly influenced by the trend in surface temperature. The correlation with atmospheric circulation as represented by the southern annular mode (SAM) index appears to be relatively weak. A case study comparing the record high in 2014 with a relatively low ice extent in 2015 also shows strong sensitivity to changes in surface temperature. The results suggest that the positive trend is a consequence of the spatial variability of global trends in surface temperature and that the ability of current climate models to forecast sea ice trend can be improved through better performance in reproducing observed surface temperatures in the Antarctic region.

Full text available upon request to the author


Article title: Forest Cover Dynamics in the Philippines from LandSAT-Derived Global Forest Cover Dataset (2000-2012)

Authors: Brent Fallarcuna and Gay Perez

Publication title: Journal of the Philippine Geoscience and Remote Sensing Society, 2016



Tropical deforestation still pervades in the developing countries despite the conservation and development programs implemented by national agencies and international community. Its effects are mostly felt not only by forest dependent rural communities, but also by other people across the globe in line with the changing climate. In relation to this, up to date and reliable estimates of forest cover is necessary in order to guide the policy makers and forest managers in implementing relevant policies and programs for sustainable forest management and conservation. However, as other tropical countries like the Philippines, forest cover maps and statistics is hampered by multiple imagery sources, varying scale and resolutions, undisclosed methods of mapping and use of different forest classes and definitions. In this study, global and freely available Landsat processed data were used in order to quantify the rates of forest loss on the provincial level and to determine its level of accuracy in the Philippines. The Landsat processed data was enhanced by defining a threshold on percent tree cover that represents forest. However, by applying previously published forest maps, non-forest features were masked out. This data was then analyzed to produce net change ranking and annual forest loss trends and correlations. This was done per forest cover type to provide more detailed insights on forest cover change. Results showed that at least 53,620.20 hectares of forest was lost in the country from year 2000-2012. Overall accuracy indicates high reliability for both loss (86.48%, kappa statistic = 0.86) and gain (92.26%, kappa statistic = 0.92). Mindanao provinces such as Zamboanga Sibugay (-11.30%), Zamboanga del Norte (-5.42%) and Basilan (-4.06) topped the provinces with negative aggregate net changes. Provinces like Surigao del Sur (r = 0.77), Surigao del Norte (r = 0.65) and Ilocos Sur (r = 0.61) had the highest increasing trend of forest loss (p-value at ¡ 0.05). For open forest alone, all Zamboanga provinces showed high aggregate net change while forest cover loss correlations with positive trend were significant in Agusan del Sur (r = 0.70), Tarlac (r = 0.69) and Davao Oriental (r = 0.69). Among the forest cover types analyzed, the open forest (secondary) type exhibited the highest aggregate net change, implying that it was most dynamic and vulnerable to deforestation. This study showed that forest cover changes could be quantified consistently given a globally available platform such as the Landsat processed data.


Article title: Enhanced Pacific Ocean Sea Surface Temperature and Its Relation to Typhoon Haiyan

Authors: Gay Perez, Josefino C. Comiso, Larry V. Stock

Publication title: Journal of Environmental Science and Management 18(1), June 2015



Typhoon Haiyan, which devastated the Visayan Islands in the Philippines on November 8, 2013 was recorded as the strongest typhoon ever-observed using satellite data. Typhoons in the region usually originate from the mid-Pacific region that includes the Warm Pool, which is regarded as the warmest ocean surface region globally. Two study areas were considered: one in the Warm Pool Region and the other in the West Pacific Region near the Philippines. Among the most important factors that affect the strength of a typhoon are sea surface temperature (SST) and water vapor It is remarkable that in November 2013 the average SST in the Warm Pool Region was the highest observed during the 1981 to 2014 period while that of the West Pacific Region was among the highest as well. Moreover the increasing trend in SST was around 0.20 degrees C per decade in the warm pool region and even higher at 0.23 degrees C per decade in the West Pacific region. The yearly minimum SST has also been increasing suggesting that the temperature of the ocean mixed layer is also increasing. Further analysis indicated that water vapor, clouds, winds and sea level pressure for the same period did not reveal strong signals associated with the 2013 event. The SST is shown to be well-correlated with wind strength of historically strong typhoons in the country and the observed trends in SST suggest that extremely destructive typhoons like Haiyan are likely to occur in the future.

Full text available upon request to the author


Article title: Prior Individual Training and Self-Organized Queuing during Group Emergency Escape of Mice from Water Pool

Authors: Caesar Saloma, Gay Perez, Catherine M. Gavile, et al.

Publication title: PLoS ONE 10(2), February 2015



We study the impact of prior individual training during group emergency evacuation using mice that escape from an enclosed water pool to a dry platform via any of two possible exits. Experimenting with mice avoids serious ethical and legal issues that arise when dealing with unwitting human participants while minimizing concerns regarding the reliability of results obtained from simulated experiments using 'actors'. First, mice were trained separately and their individual escape times measured over several trials. Mice learned quickly to swim towards an exit-they achieved their fastest escape times within the first four trials. The trained mice were then placed together in the pool and allowed to escape. No two mice were permitted in the pool beforehand and only one could pass through an exit opening at any given time. At first trial, groups of trained mice escaped seven and five times faster than their corresponding control groups of untrained mice at pool occupancy rate ρ of 11.9% and 4%, respectively. Faster evacuation happened because trained mice: (a) had better recognition of the available pool space and took shorter escape routes to an exit, (b) were less likely to form arches that blocked an exit opening, and (c) utilized the two exits efficiently without preference. Trained groups achieved continuous egress without an apparent leader-coordinator (self-organized queuing)-a collective behavior not experienced during individual training. Queuing was unobserved in untrained groups where mice were prone to wall seeking, aimless swimming and/or blind copying that produced circuitous escape routes, biased exit use and clogging. The experiments also reveal that faster and less costly group training at ρ = 4%, yielded an average individual escape time that is comparable with individualized training. However, group training in a more crowded pool (ρ = 11.9%) produced a longer average individual escape time.


Article title: Allelomimesis as escape strategy of pedestrians in two-exit confinements

Authors: Gay Perez and Caesar Saloma

Publication title: Physica A: Statistical Mechanics and its Applications 388(12), June 2009



We study the efficacy of allelomimesis as an escape strategy of mobile agents (pedestrians) that aim to leave a two-exit room within the shortest possible time. Allelomimesis is the act of copying one’s kindred neighbors. To escape, an agent employs one of the following strategies: (1) It chooses its own route independently (non-copying, α=0), (2) It imitates the actions of its neighbors at all times (blind copying, α=1), or (3) It either copies or acts independently according to a certain probability that is set by the copying parameter α(0α1). Not more than one agent could occupy a given room location. An agent’s knowledge of the two exit locations is set by its information content β(0≤β≤1). When left alone, an agent with complete knowledge of the exit whereabouts (β=1) always takes the shortest possible path to an exit. We obtain plots of the (group) evacuation time T and exit throughput Q as functions of α and β for cases where the two exits are near (on same room side) and far (on opposite sides of room) from each other. For an isolated agent, T is inversely proportional to β. The chances of escape for an isolated agent with β≤0.2 are higher with adjacent exits. However, for an agent with β>0.4 the chance is better with opposite exits. In a highly occupied room (occupancy rate R=80%) with adjacent exits, agents with β>0.8 escape more quickly if they employ a mixed strategy of copying and non-copying (0.4α0.6). On the other hand, blind copying (α≈1) gives agents with β0.1 a better chance of escaping from the same room. For the same α and R values, opposite exits allow faster evacuation for agents with β0.1 due to the likelihood of streaming and the lower probability of exit clogging. Streaming indicates an efficient utilization of an exit and it happens when the arcs that are formed are smaller and arch interference is less likely. Allelomimesis provides a simple yet versatile mechanism for studying the egress behavior of confined crowds in a multi-exit room.

Full text available upon request to the author


Article title: Self-Organized Queuing and Scale-Free Behavior in Real Escape Panic

Authors: Caesar Saloma, Gay Perez, Giovanni Tapang, May T. Lim, et al.

Publication title: Proceedings of the National Academy of Sciences 100(21), November 2003



Numerical investigations of escape panic of confined pedestrians have revealed interesting dynamical features such as pedestrian arch formation around an exit, disruptive interference, self-organized queuing, and scale-free behavior. However, these predictions have remained unverified because escape panic experiments with real systems are difficult to perform. For mice escaping out of a water pool, we found that for a critical sampling rate the escape behavior exhibits the predicted features even at short observation times. The mice escaped via an exit in bursts of different sizes that obey exponential and (truncated) power-law distributions depending on exit width. Oversampling or undersampling the mouse escape rate prevents the observation of the predicted features. Real systems are normally subject to unavoidable constraints arising from occupancy rate, pedestrian exhaustion, and nonrigidity of pedestrian bodies. The effect of these constraints on the dynamics of real escape panic is also studied.

Article title: Streaming, disruptive interference and power-law behavior in the exit dynamics of confined pedestrians

Authors:  Gay Perez, Giovanni Tapang, May T. Lim, Caesar Saloma

Publication title: Physica A: Statistical Mechanics and its Applications. September 2002



We analyze the exit dynamics of pedestrians who are initially confined in a room. Pedestrians are modeled as cellular automata and compete to escape via a known exit at the soonest possible time. A pedestrian could move forward, backward, left or right within each iteration time depending on adjacent cell vacancy and in accordance with simple rules that determine the compulsion to move and physical capability relative to his neighbors. The arching signatures of jamming were observed and the pedestrians exited in bursts of various sizes. Power-law behavior is found in the burst-size frequency distribution for exit widths w greater than one cell dimension (w>1). The slope of the power-law curve varies with w from −1.3092(w=2) to −1.0720(w=20). Streaming which is a diffusive behavior, arises in large burst sizes and is more likely in a single-exit room with w=1 and leads to a counterintuitive result wherein an average exit throughput Q is obtained that is higher than with w=2,3, or 4. For a two-exit room (w=1), Q is not greater than twice the yield of a single-exit room. If the doors are not separated far enough (<4w), Q becomes even significantly less due to a collective slow-down that emerges among pedestrians crossing in each other's path (disruptive interference effect). For the same w and door number, Q is also higher with relaxed pedestrians than with anxious ones.

Full text available upon request to the author


Papers Presented


Title: Validation of the separability measure for Rhizophoraceae and Avicenniaceae using point density distribution from lidar

Authors: Regine Anne Faelga, Enrico C. Paringit, Gay Perez, et al.

Conference title: SPIE Asia-Pacific Remote Sensing, May 2016



The extent at which mangrove forest characterization can be done through utilization of Light Detection and Ranging (LiDAR) data is investigated in this paper. Particularly, the ability of LiDAR parameters, such as its point density to provide height and structural information was explored to supplement manual field surveys which are time-consuming and requires great effort. Point cloud information was used to produce separability measure within a mangrove forest. The study aims to validate the point density distribution curves (PDDC) that were established to characterize the structural attributes between Rhizophoraceae and Avicenniaceae. The applicability of the PDDC was applied to fifteen (15) 5x5 sample plots of pure Rhizophoraceae and fifteen (15) 5x5 sample plots of pure Avicenniaceae in a one hectare (1ha) natural riverine mangrove forest. 15 out of 15 plots were correctly discriminated as Rhizophoraceae; however, Avicenniaceae plots were not correctly discriminated using the established separability measure. This study had determined that the two mangrove families are difficult to separate in terms of point density distribution alone. Enhancement of the PDDC as a separability measure should be improved to pave way for a more sensitive and robust way to separate the two families.

Full text available upon request to the author


Title: Quantifying forest cover changes in the Philippines from 2000-2012 from landsat-derived global forest cover dataset

Authors: Brent Fallarcuna and Gay Perez

Conference title: 36th Asian Conference on Remote Sensing, October 2015



Updated information of forest resources is vital in proper management and policy formulation from local to national scale. However, data generation regarding forest maps and statistics in the Philippines is hampered by multiple imagery sources, varying scale and resolutions, undisclosed methods of mapping and different forest classes and definitions used. In this study, processed Landsat forest cover data was analyzed to quantitatively measure the rates of forest loss on the regional level, to analyze its level of applicability and to validate forest changes through various geospatial and other forestry data. Raw data consists of forest tree cover, loss, gain and loss year layers. Five forest vectors were used to extract forest areas from the main data rasters before subsetting them regionally. A forest tree cover threshold of 85 percent and above was used to isolate densely forested rasters. These rasters were further analyzed to produce gain, loss and aggregate net change ranking; annual forest loss and annual forest cover trends, with their corresponding maps and statistics. Among the regions, Zamboanga Peninsula had the highest aggregate net change (-7.18%) followed by CALABARZON (-3.41%) and CARAGA (-3.16%) from 2000-2012. CARAGA had the highest average annual forest loss of-147.53 ha and a correlation of 0.7. On the other hand, Zamboanga and CALABARZON had average annual forest loss which were less than hundred hectares (-73.10 ha and-31.4 ha; r = 0.34 and r = 0.13, respectively). However, only CARAGA had a significant correlation, with p value < 0.05 (0.0098). In terms of annual forest tree cover, CARAGA had the highest average annual forest cover increment of-1,238.16 ha, followed by Zamboanga Peninsula (-830.12 ha) and CALABARZON (-527.61 ha). Forest cover increments exhibited negative values since forest recovery is always outweighed by forest extraction. Error matrix of loss, gain and no change registered an overall accuracy of 86.48% and 92.26%, for loss and gain respectively. Yet, the assessment had lower user accuracy values compared to that of producers'. This might be affected by inadequate high resolution images (mainly Google Earth and Landsat) used in the satellite based validation. The research showed that freely available forest global datasets could be used to pinpoint significant areas where forest loss is occurring. Correlation analysis with other forestry data and field validation through group discussions and interviews may also illustrate insights regarding the drivers of forest loss on a particular area.


Title: Dynamic Contact Angle Measurements on Various Fabric Surfaces Using a Simple Optical Vision System

Authors: Hernando Siy Salapare III, Jamaica Palay, Gay Perez, et al.

Conference title: Proceedings of the 25th Samahang Pisika ng Pilipinas Physics Congress, October 2007



A contact angle measuring device was developed for undergraduate Physics laboratory experiments. The new device was tested on determining the wettability of waterproof fabrics. Intel® Play™ QX3™ Computer Microscope was used in determining the changes in time of the contact angles of different waterproof fabrics washed in different conditions. At contact angles between 80 to 90 degrees, the untreated samples changed its wettability from being hydrophobic to being hydrophilic. For washed sam-ples it was determined that the change in wettability occurred at contact angles between 40 to 70 degrees. The rate of recession of the contact angle of the water droplet was related to the rate of absorbance of the waterproof fabric. It was found that high temperature of water used for washing increases the rate of recession. Measurements of the contact angle in time were able to show the dynamic behavior of the water droplet on the fabric surface.