Opengl 3.3+ incorrect shadows when using shadow mapping

Question

I am trying to implement shadow mapping on my landscape editor with OpenGL 3.3+. Using a few tutorials I have managed to get my code to compile and run but the whole landscape is in shadow except for the back row of my landscape grid (smallest z).

I am currently using the same projection, view and model matrices for my light as the camera (negative z is furthest from the camera).

Initialisation of my projection, view and model matrices (from LWJGL matrix tutorial):

modelPos = new Vector3f(0f, 0f, -20f);
modelAngle = new Vector3f(15f, 0f, 0f);
modelScale = new Vector3f(1f, 1f, 1f);
cameraPos = new Vector3f(-50f, 0f, -120f);

projectionMatrix = new Matrix4f();

float fieldOfView = 120f;
float aspectRatio = (float)width / (float)height;
float near_plane = 0.01f;
float far_plane = 100f;

float y_scale = DepthMatrixUtility.coTangent(DepthMatrixUtility.degreesToRadians(fieldOfView / 2f));
float x_scale = y_scale / aspectRatio;
float frustum_length = far_plane - near_plane;

projectionMatrix.m00 = x_scale;
projectionMatrix.m11 = y_scale;
projectionMatrix.m22 = -((far_plane + near_plane) / frustum_length);
projectionMatrix.m23 = -1;
projectionMatrix.m32 = -((2 * near_plane * far_plane) / frustum_length);

Binding my matrices when displaying scene:

Matrix4f.translate(cameraPos, viewMatrix, viewMatrix);
Matrix4f.scale(modelScale, modelMatrix, modelMatrix);
Matrix4f.translate(modelPos, modelMatrix, modelMatrix);

Matrix4f.rotate(DepthMatrixUtility.degreesToRadians(modelAngle.z), new Vector3f(0, 0, 1), modelMatrix, modelMatrix);
Matrix4f.rotate(DepthMatrixUtility.degreesToRadians(modelAngle.y), new Vector3f(0, 1, 0), modelMatrix, modelMatrix);
Matrix4f.rotate(DepthMatrixUtility.degreesToRadians(modelAngle.x), new Vector3f(1, 0, 0), modelMatrix, modelMatrix);

matrix = new Matrix4f();
Matrix4f.mul(matrix, projectionMatrix, matrix);
Matrix4f.mul(matrix, viewMatrix, matrix);
Matrix4f.mul(matrix, modelMatrix, matrix);

matrix.store(matrix44Buffer);
matrix44Buffer.flip();

matrixLocation = GL20.glGetUniformLocation(pId, "matrix");
GL20.glUniformMatrix4(matrixLocation, false, matrix44Buffer);

I have tested my FBO with storing colour in the fragment shader, the height map displays correctly (I drew the FBO texture to a small quad in the corner of my screen) and updates as I alter the height map.

I then modified my FBO to store the depth to a texture on the first pass:

depthTexture = GL11.glGenTextures();

GL11.glBindTexture(GL11.GL_TEXTURE_2D, depthTexture);

GL11.glTexImage2D(GL11.GL_TEXTURE_2D, 0, GL11.GL_DEPTH_COMPONENT, Window.getScreenWidth(), Window.getScreenHeight(), 0, GL11.GL_DEPTH_COMPONENT, GL11.GL_UNSIGNED_BYTE, (ByteBuffer)null);

GL11.glTexParameterf(GL11.GL_TEXTURE_2D, GL11.GL_TEXTURE_WRAP_S, GL11.GL_NEAREST);
GL11.glTexParameterf(GL11.GL_TEXTURE_2D, GL11.GL_TEXTURE_WRAP_T, GL11.GL_NEAREST);
GL11.glTexParameteri(GL11.GL_TEXTURE_2D, GL11.GL_TEXTURE_MAG_FILTER, GL11.GL_LINEAR);
GL11.glTexParameteri(GL11.GL_TEXTURE_2D, GL11.GL_TEXTURE_MIN_FILTER, GL11.GL_LINEAR);

GL11.glBindTexture(GL11.GL_TEXTURE_2D, 0);

fboId = GL30.glGenFramebuffers();
GL30.glBindFramebuffer(GL30.GL_FRAMEBUFFER, fboId);

GL11.glDrawBuffer(GL11.GL_NONE);
GL11.glReadBuffer(GL11.GL_NONE);

GL32.glFramebufferTexture(GL30.GL_FRAMEBUFFER, GL30.GL_DEPTH_ATTACHMENT, depthTexture, 0);

verifyFBO();

My vertex shader for the first pass (Creating the shadow map):

#version 330 core

uniform mat4 matrix;

in vec4 in_Position;

void main(void)
{
gl_Position = matrix * in_Position;
}

My fragment shader for the first pass:

#version 330 core

layout(location = 0) out float fragmentdepth;

void main(void)
{
fragmentdepth = gl_FragCoord.z;
}

My bias matrix:

[0.5f, 0.0f, 0.0f, 0.0f]
[0.0f, 0.5f, 0.0f, 0.0f]
[0.0f, 0.0f, 0.5f, 0.0f]
[0.5f, 0.5f, 0.5f, 1.0f]

My vertex shader for the second pass (rendering the scene using the shadow map):

void main(void)
{
    gl_Position = matrix * in_Position;
    ShadowCoord = biasMatrix * lightMatrix * in_Position;
}

My fragment shader for the second pass:

if (texture(shadowMap, ShadowCoord.xy).z  <  ShadowCoord.z)
{
    vec4 colour = 0.5 * out_Colour;
    out_Colour = new vec4(colour[0], colour[1], colour[2], 1.0f);
}

Answer 1

After transforming in_Position with lightMatrix, the result is not projected on screen yet.

Actual perspective projection is applied by dividing by w component.

The perspective division will give you texture coordinates and depth in [-1,1] range. At this point you use biasMatrix to transform them to [0,1] range.

So you shouldn't multiply by biasMatrix, then in your shader before the line

if (texture(shadowMap, ShadowCoord.xy).z  <  ShadowCoord.z)

add

ShadowCoord.xyz /= ShadowCoord.w;
ShadowCoord = biasMatrix * ShadowCoord;

The biasMatrix content you're showing should be stored transposed in memory. If you hesitate replace the matrix product with

ShadowCoord.xyz = ShadowCoord.xyz * .5f + float3(.5f);