Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción...

41
Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República Dominicana (REDD CCAD GIZ): Interfaz de Manejo Simplificado para el Sistema de Alerta Temprana PN 08.2211.4-001.00 Consultor: Jeffrey R. Jones. Para: IMPLEMENTACIÓN DEL SISTEMA REGIONAL DE MONITOREO DE DEFORESTACION Y DEGRADACIÓN DE BOSQUES PARA CENTROAMÉRICA Y REPÚBLICA DOMINICANA. Contrato No.: 83115498 Agencia COSTA RICA Informe de Avance: Producto 1: 17 agosto de 2012 Productos 1. Interfaz Versión 2 con visualizadores y nuevos métodos 2. Articulo científico

Transcript of Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción...

Page 1: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

Programa Reducción de Emisiones de la Deforestación y Degradación

de Bosques en Centroamérica y República Dominicana

(REDD – CCAD – GIZ):

Interfaz de Manejo Simplificado para el Sistema de Alerta Temprana

PN 08.2211.4-001.00

Consultor: Jeffrey R. Jones.

Para: IMPLEMENTACIÓN DEL SISTEMA REGIONAL DE MONITOREO DE

DEFORESTACION Y DEGRADACIÓN DE BOSQUES PARA

CENTROAMÉRICA Y REPÚBLICA DOMINICANA.

Contrato No.: 83115498 Agencia

COSTA RICA

Informe de Avance: Producto 1: 17 agosto de 2012

Productos 1. Interfaz Versión 2 con visualizadores y nuevos métodos 2. Articulo científico

Page 2: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

Tabla de contenido

1. Interfaz VCF Versión 2 con visualizadores y nuevos métodos

a. Visualizador VCF y Visualizador Amenaza Forestal

i. Capacidades

ii. Métodos

1. VisualizarVCF

2. AmenazaImg_Event

3. Secuencia_VCF_Event

2. Eliminación de áreas no clasificadas

a. Bagging ('Bootstrap Aggregation') y promediar resultados

b. Comparación de promedio y desviación estándar

i. Método meantop3

c. Detección y control de áreas nubosos

i. reliability0count01_11.tif

ii. RegionRel3CloudStack.tif

iii. Reliability3SummaryRegion3Freq.tif

iv. RegionRel3CloudMaski.tif

3. Exportación de archivo VCF_promedio.tif en Alerta Temprana

4. Artículo científico

a. MODIS Based REDD+ Regional Monitoring System for Forest Cover

Reporting of Central America and the Dominican Republic

b. 'International Journal of Forestry Research'

i. Tópico especial

5. Anexos

a. Articulo 'MODIS Based REDD+ Regional Monitoring System for Forest

Cover Reporting of Central America and the Dominican Republic'

Page 3: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

Interfaz VCF Versión 2

Un limitante de la Interfaz Grafico Versión 1 para VCF-GIZ fue su dificultad en

comunicar resultados a audiencias no técnicas, especialmente de la secuencia de datos

VCF. En un principio la Interfaz fue creada para eliminar el uso de la línea de comando

y llevar el procesamiento de datos MODIS para la creación de la imagen VCF a un

ambiente más reconocible tipo Windows. El desarrollo de la Versión 2 tuvo como

objetivo un método de presentación de datos para el usuario y para la divulgación de

resultados.

El uso de la secuencia VCF para la presentación y el análisis de comportamiento forestal

agrega un componente dinámico al análisis que aprovecha de la capacidad de rápida

actualización de MODIS/VCF. La secuencia VCF presenta cambios anuales en la

densidad de cobertura de bosque con una discriminación fina de estados de densidad. La

secuencia VCF ayuda a visualizar dinámicas complejas de cobertura forestal, producto de

intervención humana al igual que eventos naturales. Esta dinámica no ha sido accesible

con el monitoreo en el pasado por los largos periodos de actualización de datos, que

suelen ser de 5 a 10 años entre actualizaciones de mapas de cobertura forestal. Sin

embargo, la dinámica de cambio, y los datos que se utilizan para describirla, llegan a

llenar un vacío muy importante en el monitoreo para efectos de REDD o REDD+.

En un inicio el acceso a la visión dinámica de la secuencia VCF fue limitado a los

usuarios de ENVI; aunque ENVI representa la tecnología de punta en el análisis de

imágenes (es un software muy ampliamente difundido dentro de NASA), su alto costo y

la tradición de uso de ArcGIS y ERDAS resulta en pocas licencias para el uso de este

paquete. Aunque otros paquetes de software tienen la capacidad de desplegar las

imágenes VCF, les falta herramientas como el 'perfil Zeta' que ayuda en la comprensión

de las secuencias VCF.

Figure 1: Tres ventanas de la opción 'Visualizar VCF'

Page 4: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

La Versión 2 de la Interfaz VCF se dirige principalmente a la presentación de los

resultados únicos del VCF, que son la presentación de tendencias de cambio de densidad

en el mediano plazo, y la clasificación de bosque específicamente en cuanto a sus

características dinámicas de VCF.

A la Interfaz VCF Versión 1 se agregó dos botones; 'Visualizar VCF' y 'Visualizar

Amenazas' (Figura 1). Estos botones inician tres ventanas de información sobre el VCF.

En el caso de Visualizar VCF, las tres ventanas son 1) una ventana general de la región 2)

una ventana magnificada alrededor de un punto seleccionado con el mouse sobre la

ventana general y 3) una ventana de Perfil Zeta, que presenta todos los valores de la

secuencia VCF.

De las tres ventanas de la Versión 2, 'Grafico Secuencia VCF' es la más interesante.

Cada valor VCF es el porcentaje de cobertura forestal del pixel en un año particular.

Cuando se visualiza como secuencia ilustra procesos de cambio (o no cambio) con una

claridad que presta para una mayor comprensión de los procesos actuales. Por ejemplo,

la secuencia ilustra la importancia de la recuperación forestal en la dinámica de

cobertura; donde la metodología IPCC enfoca en buscar áreas de bosque que cambian a

áreas no-bosque, la secuencia VCF presenta dinámicas más complejas generadas por la

recuperación del bosque.

En la Figura 1 en la ventana 'Grafico Secuencia VCF' se presenta una secuencia donde

un bosque sufre una reducción de su cobertura de hojas, y se recupera en forma repetida.

El eje vertical de la ventana Grafico Secuencia VCF describe la densidad de bosque, y el

eje horizontal marca el paso de los años; la línea solida traza los cambios observados en

los valores VCF, y la línea intermitente describa la línea de tendencia calculada en base

de los valores observados. Esta secuencia ocurre en los lechos de un rio, entonces podría

ser resultado de crecidas de agua en el cauce del rio, o un proceso de manejo de los

sedimentos cerca al río con fines agrícolas o pecuarias. Lo que llama la atención es la

capacidad de recuperación del bosque, que después de sufrir la defoliación, repone hojas

en un año siguiente para seguir como bosque vivo. El Perfil Zeta dibuja esta dinámica de

una manera que nunca se ve en monitoreo con actualizaciones menos frecuentes.

Las ventanas regionales y de magnificación presentan los datos VCF como imagen de

'falso color', donde las intensidades de RGB (rojo, verde y azul) se deriva de los valores

de VCF en el primer, quinto y último año de la secuencia. La presentación a color da una

impresión general de la condición del bosque. Los áreas más claros son los que tienen

valores altos en cada año que compone RGB; es decir, son áreas de alta densidad de

arboles cuya densidad no cambia mucho durante el periodo. Los colores más oscuros

indican VCF bajos, por ejemplo en zonas agrícolas o en zonas de inundación donde crece

poca vegetación (vea especialmente las zonas de lagunas en la costa de la moskitia).

El botón de 'Visualizar Amenazas' presenta una interpretación de los datos VCF que sale

del análisis 'Alerta Temprana'. Al igual que 'Visualizar VCF' abre tres ventanas, 1) de

vista general, 2) de magnificación y 3) de leyenda. El mapa de amenazas es una

combinación de diferentes informaciones de la Alerta Temprana; 0) agua y áreas fuera de

Page 5: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

la región del estudio Centroamérica y República Dominicana, 1) tierra (área sin bosque),

2) otros bosques, 3) área de bosque con VCF estable, 4) bosque con VCF inestable, y 5)

Alerta Temprana. Los colores de la leyenda están programados para asociar con estas 6

clases, entonces el mapa de amenazas puede ser actualizado y desplegado siempre que

utiliza códigos 0 a 5; el cambio de colores o las definiciones de la leyenda requieren una

modificación del código base del modulo.

Las categorías del mapa están definidos desde las mascaras de tierra y de bosques de la

región, generado en base de los mapas forestales provistos por el proyecto REDD-

CCAD-GIZ, de donde salen las categorías 'agua' y 'tierra'. Las categorías de bosque se

generan en base de los valores del promedio VCF, el pendiente de la tendencia de la

secuencia VCF para cada pixel, y la desviación estándar de los valores de cada pixel. Las

definiciones de estas clases son;

– Bosque VCF estable – desv < 10, abs(pend) < 3, VCF Promedio > 70

– Bosque VCF inestable – desv > 10, abs(pend) < 3

– Alerta temprana – desv < 10, abs(pend) < 3, ΔVCF>50

– Otros bosques (que no caen en las categorías estable o inestable)

Figure 2 Visualizar Amenazas

El mensaje que presenta el mapa de amenazas es que los bosques existen en diferentes

condiciones y calidades. Hay mucho del bosque de la región que está sufriendo

degradación que no llega al punto de deforestación (tomando como referencia una

densidad de 30% de cobertura de copa), pero que puede representar la pérdida de más de

50% del contenido de carbono del bosque. Los bosques con VCF inestable tienen

variaciones grandes en el VCF entre los diferentes años, tal como se presenta en la Figura

1. Estos bosques pierden volumen de carbono, y pierden capacidad de secuestro de

carbono, por la reducción o la eliminación de su capacidad fotosintética para ciertos

periodos, e inclusive pueden estar en peligro de daños más severos.

Page 6: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

'Bootstrap aggregation' y la eliminación de brechas en VCF causadas por nubes

El aspecto menos satisfactorio de las coberturas VCF son los efectos de áreas con

coberturas casi permanentes de nubes. En la Figura 3 se compara imágenes de

Talamanca, en Costa Rica; un VCF final en Lat/long, comparada con una imagen

compuesta de 16 días en proyección Sinusoidal (por eso las imágenes tienen formas y

orientaciones un poco distintos) del mes de junio. En la imagen VCF hay zonas que

reportan poca o ninguna cobertura de bosque en las montañas del lado Caribe de

Talamanca. Es una zona conocida por sus bosques extensos. Sin embargo, se ve

claramente la persistencia de nubes en ciertos puntos de la imagen compuesta que

corresponden a 'brechas' en la cobertura VCF.

Varias estrategias fueron examinadas para detectar y eliminar estas brechas, empezando

con cambiar el modelo del árbol de decisión, siguiendo a una selección de pixeles basada

en características estadísticas, y finalmente en el análisis de índices de calidad de pixeles.

Figure 3 VCF y MOD13Q1 Talamanca, Costa Rica

La primera estrategia fue la implementación del árbol de decisión tipo M5 con 'Bootstrap

Aggregation', con la hipótesis que la aplicación de un árbol de decisión más sofisticada

ayudaría a resolver el problema de pocos datos en ciertos áreas. El modelo del árbol de

decisión utilizado en el VCF-GIZ es el 'Replicating Tree', sin uso de poda de ramas del

árbol. El modelo M5 permite poda para eliminar clases de poca probabilidad. Además,

el 'Bootstrap Aggregation' (Bagging) genera una serie de arboles de decisión variando los

puntos de inicio del análisis y finalmente saca un promedio de todas las clasificaciones

generadas.

Los resultados del M5 con Bagging no llegaron a la altura que se esperaba. Primero,

requiere un procesamiento mucho más largo por la estructura del modelo; asigna pixeles

a clases, y cada clase tiene una fórmula para integrar datos de diferentes métricas para

generar un resultado final, mientras que el Replicating Tree requiere un solo paso de

clasificación y asignación del valor final. Además, la aplicación del bagging requiere la

Page 7: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

repetición de la clasificación entera entre

10 y 30 veces. ENVI no soporta M5

entonces fue necesario desarrollar un

modulo especial para el análisis. Aun

así, muchas clasificaciones duraba

muchas horas en finalizar para una sola

tessera.

Obviamente, con una clasificación

exitosa no habrá argumento en contra al

uso del sistema más tardado en ejecutar.

Sin embargo, el resultado en general del

método fue una homogenización de

valores, por el efecto de promediar

resultados de muchas iteraciones del

árbol de decisión.

El resultado del M5 con Bagging se

presenta en la figura 4 para la misma

zona que la figura 3 elimina algunos de

los valores bajos en las zonas de nubes, pero igualmente sube los valores de cobertura

aun en zonas agrícolas del Valle de la Estrella y Changuinola, que les hacen parecer con

mas cobertura de bosque. En vista del resultado pobre de varias clasificaciones, se

decidió abandonar la aplicación del modelo M5.

Otra estrategia fue por medio del análisis estadístico. Dentro del modulo de la Alerta

Temprana se insertó unas rutinas para detectar pixeles dudosos, y corregirlos. La bondad

de este método es que el

programa Alerta Temprana ya

está hecha, y los datos

estadísticos estaban

disponibles como parte del

procesamiento de la Alerta.

El criterio de selección

aplicada fue una comparación

del valor promedio del pixel

(Promedio VCF de la

secuencia de capas) con las

desviación estándar del mismo

pixel. En la Figura 5 se

presenta el perfil de valores

VCF para una zona nubosa.

Los valores alternan entre

valores muy altos a valores casi 0, ya que cubierto de nubes la zona devuelva una firma

espectral sin árboles. Como prueba de contaminación de nubes cada pixel fue evaluada a

Figure 4 Resultado de M5 con Bagging

Figure 5 Perfil Zeta Zona Nubosa Talamanca

Page 8: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

determinar si la desviación estándar superaba el promedio VCF; esta estrategia no logra

identificar todas las zonas nubosas, entonces la corrección termina incompleta.

En la Figura 6 se presenta la corrección a la derecha, que demuestra una reducción en los

áreas de nubes, y mantiene las zonas agrícolas correctamente, pero deja sin cambiar la

mayoría de las brechas de nubes.

Esta metodología se basaba en líneas de código para la identificación de las brechas, y

otro código que extrae el promedio de los 3 valores más altos en la zona de nube. Si la

brecha pasa mucho cubierto de nubes que van a generar VCF casi 0, entonces los valores

más reales son los más altos. La rutina meantop3 recibe el vector de valores de la

secuencia VCF para el pixel identificado como nuboso, aplica un 'bubble sort' para

identificar los valores más altos, y devuelva el promedio de los 3 valores más altos.

function meantop3, origpoints, npoints, lastlayer

done = 0

nowpoints = make_array(npoints+1, /integer)

nowpoints[0:npoints-1] = origpoints

nowpoints[npoints] = lastlayer

while not done do begin

done = 1

for s = 1,npoints do begin

if nowpoints[s-1] lt nowpoints[s] then begin

tmpval = nowpoints[s-1]

nowpoints[s-1] = nowpoints[s]

nowpoints[s] = tmpval

done = 0

endif

endfor

endwhile

avgtop3 = byte(round((nowpoints[0] + nowpoints[1] + $

nowpoints[2])/3))

return, avgtop3

end

Figure 6 Corrección basada en desviación estándar y promedio

Page 9: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

La rutina meantop3 se llama desde el interior de la rutina AlertaTemprana con el

siguiente código;

if reliabilityvec[i] eq 1 then begin

newval = meantop3(y, vcfvars2_limit+1, vcfvect_last[i])

vcfvect[i,vcfvars2_limit] = newval

endif else begin

vcfvect[i,vcfvars2_limit] = vcfvect_last[i]

endelse

Estas rutinas son de interés permanente, porque sirven para aplicar la corrección con

cualquier estrategia de definición de pixeles nublados, con solo que se pone en la forma

de un vector de píxeles en el archivo VCF.

Como última estrategia para identificar y corregir brechas por nubes se volvió a los

índices de calidad de pixel que se utiliza en la construcción de las compuestas de 32 días.

La construcción de las compuestas busca el mejor pixel, según varios criterios, pero no

excluye pixeles con nubes con la expectativa que contiene alguna información que podría

ser útil. Los índices de calidad de los pixeles están actualizados para indicar la calidad

del pixel escogido, y grabado en el archivo de la imagen compuesta.

Para el análisis de nubes se utilizaba la banda Reliability, que tiene 3 valores relevantes; 0

para pixeles sin contaminación, 1 para pixeles con algo de contaminación, y 3 para

pixeles nubosos. Se evaluó dos valores; el 0 y el 3.

Todos los archivos de compuestas mensuales fueron resumido para llegar al dato de

cuantos meses de cada año tenia pixeles Reliability0, sin contaminación. Cuando se

compara la distribución de pixeles Reliability0 con las brechas de nubes no lleva una

relación muy

cercana.

El resumen de

los pixeles

Reliability3 da

mejor resultado.

Detecta que hay

pixeles hasta

con un máximo

de 105 meses

con pixeles

nublados (de los

120 meses en el

periodo

analizado), pero

que 95% de los

pixeles forestales tienen menos que 25 meses con pixeles nublados. En la figura 7 se ve

la buena correspondencia entre la concentración de altos números de pixeles nublados y

Figure 7 Comparacion VCF con Reliability3

Page 10: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

las brechas en la imagen VCF. Como ejemplo, el pixel en el centro del cuadro rojo tiene

58 días nublados del total de 120 meses. En la Figura 8, se aprecia la distribución de la

máscara de pixeles nublados con un umbral de 25 pixeles nublados.

Figure 8 Mascara Reliability3 Umbral 25 Días

En contraste, en la zona del canal de Panamá los pixeles nublados son menos frecuentes.

En una aparente contradicción, la distribución de pixeles Reliability0 clasifica la zona del

Canal muy similar a Talamanca, mientras que Reliability3 demuestra que la zona del

Canal en particular tiene una cobertura de nubes reducida.

Figure 9 Mascara nubes Canal de Panama

Page 11: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

Esta aparente contradicción se debe a la existencia de mas pixeles parcialmente

contaminadas en la zona del Canal, que todavía permiten en análisis VCF a pesar de su

relativamente baja calidad.

Conclusión Interfaz

La incorporación de los visualizadores de VCF y de Amenazas cumple con varios

objetivos. Primero, provee una manera inmediata de uso de los datos VCF, ya que la

interfaz se construye bajo una modalidad de IDL-ENVI que permite su distribución y uso

con instalaciones IDL sin licencia; es decir, se puede instalar una copia 'no licenciada' y

gratis de ENVI en una computadora y utilizar la Interfaz VCF en ese ambiente. El

usuario estará limitado en que no puede utilizar las opciones del programa ENVI o IDL

en forma independiente y no puede editar el código del programa en IDL, pero sí puede

ejecutar el archivo compilado de la Interfaz sin restricción. Segundo, facilita la

comprensión de los datos VCF por parte de los técnicos capacitados en el programa

REDD-CCAD-GIZ, ya que pueden revisar datos y compararlos con otros datos de sus

ministerios sin la licencia ENVI ni con un conocimiento profundo de los programas

ENVI o IDL. Tercero, la Interfaz permite la presentación de datos VCF a decisores para

apoyar sus conclusiones o recomendaciones en cuanto al manejo de asuntos REDD.

La leyenda del mapa de amenazas a bosques requiere una revisión 'política', ya que en la

categoría de VCF inestable se unen áreas en proceso de variación por actividad humana,

y zonas naturales con variaciones en VCF por causa de inundaciones. Un caso en

particular es el noreste del Peten de Guatemala, que contiene un bosque inundable con

relativamente poca intervención humana y una gran variabilidad en el VCF. Hay que

buscar una terminología que aclara que no toda variación en el VCF es por causa

humana.

Para mejorar el aspecto visual de las imágenes VCF, es necesario aplicar una corrección

para zonas nubosas. El análisis de valores Reliability3 sugiere que una máscara con un

umbral de 25 meses nubosos en la década identifica zonas cuyo valor VCF está

claramente afectado por las nubes. La precisión de los pixeles bajo la máscara de nubes

puede mejorarse por medio del análisis de la secuencia VCF y la aplicación de valores

del último año en que un valor realista VCF fue posible; esta corrección obedece a la

misma lógica que la construcción de las compuestas de 32 días, en el sentido que se

extiende el periodo de buscada de meses a años en vista de las condiciones

meteorológicas. Como sugerencia, se puede construir unas capas índices para los VCF

que indica cuales son las zonas de máscara nubes, y cuál es la frecuencia con que se

obtiene datos razonables para su corrección.

Artículo científico

En vista del interés en dar publicidad al Proyecto REDD-CCAD-GIZ, se han hecho varias

presentaciones del sistema de monitoreo en talleres y capacitaciones. Los sitios de estas

presentaciones han sido en el IICA, el CATIE, el CENAT (todos en Costa Rica) y una

presentación en Quebec, Montreal, Canadá.

Page 12: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

En octubre 2011 el sistema de monitoreo fue presentado en una conferencia en Laval

University en Quebec, Canadá. El titulo de la conferencia fue 'Taller Internacional sobre

la Forestería Comunitaria en el Marco de REDD+'. Los organizadores de la conferencia

buscaron por medio de contactos en el CATIE alguien para hacer una presentación de un

sistema regional de monitoreo, entonces se hizo la presentación 'Sistema Regional VCF

para el Monitoreo de Deforestación y Degradación de Bosques para Centroamérica y la

República Dominicana', que incluía una descripción del proyecto REDD-CCAD-GIZ

como el contexto del desarrollo del sistema.

En el mes de noviembre después de la conferencia, una nota fue circulada por los

organizadores del interés de unir las presentaciones de la conferencia en un edición

especial de la revista 'International Journal of Forestry Research' sobre el tópico especial

'REDD+ Mechanism in Developing Countries'. Este aviso fue compartido con GIZ, pero

no se llegó a ninguna conclusión sobre el desarrollo de tal articulo inmediatamente. En

abril y mayo se iniciaron discusiones sobre el interés en dar seguimiento a la invitación,

entonces un tiempo fue reservado en el contexto de esta consultoría.

En la página web de la revista se publicó la siguiente información, en adición a las

invitaciones personales extendidas por los editores, quienes todos participaron en el

Taller de Quebec: REDD+ Mechanism in Developing Countries

PDF Call for Papers | HTML Call for Papers Guest Editors: Damase Khasa, Alison Munson, Mariteuw Chimère Diaw, Nancy Gelinas, Nadine T. Laporte, Glenn Bush Manuscript Due: Friday, 3 August 2012 Publication Date: Friday, 21 December 2012

Al final, se definió generar el articulo 'MODIS Based REDD+ Regional Monitoring

System for Forest Cover Reporting of Central America and the Dominican Republic',

como esfuerzo conjunto entre Abner Jiménez, Laszlo Pancel y el consultor. Un plan de

articulo fue desarrollado por medio del intercambio de ideas escritas y algunas

conferencias Skype; el articulo fue terminado y entregado a los editores de la revista el 2

agosto. En este momento, sigue en el proceso de evaluación y edición para definir si se

incluye en la revista.

El articulo en formato Word esta anexo a este informe al igual que el anuncio de la

edición especial.

Resumen

Durante este periodo se enfocó en tres tareas;

1) La generación del la Versión 2 de la Interfaz VCF, con una capacidad de desplegar

datos VCF sin requerir licencias de ENVI

2) La evaluación de alternativas para identificar las brechas en los valores VCF donde

hay cobertura persistente de nubes (para distinguirlas de zonas donde realmente no hay

Page 13: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

cobertura forestal) y la identificación de una estrategia para remplazar los valores bajos

de las brechas con valores más realistas.

3) La creación de un artículo científico documentando el uso del Sistema de Monitoreo

VCF-GIZ en el contexto de REDD+

Page 14: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

Materiales Anexos

1. Artículo Científico: JimenezJonesPancelMODISMonitoringCADR2.docx

2. Descripción de la edición especial para artículo:

IntJForResRMDC22march12AM.pdf

3. Recibido de articulo de editores de revisa (correo electronico)

4. Mascara Reliability3 Umbral 25: RegionRel3CloudMaski.tif

5. Modulo compilado VCF_InterfaceV2.sav

Page 15: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

Anexo 1: Artículo Científico: JimenezJonesPancelMODISMonitoringCADR2.docx

Page 16: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

MODIS Based REDD+ Regional Monitoring System for Forest Cover Reporting of Central America and

the Dominican Republic Abner J. Jiménez

1; Jeffrey R. Jones

2; Laszlo Pancel

3 1. Pasante de Doctorado en Tecnología de la Información Geográfica de la Universidad de Alcalá, Spain, y Especialista Sectorial del Programa Regional REDD-CCAD/GIZ. Email: [email protected]; [email protected] 2. Consultant, Regional REDD-CCAD/GIZ Program. Email: [email protected] 3. Asesor Principal, Programa REDD-CCAD/GIZ, San Salvador, El Salvador. email: [email protected] ABSTRACT REDD and REDD+ monitoring place special demands on remote sensing techniques. Forest degradation or enhancement need to be detected with annual or more frequent updates. The high temporal resolution of MODIS data compensates for its moderate spatial resolution by enabling the creation of monthly VCF (Vegetation Continuous Field) images, based on normalized phenological metrics which permit the differentiation of degradation from phenological cycling. Annual VCF images are analyzed to detect patterns of forest canopy change using basic statistics of annual values for individual pixel positions. A classification of pixel statistics distinguishes the class of Stable Non-Intervened Forest from Stable Intervened Forest, representing untouched natural forest stands, and forest stands experiencing a variety of human interventions, respectively. These classes cover 20% and 50% of the region's forest area. The larger area of Stable Intervened Forest mark regions where human efforts to enhance forest structure might be implemented and monitored for recognition under REDD guidelines. Key words: REDD, REDD+, MRV, MODIS, VCF, Monitoring, Central America, Dominican Republic 1. INTRODUCTION

Requirements for forest monitoring have changed significantly with the

introduction of REDD and REDD+ (Reduction of Emissions from Degradation and

Deforestation), with their focus on primary forest carbon, and an expanded

definition including secondary forest carbon, respectively. Both are referred to

hereafter as REDD, where the primary concern is the quantity of carbon, rather than

amounts of available lumber. REDD has also increased the expected tempo of forest

cover monitoring (MRV in REDD terminology - Monitoring, Reporting and

Validation), where results are required in a usable format on a monthly, or at most, a

yearly basis as a mechanism to document and justify monetary compensation for

carbon sequestration (GOFC-GOLD, [5]). Significantly, REDD also requires

information on forest degradation, the loss of carbon within standing forests, which,

as noted by Herold et.al. [11] represents nearly 10 times the carbon loss caused by

outright deforestation. The human population is the primary agent of different

forms of forest degradation and their extents, and its activities must be considered

Page 17: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

in understanding and addressing degradation. Forest monitoring objectives require

information on multiple scales; broad region wide detection of current trends, and

detailed local information to confirm the existence and nature of the trends; the

MODIS VCF focus will focus on broad region wide data.

The MODIS based VCF methodology addresses several key problems of MRV,

especially timely coverage, and information on degradation. The MODIS satellite

constellation was launched expressly to monitor earth surface conditions, featuring

daily coverage and a high level of spatial and spectral accuracy. Early in MODIS'

implementation a research process was begun to improve on the AVHRR legacy-

monitoring-methods through the introduction of the VCF metric (Vegetation

Continuous Field - Hansen et.al. [6]). VCF represented a departure from earlier

forest monitoring efforts which focused on forest classification rather than carbon

measurements, because this was expected to provide more realistic information

regarding the status of carbon sinks in order to draw conclusions regarding Green

House Gas emissions and sequestration.

In spite of its lower spatial resolution, MODIS is a powerful tool for forest

monitoring because of its high temporal resolution. The smallest pixel size for

MODIS products is a nominal 250m, nearly 10 times the 30m of a LANDSAT pixel,

but the daily MODIS coverage greatly increases the probability of finding high

quality pixels to use for regular analysis. Comparisons of MODIS products in the

analysis of tropical forest coverage in Central America finds them to duplicate

results of higher resolution products with reasonable levels of accuracy (Hayes

2008). Similar comparisons, specifically for VCF, in Brazil, Indonesia and Africa

arrive at similar conclusions (Hansen et.al. [7-9]).

MODIS products have been used to construct the relatively simple Disturbance

Index (DI) based on MODIS EVI (Enhanced Vegetation Index) and LST (Land surface

temperature) as an efficient method for detecting remote disturbances without field

level monitoring (Mildrexler [15]). Coops [1] successfully applied DI to detect fires

in remote Canadian forests, and a similar approach is used in the The Global Forest

Disturbance Alert System[22].

VCF expands on the capability of DI by not only indicating locations of disturbances,

but also the magnitude of variations from year to year. Information on magnitude of

annual variation in forest cover offers key insights into the health of forests, and the

types of disturbances that might be affecting them.

Population pressure is an important motor in the changes of forest cover, although

in the past analyses of land cover change focused on replacement of forest by

farmland or pastures. Central American farmers interact with forests in more

complex ways than simply cutting them down, and these interactions can be seen to

a certain extent through VCF monitoring.

Page 18: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

Deforestation in Central America and the Dominican Republic broadly follows the

trend in population increase. A population increase of over 30% between 1990 and

2005 was accompanied by a decrease and a degradation in forest cover, especially

in the largest countries of the region, as forests are slowly cut down or occupied by

land-seeking farmers.

Table 1: Population by country and by year: 1990 - 2005

Country 1990 1995 2000 2005 Change

Belice 186 214 242 270 145.16%

Costa Rica 3076 3475 3925 4322 140.51%

El Salvador 5110 5669 6276 6875 134.54%

Guatemala 8908 10004 11225 12700 142.57%

Honduras 4879 5654 6485 7347 150.58%

Nicaragua 3960 4477 4957 5483 138.46%

Panamá 2411 2670 2948 3228 133.89%

República Dominicana 7066 7705 8396 9100 128.79%

Source: ECLAC, 2009 [3]

Table 2: Forest Cover and Forest Cover Change 1990 - 2005

Area in forest: Thousands of Has.

Deforestation

Country/Year 1990 2000 2005 2010

Belice 1586 1489 1441 1393 -13.85%

Costa Rica 2564 2376 2491 2605 1.57% El Salvador 377 332 309 287 -31.36%

Guatemala 4748 4208 3938 3657 -29.83%

Honduras 8136 6392 5792 5192 -56.70%

Nicaragua 4514 3814 3464 3114 -44.96%

Panamá 3792 3369 3310 3251 -16.64%

República Dominicana 1972 1972 1972 1972 0.00%

Region 27689 23952 22717 21471 -28.96%

Source: FAO, 2010 [4]

The weakness of the region in the development of urban employment opportunities

in manufacturing or services historically has contributed to deforestation in Central

America and the Dominican Republic. The countries of the region for most of the

last century were internally perceived as frontier societies, with uninhabited forest

lands available for those with the initiative to occupy them (Jones [14]), a condition

that continues today on an individual level even though most governments have

prohibited or discouraged deforestation for land settlement . Heckadon [10]

documents the example of Panama in a period when official policies began to

discourage land clearance at the same time that individual farmers continued to

view forest clearance as a legitimate path to financial security. Although a part of

the land clearance may be carried out directly or indirectly by large land holders,

Page 19: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

land clearance represents an economically attractive option for poor farmers, both

in the exploitation of forest resources, and also as a means to access farmland for

self-capitalization (Sewastynowycz [21]).

An important element of tropical deforestation by small farmers is its partial and

heterogeneous nature. A variety of climatic and topographical conditions make the

complete clearance of forests uncommon, creating what have come to be recognized

as ‘agroforestry systems’ where trees coexist with a variety of productive activities.

The forest cover associated with these systems varies between 10% in pastures and

cropland to more than 80% in coffee and cacao, and offer an attractive livelihood to

poor farmers even when lands remain partially forested. From a carbon

standpoint, these systems duplicate, at least partially, the characteristics of forest

systems (Rapidel et.al. [18]).

VCF provides a tool for the detection and quantification of forest degradation for

Central America and the Dominican Republic. The occupation of forests by the

region’s growing population does not result in immediate deforestation, but rather a

mosaic of cleared areas interspersed with dense forest stands and agroforestry

systems of different kinds, which maintain many aspects of forest cover but make a

major contribution to changes in the carbon stock of the region through the

elimination of some forest species and the introduction of more homogeneous

ecosystems based on crops such as coffee or cacao. As a result, forests are rarely

completely eliminated from Central American landscapes, but are reduced to small

patches or low densities to accommodate other activities.

The availability of a variety of satellite based imagery products, and the existence of

National Forest Inventories, permits the development of multi-scale monitoring

systems which use different data sources for complementary functions in the

monitoring process. This is implementation is described for the national forest

monitoring system of Canada in Wulder et.al [24]. Forest inventory plots provide

the most accurate assessments of carbon concentrations, but these are logistically

limited to extremely small areas and/or long update cycles. Detailed imagery

products such as aerial photography, LIDAR and LANDSAT, represent a bridge

between inventory plots and regional evaluations, through their ability to detect and

generalize the extent of forest classes over significantly larger areas the inventory

assessments. Detailed imagery products suffer a similar logistical limitation to

inventory plots, in that the relatively high costs of collecting and analyzing these

data lead to relatively long update cycles. The large pixel size of MODIS permits the

evaluation of extremely large areas (Mildrexler [15]), but at the same time limits the

accuracy of ground observations both spatially and spectrally, as many pixels

represent mixed ground coverage types. MODIS products are recognized as a key

element in the detection of short term events that affect inventory but which cannot

be captured through the tools of field level inventories or detailed imagery.

Page 20: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

MODIS quick-update products improve temporal response at the cost of increased uncertainty in locations of phenomena, or their exact dimensions, requiring the use of complementary higher resolution imagery to improve precision. A MODIS pixel covering approximately 5ha. with its VCF dropping from 90% to 80% might indicate either a degradation of 10% over the entire 5has, or the complete deforestation of 0.5ha, and no change in the remaining 4.5ha. The use of aerial photography or other higher resolution imagery will be required to distinguish what observed changes mean at the most detailed level, with the benefit that the number of images to be reviewed is greatly reduced by having an exact location to evaluate.

2. BACKGROUND

The “Sourcebook of methods and procedures for monitoring and reporting

anthropogenic greenhouse gas emissions and removals caused by deforestation,

gains and losses of carbon stocks in forests remaining forests, and forestation”

(GOFC-GOLD [5]) clearly identifies on its first page the monitoring needs, and

monitoring challenges, for the implementation of REDD+ goals:

“At current status of negotiation five forest-related activities have been listed to be

implemented as mitigation actions by developing countries, namely: reducing

emissions from deforestation (which implies a land-use change) and reducing

emissions from forest degradation, conservation of forest carbon stocks, sustainable

management of forest land, Enhancement of forest carbon stocks (all relating to

carbon stock changes and GHG emissions within managed forest land use). . . The

book emphasizes the role of satellite remote sensing as an important tool for

monitoring changes in forest cover ”

The Sourcebook recognizes several different methods for evaluating forest change.

One is the comparison of classified images (GOFC-GOLD [5]: p2-15); classified

images are developed on the basis of careful field work, and cover change is

evaluated through a comparison of classifications from different time periods. The

output identifies how many and where pixels have changed from one class to

another, especially from forest to a non-forest class. This strategy is used to assess

global carbon stocks based on classifications of land cover, to capture categorical

differences between major types of land cover (Saatchi et.al. [20], Pan [17]).

Through a combination of land class information and forest inventory data from a

variety of forests, an initial estimate of carbon existence can be derived. This

analysis demonstrates the power of remote sensing to create global scale

summaries of information, which can incorporate the most recent information on

carbon concentrations from identified forest classes, and generate a dramatic

worldwide view of carbon stocks.

A major theoretical obstacle confronts the production of the forest class images

used in the above analysis; this same obstacle is relevant for the entire activity of

Page 21: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

REDD forest change evaluation. Forest classifications inevitably are faced with

border errors, where a gradual decrease in forest density creates doubt as to the

exact location of the forest/non-forest division; this is a normal characteristic of

natural systems whose vegetation changes are gradual, and are seldom as clear cut

as would appear on a land classification map. The same problem confronts the

analysis of forests themselves, because forest densities are not homogeneous even

in their interiors. Variations in soils, water availability, exposure, etc, lead to

differences in density within a single type of forest, which do not change the overall

forest type, but change the density of carbon within it.

The application of VCF values to forest classification maps permits a more realistic

view of carbon density than a simple classification. In an assessment of REDD

monitoring capabilities for Non-Annex 1 countries, Romijn et.al. [19] identify MODIS

management and VCF as one of the evaluation criteria to estimate country

capabilities. REDD+ monitoring requires a more nuanced and dynamic view of

forest cover which can be provided by VCF and MODIS. It is necessary to recognize

variations within existing forests as a baseline for documenting subtle differences in

forest quality resulting from both degradation processes, and mitigation efforts in

each individual country.

The high temporal density of the MODIS sensor provides a basis for the solution to

the problems posed by REDD+ monitoring (GOFC-GOLD [5]: 2-21). Daily data can

be composited into cloud ‘free’ images of periods appropriate for the climatic

conditions of the study area (Huttich et.al. [13]). The advantage of the compositing

strategy is that imagery is available throughout the annual cycle, providing a

balance to potential errors caused by phenological cycling when monitoring is

attempted with a single date image.

The method for MODIS data management cited in the Sourcebook is the VCF

methodology developed by academic researchers in coordination with the MODIS

team (DeFries et.al [2], Hansen et.al. [6]). This methodology is based on a series of 8

to 12 composite images throughout the year, and a set of metrics which highlight

different phases of the yearly phenological cycle, normalized for annual variations

in climate.

VCF coverage output has been adopted as a MODIS product, issued with the

designation MOD44B in conjunction with the University of Maryland. These data

have been updated completely first for 2001 through 2005, and more recently, for

2005 through 2010 at a pixel size of 250m. These represent very high quality

products produced through the application of the combined resources of the

Department of Geography, Earth System Science Interdisciplinary Center, and the

Institute for Advanced Computer Studies, of the University of Maryland.

The implementation of the GIZ-VCF Monitoring and Early Warning address the

needs of REDD forest cover monitoring. The goal of monitoring cannot be limited

Page 22: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

simply to the differentiation of forest classes, but must be able to provide

information on gradual spatial and temporal changes within classes, and the implied

GHG emissions changes. This goal is especially pertinent with regard to the above

referenced actions of “reducing emissions from forest degradation” and the

recapture of emissions through the “enhancement of carbon stocks”, outcomes to be

expected from the implementation of community based REDD projects. These goals,

highlighted and specifically identified in the first pages of the Sourcebook, require

an approach to monitoring which reports changes in forest density across time and

space.

3. METHODS AND DATA

MODIS - Characteristics and specifications

MODIS satellites were launched with the intention of providing global forest cover

information, to continue, and improve, on the information collected through the

earlier AVHRR program. The use of MODIS for national level forest monitoring

strikes a discordant note with national institutions accustomed to the use of

LANDSAT for monitoring purposes, in part because of the differences in pixel

resolution, and partly in response to the inability to interpret new information

generated by the MODIS data set. Nevertheless, through the use of composited

images based on daily data, MODIS provides significant improvement in satellite

data coverage, and introduces new opportunities for forest monitoring through its

increased temporal density.

The MODIS satellite constellation continued the global surface monitoring effort

initiated with AVHRR data to determine surface temperature conditions and

vegetation. AVHRR satellites monitored four, five and six bands of data in its

successive versions (NOAA [16]) which limited capabilities with regard to surface

cover monitoring. Beginning in 1981 AVHRR permitted the first continuous, global

view of vegetation condition with frequent updates, in a globally consistent format,

specifically through the use of NDVI at a 1000m resolution. The MODIS program

greatly expanded the available data by generating 36 bands including sources of

data for terrestrial, oceanic and atmospheric monitoring at a nominal resolution of a

250m, and incorporated as one of the primary products the generation of NDVI to

continue the sequence initiated by AVHRR (Zhan [25]).

An advantage of the lower MODIS resolution is the ability to observe all of Central

America and the Dominican Republic with only 4 images, easily viewable and

available through GLOVIS (USGS Global Visualization Viewer)(USGS [23]).

Preparation and processing time for the entire region is greatly reduced over a

similar undertaking with higher resolution images; processing is further facilitated

by the availability of preprocessed MODIS Data Products which include composite

images for different time periods. Of special interest is the 16 day composite

Page 23: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

MOD13Q1 which provides vegetation indices and 4 spectral bands at 250m

resolution (Huete et.al. [12]). The reduced data processing requirements permit the

generation of new data sets very quickly, bringing results much closer to ‘real time’

(less than 4 weeks from the end of the data period) than would be possible with a

higher resolution system.

Monthly composite files have MODIS quality indicators described by Huete et.al.

[12], which provide a basis for pixel quality evaluation. Sixteen day composite files

are compared with special emphasis on MODIS quality bits, replacing pixels when

there is a clear superiority of one over the other. Composite images of longer

duration (32 vs 16 days) have a higher likelihood of good quality pixels being found

for cloudy areas.

Focus on phenological cycle

The VCF methodology seeks to identify the density of forest cover. The

methodology calls for the development of a set of ‘metrics’ tied to maxima and

minima of band values throughout the course of a year.

The quantity of tree leaves in a pixel is a major component in the values reflected to

the satellite sensor. The quantity of leaves is not constant, since it varies according

to the stage of the phenological cycle; trees frequently shed leaves during summer

droughts, and replace when rainfall increases. In order to normalize the detection

of forest cover for annual variations in onset dates of rainy and dry periods, metrics

must be constructed which are tied to the phenological cycle, rather than to the

calendric cycle; rains may begin in March of one year, and May of another.

VCF metrics are constructed to detect periods of maxima and minima which reflect

stages in the phenological cycle (Hansen [6]). One set of statistics is developed at

the period of maximum NDVI, and another at the period of maximum temperature,

corresponding to periods of maximum leaf coverage, and minimum leaf coverage,

respectively. This approach minimizes the possibility of false reports of change

caused by the comparison of different stages in the phenological cycle in succeeding

years.

In addition to the maxima and minima for phenologically defined stages, VCF

metrics also capture the length and intensity of the dry season and the rainy season.

Multimonth averages are compared for phenological maxima and minima. A total of

86 metrics are collected, which are complemented by elevation and rainfall data.

Page 24: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

The VCF value produced is normalized for climatic variations throughout the year,

and by incorporating information of lengths of dry and wet seasons also tends to

normalize interannual differences.

Sequence Statistics

The analysis of forest change dynamics is carried out on the sequence of VCF layers

for the years 2001 through 2010. The analysis of change is carried out by a pixel by

pixel mathematical analysis, to determine characteristics of each pixel sequence.

Statistics collected on the pixel by pixel analysis are mean, standard deviation, and

slope, based on, in this case, the 10 yearly pixels from each corresponding layer.

Pixels are grouped by their behavior with respect to these statistics. Mean values

r

e

f

l

e

c

Figure 3 Distribution of Standard Deviations

Figure 1 Distribution of Means

Figure 2 Distribution of Slopes

Page 25: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

t a general level of forest cover, without providing information as to the stability of

the observed formations. A slope is calculated for each pixel set, to reflect the

general tendency of the forest cover, along with a standard deviation based on

residual values around the line of the calculated slope.

The histogram of means shows that he majority of land cover falls below 50% forest

density. In the 0 – 20% range abrupt changes may reflect distinct agroforestry

systems while denser forest systems show gradual changes in the histogram

reflecting the variety of forest types and distributions, with distinct peaks at 30%

and 90% forest cover.

Slopes of pixel data were calculated after eliminating null data (cases where no VCF

classification was achieved). Predictably, the distribution of slopes clusters around

0, which indicates stable land cover patterns, and a slight skew toward negative

slopes, indicating processes of degradation or deforestation.

Similar to the previous calculation, the standard deviation is calculated after

eliminating null values. The standard deviation identifies how much variability in

VCF values are seen over the decade. Sixty-nine percent of the non-zero standard

deviations (zero values only occur where there are no data) are greater than the

value of 10.

Data Anomalies and Forest Disturbance

MODIS VCF data generates many anomalies. Some are common to all kinds of

remotely sensed imagery, and others arise from the attempt to utilize the high

temporal density of the MODIS data set. A close analysis of these anomalies reveals

that many are ‘anomalies’ only from the perspective of LANDSAT type image

processing, and in fact present a new enriched perspective on forest dynamics in the

tropics.

Anomalies common to all remotely sensed data are typically atmospheric effects.

Figure 10 Detail of Fire Anomaly

Page 26: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

More notable are clouds, along with their deep shadows which produce values very

different from non-shadowed pixels of the same land cover. In this same category

fall haze, jet contrails, smoke, or any other phenomenon which decreases the

transparency of the atmosphere.

Anomalies specific to MODIS-VCF analysis highlight a new set of transitory

phenomena. The most striking anomaly is the periodic ‘disappearance’ of forest

cover, or its degradation, quickly followed by ‘reforestation’ in the space of one or

two years. Two mechanisms have been identified which create this anomaly.

The first, and most common cause of rapid forest disappearance and re-appearance

is fire. In Figure 4, three data sets are compared; the 2006 MODIS coverage of fire

occurrence for the entire year, the 2010 land use classification for Central America

carried out by CATHALAC (Centro de Agua para los Tropicos Humedos de

Latinoamerica y el Caribe), and a sequence of 10 annual VCF images from 2001 to

2010, all focused on central Honduras. The three images are geographically linked,

showing that the point of the anomaly, identified by the VCF profile in the bottom

center of the image (where each data point corresponds to a VCF value for a single

year), coincides with a pine forest which experienced burning over an area of some

500ha during 2006. In Honduras, forests are used as a pasture area for cattle, and

burning is a common practice to eliminate ticks, which create health problems for

the cattle. Burning also serves to ‘renew’ grasses, burning off old tough stems which

are then replaced with new growth. The intention of the burning is not land

clearance, but the destruction of dry grasses and pests which live in the vegetation.

This low intensity fire eliminates a great deal of the tree foliage, and also covers the

landscape with soot making the forest 'disappear'. In the following rainy season,

trees set new leaves, and soot is washed from the landscape, so the forest becomes

visible once again. Sixty-seven percent of locations of the 8129 fires in forest areas

for 2006 correspond to areas showing this anomaly during the period 2001-2010;

the remaining 33% of fires occurred in areas reported as stable forest in 2006.

Of the remaining anomalies, 21% are found in flood plains and river beds. In flood

periods, rising waters lead to decreased visibility of forest vegetation and reduce

the apparent forest cover for a period of days or weeks. VCF values during the

period may drop dramatically, only to rise again in the following year.

Other causes of brief ‘deforestation’ anomalies are landslides, river bed scouring,

high winds (e.g. hurricanes), droughts and possibly insect attacks. While there can

be no doubt that hidden within the deforestation anomalies are cases of erroneous

data, it is clear that ‘anomalous’ patterns of change reflect real phenomena which

have been unrecognized due to the use of single data imagery for land cover

evaluation. An awareness and understanding of the forces which cause forest

anomalies will contribute to a deeper understanding of forest processes, and help

provide realistic approaches to degradation mitigation.

Page 27: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

Stable Intervened and Non-Intervened Forests

The combination of the annual sequence statistics, mean, standard deviation and

slope, permit a classification of patterns of forest dynamics. Two patterns which

account for over 75% of all nationally identified forest aeas, are of special interest

(see Table 3). The patterns detected are:

- Stable, non-intervened forest; average forest cover above 60%, slope less than 3

(300 in the accompanying table) and a standard deviation less than 10.

- Stable, intervened forest; slope less than 3 and a standard deviation greater

than 10. Threatened forest.

Table 3: Forest dynamics as a percentage of total forest area, by country, 2011

Belize

Costa

Rica

República

Dominicana

El

Salvador Guatemala Honduras Nicaragua Panamá Total

1. Stable Intervened Forest 34.88% 53.23% 54.18% 64.31% 43.11% 51.07% 48.03% 52.73% 29.24%

2. Stable Non-intervened Forest 51.77% 26.56% 27.19% 20.97% 31.84% 33.36% 24.88% 20.69% 48.85%

3. Forest with constant reduction 8.02% 7.38% 6.28% 7.54% 15.51% 5.91% 13.70% 7.06% 9.68%

4. Forest with constant increase 1.37% 2.20% 5.50% 3.63% 1.64% 3.54% 2.31% 2.25% 2.56%

5. Highly intervened,

tendency toward

change 0.90% 5.09% 2.52% 0.89% 2.18% 2.18% 3.32% 8.98% 3.79%

6.Other 3.06% 5.55% 4.33% 2.66% 5.72% 3.94% 7.76% 8.29% 5.88%

Total

100.00% 100.00% 100.00% 100.00% 100.00% 100.00% 100.00% 100.00% 100.00%

From the perspective of community forest management, the category of stable,

intervened forest is of special interest, as these are forests which most likely are the

result of a long term presence of human management practices. (The exceptions to

this generalization regarding highly intervened forests without change tendencies

are swamps and riverbeds, where regular, nearly annual, flooding occurs to cause

periodic fluctuations in the apparent forest cover.) This observation explains the

distribution of forests with high and low standard deviations (see Figure 5): the

largest areas of low standard deviations are seen in the most remote forest areas,

the coastal area of the Honduran and Nicaraguan Moskitia, south-central Belize, and

the lowlands of Petén. Forests found near areas of intense human activity are more

likely to have high standard deviations, although there are many patches of non-

Page 28: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

intervened forest even in the more populous areas of the region.

Figure 11 Distribution of Stable Intervened and Non-Intervened Forests

It is important to draw the contrast between high and low standard deviation

forests. Low

standard deviation

forests suffer changes

in density, but at a

low level which

would be consistent

with changes in

rainfall totals,

occasional high winds

or hurricanes.

Apparent forest cover

does not change

dramatically. High

standard deviation

forests are defined

with standard

Figure 12 Detail of Distribution of Intervened Forests

Page 29: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

deviations greater than 25, representing changes of more than 25% on average from

the mean tendency of forest cover; logically, some years will show more than 25%

change to reach an average of 25%. High standard deviation forests experience

intense human intervention, through the near complete loss of cover on a regular

basis; although these forests are intensely managed, and the objective is clearly not

one of deforestation.

A higher resolution view of the Guatemalan highlands (Figure 6), where there is a

tradition of community forest management, shows a complex pattern of intervened

and non-intervened forests. The linear features running northwest to southeast

emphasize the intensive geological folding which generated the irregular

topography of the region.

Early Warning

Early warning is the implicit goal of a monitoring system. With a delay of 4-6 weeks

in the presentation of monitoring results, the VCF based monitoring system is not

the ideal early warning platform; nevertheless, it has a much quicker response time

than forest inventory or forest monitoring systems in the past, especially in

identifying where most recent changes are ongoing across a large landscape.

The VCF based monitoring system is built on the analysis of year-long image

sequences, to capture phenological information, and distinguish when changes vary

from normal fluctuation. The proper operation of an early warning system is

plagued by ‘false positives’ of degradation, caused as a result of ‘normal’ decreases

in forest cover. Phenological variation makes it impossible to absolutely identify

degradation events in the short term, since forests may suffer temporary decreases

in leaf cover one or more times a year, which are indistinguishable in their early

stages from long term degradation events.

Even more challenging than false positives caused by the phenological cycle are fire

events such as those described above. A fire-caused decrease in tree cover might

represent the beginning of a deforestation or degradation process, or it may be one

more cycle in a homeostatic management cycle which temporarily decreases

vegetation as a means to improve the quantity and quality of vegetation over a

longer period. The final conclusion regarding the impact of a vegetation decrease

event may take years to define, especially when it occurs in the context of this kind

of forest management.

To reduce the probabilities of incorrect identification of degradation events, early

warning analysis is only carried out in stable forests, with low standard deviations

in VCF values. Sudden decreases in forest cover outside the normal range of

variability most likely represents a degradation event. Since early warning is

calculated on the basis of mean VCF values, forest with constantly changing VCF's,

Page 30: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

even with low standard deviations, cannot be included because the overall trend

tends to obscure variations from the mean.

4. RESULTS AND DISCUSSION

VCF analysis provides increased depth to the evaluation of forest trends in Central

America and Dominican Republic. Within the optimistic statistic of a decrease in

deforested areas in the decade 2010-2011 (Table 4), a trend toward degradation of

primary forests can be distinguished, and an increase in forests suffering regular

human intervention.

Table 4 shows that the only category of forest to increase in the 2010-2011 was

stable intervened forest, while all other categories, including deforested land,

decreased.

Table 4: Type of forest dynamic %2010 %2011

Stable mean >= 60 and sdev <= 10 and abs(slope) <= 300 29.32% 27.03%

Stable intervened sdev > 10 and abs(b2)<= 300 44.96% 48.73%

Constant tendency to deforestation sdev <= 30 and slope<= -300

1.17% 0.73%

Stable agroforestry : mean >= 60 and mean >= 30 and sdev <= 10 and abs(slope)<= 300

1.37% 1.26%

Legally deforested <30% : mean < 30 23.18% 22.25%

TOTAL 100% 100%

Table 5: Early Warning by Trimesters 2010 - 2011

EARLY WARNING

Year 2010 Year 2011

I II III IV I II III IV

Total Warnings

24635 25095 29472 36963 17058 29733 36962 23182

New Warnings

24635 16236 17112 32440 4315 22202 20244 13419

Accumulated total

24635 40871 57983 90423 94738 116940 137184 150603

Page 31: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

In the category of stable forests, trimestral reports of early warnings accumulate to

a total of more than 8000km (Table 5). The method for constructing the trimestral

warnings was to compare each trimester with the previous calendar year; each

trimester of 2010 was compared to trends through 2009. This method allowed

early persistent warnings to be repeated, if VCF values did not recover during the

trimester. The figures for each trimester are broken down by total warnings, total

warnings for the trimester which do not duplicate earlier warnings, and the

accumulated non-duplicated warnings over 8 trimesters, in pixels (figure 4).

Figure 13 Loss of Primary Forest

warnings

Area accumulated warnings (km2)

1314.55

2180.92

3094.03

4825.06

5055.31

6240.04

7320.28

8036.33

Area new warnings (km2)

1314.55 866.37 913.11 1731.03 230.25 1184.72 1080.24 716.05

Page 32: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

-

Figure 14 Cumulative Early Warning 2010 - 2011

The comparison of stable forest and early warnings gains credibility with the

accumulation of data. The same data appears in Figure 8, with stable forest for

March 2010 presented in green; the black area of the image is largely covered by

intervened forest, but indicative of human activity. This image shows Belize and its

coastline on the right, including the area of Belize City and Belmopan, and extends

into Guatemala on the left. Accumulated trimestral early warnings from June 2010

to December 2011 are presented in red, and clearly define a swath across Belize

being cleared as part of a rice expansion program supported by the government. A

large number of scattered early warnings appear throughout the image, indicating a

process of more individualized, and possibly clandestine forest degradations; since

early warnings are defined as a decline of 25% forest cover, these same areas may

simply be degraded, but remain under forest.

5. CONCLUSION

A VCF based regional monitoring system for REDD reporting contains tools to address changes (or non-changes) in forest cover required for complete Green House Gas reporting: "reducing emissions from deforestation (which implies a land-use change) and reducing emissions from forest degradation, conservation of forest carbon stocks, sustainable management of forest land, Enhancement of forest carbon stocks". Periodic updates of the percent forest cover image for the region are a basis for the observation of degradation or enhancement trends in forest cover. A breakdown of statistical patterns based on pixel by pixel multiyear trends identifies two broad forest categories; dense forest with little interannual variability in density values, and forests with high interannual variability.

Page 33: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

The low variability forest represents non-intervened forest experiencing normal variations in density due to natural conditions. The stability of the VCF values permit the application of an Early Warning analysis to this type of forest, identifying moments when forest degradation begins, but requiring ground checking or more detailed photographic analysis to determine if this is an anthropogenic effect or a natural process. The latter category of high variability represents intervened, degraded forest, which can be evaluated in terms of average forest density to identify trends toward improved or degenerating forest cover. As noted by Herold [11], carbon loss through degradation may represent up to 10 times the carbon loss from simple deforestation, due to the large areas affected. In Central America and the Dominican Republic, stable natural forests represent only 20% of the regional forest area, while intervened forests represent nearly 50%. The ability to monitor levels of forest degradation, or the evolution of forest density in agroforestry systems, provides a necessary element in the compensation of broad community based forest management efforts. Changes in density can be used as a complement to ground point controls, to demonstrate the effectiveness of community management activities. In view of the potentially large area affected by this type of management, VCF monitoring of carbon concentrations in highly intervened forests stands to make very large contributions to the documentation of carbon sequestration efforts. ACKNOWLEDGMENT The development of the Central American MODIS Monitoring System was financed by Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República Dominicana (REDD - CCAD - GIZ: http://www.reddccadgiz.org/) from 2010 through 2012. REFERENCES 1. Coops, Nicholas C., Michael A. Wulder, Donald Iwanicka. 2009. "Large area

monitoring with a MODIS-based Disturbance Index (DI) sensitive to annual and seasonal variations". Remote Sensing of Environment 113 (2009) 1250–1261.

2. DeFries, R. S., Hansen, M. C., Townshend, J. R. G., and Janetos, A. C., 2000, A

new global 1-km dataset for percentage tree cover derived from remote sensing. Global Change Biology 6, 247-254.

3. ECLAC 2009. Latin America and the Caribbean Demographic Observatory. United Nations. Santiago, Chile.

4. FAO, 2010. Evaluación de los recursos forestales mundiales 2010. Informe principal. Roma, Italia.

5. GOFC-GOLD, 2011. A sourcebook of methods and procedures for monitoring and reporting anthropogenic greenhouse gas emissions and removals caused by deforestation, gains and losses of carbon stocks in forests remaining forests, and

Page 34: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

forestation. GOFC-GOLD Report version COP17-1, (GOFC-GOLD Project Office, Natural Resources Canada, Alberta, Canada).

6. Hansen, M.C., R.S. DeFries, J.R.G. Townsend, R. Sohlberg, C. Dimiceli, M. Carroll. 2002. “Towards an operational MODIS continuous field of percent tree cover algorithm: examples using AVHRR and MODIS data”. In Remote Sensing of Environment 83: 303 – 319.

7. Hansen, Matthew C. Yosio E. Shimabukuro, Peter Potapov, Kyle Pittman. 2008a. "Comparing annual MODIS and PRODES forest cover change data for advancing monitoring of Brazilian forest cover". Remote Sensing of Environment 112 (2008) 3784–3793.

8. Hansen, Matthew C., Stephen V. Stehman, Peter V. Potapov, Thomas R. Loveland, John R. G. Townshend, Ruth S. DeFries, Kyle W. Pittman, Belinda Arunarwati, Fred Stolle, Marc K. Steininger, Mark Carroll, and Charlene DiMiceli. 2008b. "Humid tropical forest clearing from 2000 to 2005 quantified by using multitemporal and multiresolution remotely sensed data " PNAS , July 8, 2008, vol. 105 , no. 27: 9439–9444.

9. Hansen, Matthew C., David P. Roy, Erik Lindquist, Bernard Adusei, Christopher O. Justice, Alice Altstatt. 2008c. " A method for integrating MODIS and Landsat data for systematic monitoring of forest cover and change in the Congo Basin ". Remote Sensing of Environment 112 (2008) 2495–2513.

10. Heckadon, Stanley. 1981. Dinamica social de la cultura del potrero. RENARE

(Ministerio de Recursos Naturales Renovables, Panamá), Panamá.

11. Herold, Martin, Rosa María Román-Cuesta, Danilo Mollicone, Yasumasa Hirata, Patrick Van Laake, Gregory P Asner, Carlos Souza, Margaret Skutsch, Valerio Avitabile and Ken MacDicken. 2011. " Options for monitoring and estimating historical carbon emissions from forest degradation in the context of REDD+". Carbon Balance and Management 2011, 6:13.

12. Huete, Alfredo, Chris Justice, Wim Van Leeuwen. 1999. MODIS VEGETATION INDEX (MOD 13) ALGORITHM THEORETICAL BASIS: DOCUMENT Version 3. Tucson, Arizona.

13. Hüttich, Christian, Martin Herold, Martin Wegmann, Anna Cord, Ben Strohbach Christiane Schmullius, Stefan Dech. 2011. “Assessing effects of temporal compositing and varying observation periods for large-area land-cover mapping in semi-arid ecosystems: Implications for global monitoring”. In Remote Sensing of the Environment 115: 2445 – 2459.

Page 35: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

14. Jones, Jeffrey. 1990. Colonization and Environment: Land Settlement Projects in

Central America. United Nations University Press, Tokyo.

15. Mildrexler, D. J., Zhao, M. S., Heinsch, F. A., & Running, S. W. 2007 . A new satellite based methodology for continental-scale disturbance detection. Ecological Applications, 17, 235−250.

16. NOAA 2012. NOAAA Satellite Information Service

(http://noaasis.noaa.gov/NOAASIS/ml/avhrr.html)

17. Pan, Yude, Richard A. Birdsey, Jingyun Fang, Richard Houghton, Pekka E. Kauppi, Werner A. Kurz, Oliver L. Phillips, Anatoly Shvidenko, Simon L. Lewis, Josep G. Canadell, Philippe Ciais, Robert B. Jackson, Stephen W. Pacala, A. David McGuire, Shilong Piao, Aapo Rautiainen, Stephen Sitch, Daniel Hayes. 2011. "A Large and Persistent Carbon Sink in the World’s Forests". SCIENCE, VOL 333, 19 AUGUST 2011 : 988 - 993.

18. Rapidel, Bruno, Olivier Roupsard, Jean-François Le Coq and Muriel Navarro. 2009. Modelling Agroforestry Systems : Workshop Proceedings . CATIE, Costa Rica, 25–29 February 2008. CATIE, Serie Tecnica: Reuniones Tecnicas #14. 2009.

19. Romijn, Erika, Martin Herold, Lammert Kooistra, Daniel Murdiyarso, Louis Verchot. 2012. “Assessing capacities of non-Annex I countries for national forest monitoring in the context of REDD+”. In Environmental Science and Policy 19-20, 2012: 33-48.

20. Saatchi, Sassan S., Nancy L. Harris, Sandra Brown, Michael Lefsky, Edward T. A. Mitchard, William Salas, Brian R. Zutta, Wolfgang Buermann, Simon L. Lewis, Stephen Hagen, Silvia Petrovac, Lee Whiteh, Miles Silmani, and Alexandra Morelj. 2011. “Benchmark map of forest carbon stocks in tropical regions across three continents”. In PNAS June 14, 2011, Vol 108 #24: 9899 – 9904.

21. Sewastynowycz, James. 1986. “Two Step Migration and Upward Mobility on the

Frontier. Economic Development and Cultural Change 34(4): 731-753l.

22. The Global Forest Disturbance Alert System (GloF-DAS):

http://rainforests.mongabay.com/deforestation-tracker/

23. USGS ND. Global Visualization Viewer (GloVis). http://glovis.usgs.gov/

24. Wulder, Michael A., Joanne C. White, Mark D. Gillis, Nick Walsworth, Matthew

C. Hansen, Peter Potapov. 2010. "Multiscale satellite and spatial information and analysis framework in support of a large-area forest monitoring and inventory update". Environmental Monitoring and Assessment (2010) 170:417–433.

Page 36: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

25. Zhan, X. , R. DEFRIES, J. R. G. TOWNSHEND, C. DIMICELI, M. HANSEN, C.

HUANG and R. SOHLBERG. 2000. The 250m global land cover change product from the Moderate Resolution Imaging Spectroradiometer of NASA’s Earth Observing System. Int. J. Remote Sensing, 2000, vol. 21, no. 6 & 7, 1433–1460.

Page 37: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

Anexo 2: Descripción de la edición especial para artículo:

IntJForResRMDC22march12AM.pdf

Page 38: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

International Journal of Forestry Research

Special Issue on REDD+ Mechanism in Developing Countries

Call for Papers Deforestation and forest degradation, through agricultural and animal husbandry activities,

infrastructure development, destructive logging, fires etc., account for nearly 6-17% of global

greenhouse gas emissions; in some years these emissions represent more than the entire

global transportation sector and are second only to the energy sector. It is now well

established that it will be essentially impossible to stabilize the climate within two degrees

Celsius without reducing emissions from deforestation and forest degradation. This reduction

must be in addition to increased Annex I countries’ mitigation pledges before 2020, which

are absolutely necessary in order to avoid the worst impacts of anthropogenic interference

with the climate system. Since the REDD (Reducing Emissions from Deforestation and

Forest Degradation) action plan was adopted at the 13th Conference of the Parties of the

United Nations Framework Convention on Climate Change (UNFCCC) (COP13) on 8

December 2007, in Bali, Indonesia, its scope has expanded to include not only forest

conservation activities, but also enhanced and sustainable forest management activities (the

so-called REDD+ agenda). REDD+ is a mechanism being negotiated through the UNFCCC

for mitigating climate change, reducing rural poverty and achieving conservation and

sustainable development of tropical forests. REDD+ promotes co-benefits such as non-

carbon benefits, ecosystem services, employment, livelihood, climate change adaptation, and

cultural services, depending on the social, ecological and institutional context in which it is

implemented.

We invite authors to submit original research and review articles that seek to define the

interaction between REDD+ and activities leading to community development. We are

interested in articles that explore all aspects of REDD+ as a mechanism for reducing rural

poverty and achieving conservation and sustainable management and development of tropical

forests. Potential topics include, but are not limited to: REDD+ and community forestry

REDD+, forest conservation and enhanced forest management

REDD+, advanced technologies and tools for measurement of carbon sequestration

REDD+, valorisation of timber and non-timber forest products for community sustainability and bioeconomy

REDD+, exchange of experience and strategies between states

REDD+, institutional and governance issues

REDD+ and financing mechanisms

REDD+ and capacity building

Before submission, authors should carefully read over the journal’s Author Guidelines, which

are located at http://www.hindawi.com/journals/ijfr/guidelines.html. Prospective authors

should submit an electronic copy of their complete manuscript through the journal

Manuscript Tracking System at http://mts.hindawi.com/ according to the following timetable:

Manuscript Due

Manuscript Due August 3, 2012 Manuscript Due

First Round of Reviews October 26, 2012 First Round of Reviews Publication Date December 21, 2012 Publication Date

Page 39: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

Lead Guest Editor

Damase Khasa, Centre for Forest Research, Canada Research Chair in Forest and

Environmental Genomics and Institute of Systems and Integrative Biology, Canada;

[email protected]

Guest Editors Alison Munson, Centre for Forest Research, Université Laval, Québec, Canada;

[email protected]

Mariteuw Chimère Diaw, African Model Forest Network, African Model Forests Network

Secretariat, Cameroon; [email protected]

Nancy Gelinas, Department of Wood and Forest Sciences, Université Laval, Québec,

Canada; [email protected]

Dr. Nadine Laporte, The Woods Hole Research Center, USA ; [email protected]

Dr. Glenn Bush, The Woods Hole Research Center, 149 Woods Hole Road Falmouth, MA

02540–USA ; [email protected]

Page 40: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

Anexo 3: Recibido de articulo de editores de revisa (correo electronico)

Page 41: Programa Reducción de Emisiones de la Deforestación y … · 2019-09-29 · Programa Reducción de Emisiones de la Deforestación y Degradación de Bosques en Centroamérica y República

International Journal of Forestry Research [email protected] via amazonses.com

Aug 2

to me, abner.jimenez, laszlo.pancel

Dear Dr. Jones, The Research Article titled "MODIS Based REDD+ Regional Monitoring System for Forest Cover Reporting of Central America and the Dominican Republic," by Abner Josue Jimenez Galo, Jeffrey Jones and Laszlo Pancel has been received and assigned the number 805645. The special issue for which the paper is being processed is "REDD+ Mechanism in Developing Countries" An editor will be assigned to handle the review process of your manuscript, and he/she will inform you as soon as a decision is reached. All authors will receive a copy of all the correspondences regarding this manuscript. However, only the submitting author will be able to upload any revisions to the journal's Manuscript Tracking System. Thank you for submitting your work to International Journal of Forestry Research. Best regards, Fathya Sayed Editorial Office Hindawi Publishing Corporation http://www.hindawi.com